Topics addressed in this article

Natalizumab is an effective therapy for multiple sclerosis. In rare cases, it is too potent in blocking entry of T cells into the CNS. Natalizumab also changes the ecology of immune cells in the brain. This suggests new principles about multiple sclerosis therapy, T cell/endothelial cell interactions, CNS viral infections, and CNS immunology.

Natalizumab is a human IgG4 kappa monoclonal antibody with short complementarity determining region segments of a potent murine anti-VLA-4 antibody inserted at the antigen recognition site. The IgG4 isotype does not bind complement and persists for a long time in the circulation. Natalizumab binds to the alpha4-chain of VLA-4 glycoprotein (alpha4,beta1 integrin; “very late activation 4” antigen) found on immune cell membranes and also affects immune cell activation. Antibody-bound VLA-4 no longer interacts with vascular cell adhesion molecule 1 (VCAM-1) on endothelial cells to prevent adhesion. After an impressively short trip from lab bench to bedside, natalizumab was approved by the U.S. Food and Drug Administration in 2004 for the treatment of relapsing-remitting forms of multiple sclerosis. It is also effective in moderately to severely active Crohn disease and possibly in rheumatoid arthritis.

Topics 1 and 2: VLA-4 and immune cell adhesion in many organs

(1) VLA-4 affects lymphocyte adhesion, migration, and organ development.

Adhesion molecules allow immune cells to attach to other immune cells and to endothelial cells. Important adhesion interactions include lymphocyte function associated antigen (LFA-1) attaching to intercellular adhesion molecule (ICAM-1), and very late activation antigen-4 (VLA-4) binding to VCAM-1.

VLA-4 is present on chronically activated T cells, and on B cells, monocytes, eosinophils, and basophils, but not on polymorphonuclear neutrophils (PMNs). VLA-4 is an alpha4beta1 integrin (CD49d plus CD29 subunits) that binds to vascular cell adhesion molecule (VCAM-1), mucosal addressin CAM (MAdCAM), fibronectin, osteopontin, and thrombospondin (19). These ligands are present on vascular endothelial cells, including cells of the blood-brain barrier, on macrophages and microglia, in gut and genital mucosa, and in the extracellular matrix of the brain (astrocytes and neurons). VLA-4 ligands induce neurite outgrowth (113). Blocking VLA-4 in the brain could theoretically have adverse consequences on developing or regenerating neurons.

A related integrin, alpha4beta7 (LPAM-1), binds MAdCAM and fibronectin. VLA-4 has slightly lower affinity for MAdCAM than for VCAM. MAdCAM is a mucosal and CNS addressin; it directs immune cells to the gut and brain. In Crohn disease, the VLA-4 and MAdCAM interaction enhances migration into the gut of T cells and especially Th17 cells, which express more beta7 than Th1 cells. PMNs do not use alpha4 integrins for migration, suggesting that VLA-4 blockers could have less effect in IL-17/PMN-mediated diseases such as Crohn disease, the East Asian form of multiple sclerosis, and neuromyelitis optica. VLA-4/MAdCAM is important for T cell migration in chronic experimental allergic encephalomyelitis and possibly in chronic multiple sclerosis.

Four steps allow immune cells to adhere to vascular endothelium:

(1) Short-term “tethering” slows the rapid movement of lymphocytes through the post-capillary venules. Glycoprotein L-selectin on leukocytes binds to E- and P-selectins on endothelial cells to slow the cells. L-selectin expression is low on immune cells in some multiple sclerosis patients (topic 16).

(2) “Rolling” is weak and transient and further slows immune cells. In addition to L-selectin, P-selectin, and PNAd on T cells that bind to endothelial cells, alpha4beta1 and alpha4beta7 on T cells bind to VCAM and MAdCAM on endothelial cells. PSGL1/CLA, on Th1 cells, but not on Th2 cells, is increased in multiple sclerosis and binds to E- and P-selectins on endothelial cells (215). The plasma membrane is 5 to 10 nm thick, and the extended form of the integrin extracellular domain is approximately 15 nm, forming a significant net for adhesion (121).

(3) “Activation” changes the conformation of VLA-4, causing clustering of these adhesion molecules, and transforms them to a high-affinity state on lamellipodia or filopodia at the leading edge of the immune cell (121). CXCR4 chemokine receptors and intracellular Rap1 co-localize in filopodia and aid in structural activation of VLA-4.

(4) “Strong adhesion” is necessary for penetration through the vessel wall. Here, VLA-4 binds to VCAM-1 or MAdCAM, and LFA-1 binds to ICAM-1, leading to “outside-in activation.” Natalizumab blocks strong stable adhesion but not the initial steps of contact, rolling, or capture.

In step 3 of the adhesion sequence, VLA-4 is activated, and the molecule extends. This molecule opens like a jackknife, the VLA-4 tails separate in the membrane, and adhesion increases 100-fold (54).

VLA-4 affinity is increased by mechanical stretching during adhesion, by reducing agents (cellular isomerases that break disulfide bonds), and by Ca++ and Mn++ ions (“outside-in” activation). In culture, Mn++ causes 8-fold more adhesion of human T cells to human umbilical vein endothelial cells and pericytes (282).

Affinity, and not simply VLA-4 expression, determines adhesion and migration. Immune cell subsets express VLA-4 of various affinities, so simple quantitation of VLA-4 expression may not reflect the ability to migrate. In the blood, VLA-4 is expressed in a low affinity conformation by most cells, including almost all T cells; however, VLA-4 is in an active conformation on some B cells, NK cells, and monocytes (Table 1). With activation, VLA-4 on T and B cells becomes high affinity, even though the amount of VLA-4 does not change (120). Activation with phorbol esters increases VLA-4 affinity on human memory T cells but not on naïve T cells (235).

There are multiple activation sites on the extracellular portion of the VLA-4 molecule. VLA-4 is activated from “inside-out” by intracellular second messengers that are induced by phorbol esters, the T and B cell receptors (ie, activated immune cells), prolactin, complement, LTB4, membrane-bound chemokines (acting through G-protein coupled receptors, GPCR), and fever. Soluble chemokines do not cause extension of the VLA-4 molecule. However, endothelium-expressed chemokines activate RhoA and cause a nearly instantaneous extension of LFA-1, which is then immediately activated by adjacent ICAM-1 ligand (254). Some anti-VLA-4 antibodies (eg, TS2/16) are even more potent than these physiological activators (54). Mn++ plus an activating MAb (TS2/16) increase binding affinity 1000-fold (51).

In contrast, pentoxifylline decreases integrin affinity and inhibits integrin-mediated adherence to activated endothelial cells (154). Pentoxifylline blocks CXCR12 chemokine-activated G proteins that activate Rap1 and redistribute and activate VLA-4 (121). Pentoxifylline and other cAMP inducers (beta-adrenergics, prostaglandins) could potentiate the effects of anti-VLA-4 antibodies.

Table 1. Cells in Mouse Blood and Lymphoid Organs Express High Affinity VLA-4

CNS vascular endothelium and peripheral endothelium differ in adhesion properties. The blood-brain barrier endothelium is highly specialized and less permeable to solutes and immune cells compared to peripheral endothelium. Secondly, there is little selectin on CNS vessels, although selectin has high expression in the choroid plexus. Consequently, no lymphocyte rolling is seen in the spinal cord using window preparations that visualize the microcirculation. Instead, encephalitogenic T cell blasts interact almost exclusively with VCAM-1 (280). Thus, blockade of VLA-4 adhesion will have profound effects on immune cell homing to the CNS.

Multiple sclerosis microvessels differ from those in normal brain. Endothelial LFA-1 assists T cells in migrating through multiple sclerosis vessels in a process that takes 3 to 8 hours. Microvessels with activated endothelial cells are linked to activated T cells in a positive feedback loop.

After breaching the blood-brain barrier, T cells and monocytes pause in the brain perivascular Virchow-Robin space. Only a few T cells will enter the parenchyma, and these are usually activated antigen-specific encephalitogenic cells (0.01 cells per gram of tissue per 100 injected cells, 100-fold less than in spleen or lung) (80). They persist in the perivascular space if they recognize brain antigens expressed on antigen-presenting cells. VLA-4 on T cells also interacts with VCAM-1 on pericytes, in addition to the endothelial cells.

The choroid plexus and meninges form a different type of barrier. P-selectin in veins of these structures allows central-memory T cells to enter the human brain and perform immunosurveillance. This multistep process contrasts with the rapid adhesion of activated anti-brain T cells to parenchymal endothelium through VLA-4/VCAM-1 and, to some extent, LFA-1/ICAM-1 (80).

(2) The role of VLA-4 and integrins in adhesion between cells in the bone marrow and thymus and effects of the sympathetic nervous system on adhesion.

Monocytes, NK and B cells, and especially immature B cells just released from the bone marrow, use VLA-4 adhesion to traffic into the CNS. T cells need to be activated and then express “very late activation 4” antigen to penetrate into the CNS.

Conversely, interactions between VLA-4 and VCAM-1, and between alpha4beta7 and MAdCAM-1, control the exit of developing T cells from the thymus and the exit of B cells from bone marrow (136). High affinity VLA-4 retains lymphocytes in thymus and bone marrow. With immune cell maturation, VLA-4 affinity and expression are reduced, and the cells migrate to the periphery. Antibodies to VLA-4 mobilize hematopoietic progenitor egress from the bone marrow by 200-fold within 72 hours of infusion (208).

Anti-VLA-4 antibodies are even more potent than granulocyte-colony stimulating factor (G-CSF) in causing release of CD34 stem cells from bone marrow (35; 303). Natalizumab also prevents blood CD34 cells from migrating back to the bone marrow (247). Natalizumab-induced CD34 cells have lower levels of adhesion molecules and secrete less matrix metalloprotease-9 than G-CSF-mobilized hematopoietic stem cells (128). Together, G-CSF and anti-VLA-4 antibodies synergize to increase mobilization of stem cells by a further 10-fold. This synergy is important if natalizumab is to be combined with chemotherapy, bone marrow transplantation, or possibly erythropoietin. Drug interactions that affect stem cell mobilization are important because CD34 stem cells are reservoirs for the JC virus that causes PML.

Granulocyte colony-stimulating factor (G-CSF) acts through the sympathetic nervous system to loosen the embrace of stem cells by osteoblasts (135). This myeloid cytokine upregulates dopamine receptors and beta2-adrenergic receptors on immature CD34 cells and increases their motility, proliferation, and secretion of matrix metalloproteases (263). It also generates type 2 dendritic cells that enhance development of Th2 cells. Of note, CD34 stem cells are mobilized into the blood within 2 days of an ischemic stroke.

Genes involved in B cell activation and differentiation are upregulated by natalizumab therapy (167). Natalizumab also elevates basophils, eosinophils, lymphocytes, and reticulocytes (227).

Natalizumab’s ability to increase peripheral white blood cells and release stem cells from bone marrow is reminiscent of a stress response influenced by the sympathetic nervous system that innervates bone marrow. The sympathetic nervous system releases vasoactive intestinal polypeptide, neuropeptide Y, and norepinephrine, which reduces CXCL12 on osteoblasts. Low CXCL12 causes bone loss but also allows hematopoietic stem and progenitor cells to be mobilized from their usual close association with osteoblasts. (CXCL12 increases the affinity of VLA-4 and, thus, increases retention.) Beta-adrenergic agonists enhance stem cell mobilization; beta blockers inhibit mobilization. There is strong evidence for sympathetic nervous system disruption in multiple sclerosis, presumably caused by hypothalamic and spinal cord plaques (133). This disruption causes denervation hypersensitivity, with increased expression of beta2-adrenergic receptors on immune cells (132), and possibly on CD34 cells, osteoblasts, and osteoclasts.

Sympathetic input inhibits osteoblasts and causes osteopenia. Osteoclasts produce chemokines and present antigen to CD8 cells; they also induce suppressor and regulatory CD8 cells (143). Multiple sclerosis patients have low bone density, disrupted sympathetic outflow, and low regulatory CD8 T cells. It is likely that the bone marrow milieu in patients with sympathetic nervous system damage may be more sensitive to common drugs such as terbutaline (beta agonist; enhances mobilization) and propranolol (beta blocker; inhibits stem cell mobilization). These agents could modify stem cell release provoked by VLA-4 blockers. Natalizumab effects on osteoclasts, derived from the monocyte lineage, and associated bone-resorbing dendritic cells are also possible.

Topics 3 and 4: Integrins affect number and activation of immune cells

(3) Integrins control immune cell homeostasis in the periphery and CSF.

The abundance of white blood cells in the circulation is kept constant by “immune homeostasis” regulated by feedback between the number of lymphocytes and their absorption of the serum cytokines, IL-7 and IL-15. More IL-7 is available when lymphocytes are missing, and the IL-7 then induces new lymphocytes, especially CD8 memory T cells, as well as CD4 regulatory and CD4 memory T cells. Some of these peripherally expanded cells are autoreactive or poor suppressors because their tolerance was not controlled by development in the thymus. Peripheral immune expansion and poor tolerance is especially relevant in multiple sclerosis, where thymic output falls to that of a person approximately 30 years older. IL-4, IL-2, IL-21, B-cell-activating factor, and type I interferons also enhance homeostatic expansion. Levels of proliferative cytokines are governed in the blood by (1) population pressure and uptake from immune cells and (2) the recent state of immune cell competence--influenced strongly by gut microflora (142). In multiple sclerosis, there are fewer recent (naïve) thymic emigrants (TREC+); polymorphisms in the IL-7 receptors are linked to onset of multiple sclerosis; oral antigens or parasites may be therapeutic; and unknown environmental factors influence onset and relapses of multiple sclerosis.

During natalizumab therapy, the peripheral blood lymphocyte count doubles. This lymphocytosis correlates with better clinical and MRI response (258). After treatment, there are twice as many circulating B cells and 7 times more immature B cells, but they produce less immunoglobulin (156; 287). The lymphocytosis is likely from release of B-cell and T-cell precursors from the bone marrow and thymus, evidenced by more immunoglobulin kappa-deleting recombinant excision circles and more T-cell receptor excision circles (301); both are under VLA-4 control. Secondly, natalizumab prevents lymphocyte migration into the CNS. It is possible that “excess” memory T cells, formerly destined to enter the brain, are gradually deleted from the blood. Similarly, T cells are deleted by apoptosis during interferon beta therapy (96), possibly through activation of overexpressed interferon-gamma receptors (01). Monoclonal anti-VLA-4 (antibody 9C10) causes apoptosis of activated T cells, but other anti-VLA-4 antibodies do not (antibodies PS/2 and R1/2). Nonetheless, all of these antibodies block VLA-4/VCAM-1 interaction (275; 165).

Most CSF immune cells, especially T cells, disappear during natalizumab therapy. Shifts in CSF immune cells after anti-VLA-4 therapy are profound. IL-7 levels are low in CSF during multiple sclerosis relapses (122), but the influence of low IL-7 on rare immune cells in multiple sclerosis CSF is unknown.

CSF oligoclonal bands, described as an unchanging “fingerprint” (Tourtellotte 1998, personal communication), do not usually change with multiple sclerosis therapies. In 4 of 6 natalizumab-treated patients, however, bands disappeared after 10 infusions, and local IgG production was decreased (285).

(4) Variable integrin expression and affinity on immune cells in multiple sclerosis.

Natalizumab binds to the alpha4 component of alpha4beta1 (VLA-4) and to alpha4beta7 (LPAM-1), inhibiting migration through brain endothelial cells by the vast majority of T cells and by 75% of monocytes (67). Therapy halves soluble VCAM in the serum, preventing soluble VCAM from activating blood-brain barrier endothelial cells to generate reactive oxygen species and matrix metalloprotease activity, which are required for leukocyte migration. Natalizumab reduces the number of VLA-4 positive T cells from 92% to 3%; others say the reduction is 50% (198; 186).

At baseline, VLA-4 is twice as high on blood B cells and monocytes as on T cells, twice as high on CD8 as on CD4 T cells, higher on memory T and B cells than on naive cells, and low on FoxP3 regulatory T cells (76; 198; 268). Cells with low VLA-4 expression have difficulty migrating through fibronectin in a 2-compartment Boyden chamber, a model of the blood-brain barrier (without VCAM) (198). With anti-VLA4 therapy, there is minimal entry of B cells and activated Th1 cells into the CSF and a parallel rise in the peripheral blood. Natalizumab therapy reduces expression of VLA-4 on all MNC; the effect is twice as pronounced on CD4 and CD8 T cells (50% of their original VLA-4 expression) as on B cells and monocytes (30% reduction) (199) within 90 minutes of the infusion (268). There are 2 cautions regarding this report of VLA-4 fluctuation. VLA-4 expression rebounds to normal in blood before the next monthly infusion, yet CSF immune cell percentages are not affected (198). Secondly, the high-affinity form of VLA-4 drives adhesion, but affinity was not measured.

Natalizumab therapy has differential effects on lymphocyte subpopulations. VLA-4 expression on naive and memory T cells is found to be low in multiple sclerosis by some authors (180; 74; 91), but not by others (78). Subnormal VLA-4 expression in multiple sclerosis is surprising, as chronic antigenic exposure elevates alpha4 integrin and osteopontin mRNA (eg, after bee venom immunotherapy to treat bee-sting hypersensitivity) (152). Already low VLA-4 levels on peripheral immune cells would suggest that anti-VLA-4 therapy in multiple sclerosis will be potent. VLA-4 expression is greater on lymphocytes in secondary progressive multiple sclerosis than in primary progressive multiple sclerosis and relapsing-remitting multiple sclerosis, suggesting that therapeutic benefit could vary with disease state. Note that these studies measure expression, not binding affinity.

Topics 5 and 6

(5) Costimulation and adhesion are targets for therapy in multiple sclerosis and may enhance peripheral immune activation by natalizumab.

Costimulatory interactions that enhance immunity include (1) CD80 (B7-1) and CD86 (B7-2) on antigen-presenting cells paired with CD28 (stimulatory) and CTLA-4 (inhibitory) on T cells, (2) CD40 on antigen-presenting cells with CD154 (CD40L) on T cells, and (3) LFA-3 on endothelial cells with CD2 on T cells. Blockers of costimulation and adhesion have been tested in multiple sclerosis.

CD80 expression is increased on B cells and on some activated T cells during multiple sclerosis exacerbations (92). Interferon-gamma and TNF-alpha increase during exacerbations; they induce CD80 and upregulate CD86 on endothelial cells. This is important because CD80 potently enhances antigen-specific B cell responses. Potentially part of its broad therapeutic effects, interferon beta downregulates CD80 (92). In a disappointing treatment effort in multiple sclerosis, CTLA-4-Fc constructs that blocked the interaction of CD80 with its ligands had no benefit (141).

CD40/CD40L costimulation is increased in multiple sclerosis immune cells as part of a positive feedback loop that generates IL-12 and interferon-gamma (15). CD40/CD40L induces chemokine secretion and enhances VCAM-1 and ICAM-1 expression by human endothelial cells. Trials to block this interaction in multiple sclerosis are in progress.

Based on animal studies, CD2/LFA3 is much less important than VLA-4/VCAM-1 in T cell penetration of the blood-brain barrier. CD2 expression on lymphocytes is abnormal in multiple sclerosis and less capable of transducing activation signals (231). Trials with blockers of this interaction are contemplated in multiple sclerosis.

Adhesion blockade has been ineffective as a therapy for multiple sclerosis apart from anti-VLA-4 antibodies. Anti-LFA-1 antibodies that block adhesion to ICAM-1 had no clinical effect on multiple sclerosis in a well-designed trial, although oral infections were more frequent (170).

VLA-4 interactions enhance immune activation in multiple sclerosis and in its disease models. Interaction of VLA-4 with its ligands is costimulatory and enhances activation, proliferation, and immune reactivity of T and B cells (255). T cells transfected with CD49d (the alpha4 component of VLA-4) proliferate strongly to self-cells and to otherwise suboptimal levels of antigen (191).

The importance of VLA-4 in CNS inflammation was established in experimental allergic encephalomyelitis, an animal model of multiple sclerosis. Anti-MBP T cell clones expressing high levels of VLA-4 were better at inducing experimental allergic encephalomyelitis than clones expressing low VLA-4 (16). Several anti-VLA-4 Abs significantly blocked T cell and monocyte adhesion to brain endothelial cells and reduced severity of experimental allergic encephalomyelitis (300). These experiments were with nonactivating anti-VLA-4 Abs (HP2/1, R1-2 – anti-alpha4 Abs and AIIB2 – anti-beta1 integrin).

Certain anti-VLA-4 Abs (P4G9, HP1/7, and PS/2) can activate immune responses (300; 276); others may block B cell activation. The PS/2 antibody induces a more active conformation of the VLA-4 molecule and enhances costimulation (276). Consequently, this antibody activates Th1 cells and interferon-gamma secretion and induces CD4 migration into the CNS. The PS/2 antibody, given in the first week after induction, ameliorates experimental allergic encephalomyelitis. However, if PS/2 is given at the peak of disease or during remission, it worsens experimental allergic encephalomyelitis. VLA-4 also enhances the ability of myelin basic protein-reactive T cells to induce proinflammatory cytokines in microglia (65). VLA-4 tethers to VCAM-1 on B cells and enhances B cell signaling. In parallel, the B cell receptor enhances the interaction and tight adhesion.

Ninety minutes after infusion of natalizumab into multiple sclerosis patients, their immune cells are refractory to stimulation. Natalizumab transiently reduces expression of VLA-4 on T cells by 49%, on B cells by 29%, and on monocytes by 25%, but the numbers return to baseline before each infusion (198). Natalizumab mildly upregulates IL-2, IL-17, and interferon-gamma in vitro and after the first infusion, likely inducing lymphocytosis.

Anti-VLA-4 therapy induces multiple RNA changes in immune cell subsets, as seen on 22,000-gene microarrays. It induces a large number of B cell activation and differentiation markers (167). Some of these may enhance JC virus replication, such as the Spi-B transcription factor, which elevates the neurotrophic JC virus variants in B cells and stem cells (176) (section 12). Natalizumab also induces T cell and monocyte expression of Tbet, a Th1 and possibly Th17 marker that induces both inflammatory and anti-inflammatory cytokines in serum (90).

(6) VLA-4 and interferon beta affect T cells and endothelial cells and modify penetration of the blood-brain barrier.

The postcapillary venule is the site of T-cell penetration through blood vessels. Blood flow slows past the capillaries, allowing more time for attachment. High-endothelial venules at this site have unique morphology and overexpress adhesion molecules that beckon tissue-specific immune cells.

The lesions of multiple sclerosis correspond to the paths of CNS veins, reflecting the importance of lymphocyte adhesion to postcapillary venules. Ninety-four of 95 plaques surround veins on MR venography (273). Venous anatomy, adhesion molecules, and flow dynamics suggest why multiple sclerosis lesions predominate in periventricular white matter. Cerebral blood volume is highest in the periventricular white matter of the brain (ie, many white blood cells pass through this area) compared to the centrum semiovale and intermediate white matter regions. The blood flow rate in the periventricular white matter in multiple sclerosis is more than 2-fold slower than in normal brain, allowing more time for adhesion (163). MRI visualizes relatively large venules downstream from the activated high endothelial venules. However, the lesions are likely to start at small-lumen postcapillary sites with activated high endothelial venules (see figure, Lymphocyte penetrating blood-nerve barrier).

The interaction between T cells and endothelial cells is dynamic and mutually reinforcing. Activated T cells and monocytes secrete cytokines that induce more MHC class II and costimulatory molecules, as well as P- and E-selectins, ICAM-1, and VCAM-1 on endothelial cells (232). Contact with CD8 and NK cells induces MHC class II on EC, independent of cytokines. Activation of endothelial cells in turn leads to stronger adhesion by T cells (282). Activated endothelial cells secrete chemokines, which induce the active conformation of VLA-4 on leukocytes. Interferon beta rapidly disrupts the positive feedback loop between white blood cells and endothelial cells; gadolinium-positive MRI lesions largely disappear within days of starting interferon beta therapy (43).

Penetration of the blood-brain barrier is a 2-step process. Cells from inside postcapillary venules must first cross the endothelial cell/tight junction of the blood-brain barrier. Many assume this occurs by breaking the strong connection between 2 endothelial cells. However, unique “tight” junctions in the brain endothelium do not express VCAM-1 and are nearly impervious, compared to peripheral junctions. Thus, one route into the brain may be through endothelial cells.

In this model, immune cells bind strongly to VCAM-1 and ICAM-1 on the surface of endothelial cells. They are then internalized and penetrate transcellularly, moving directly through endothelial cells (“emperipolesis” = “wandering around inside”) (119). This was first seen with lymph node endothelial cells (175), and later demonstrated in experimental allergic neuritis (EAN) (12) and experimental autoimmune encephalitis (295). In EAN, the immune cells appear to be activated, but the postcapillary venules do not seem activated and are not pseudo-columnar (unlike high endothelial venules).

The second step above, crossing the basal lamina from the perivascular space, requires gelatinases. These matrix metalloproteases secreted by T cells and monocytes are able to degrade dystroglycan and lyse the dense subendothelial lamina. Natalizumab may not affect this step. However, interferons prevent matrix metalloprotease secretion, suggesting potential synergy with natalizumab in preventing cells from entering the CNS.

Immune cell subsets differ in their ability to penetrate the blood-brain barrier:

(1) Fluor-labeled activated human cells injected into mouse carotid arteries adhere to lipopolysaccharide-activated venules, as seen with intravital microscopy of mouse brain. CD8 memory and effector cells, but not CD4 cells, from patients with acute multiple sclerosis relapses display excessive rolling and arrest (17). P-selectin and endothelial selectin binding is predominant by CD8 cells (and possibly more important in mice than humans), whereas VLA-4 and VCAM-1 interactions are most important for CD4 adhesion. Selectin binding by CD8 cells could account for the higher CD8/CD4 ratio in CSF during natalizumab therapy.

(2) With in vitro models of lymphocyte migration, ConA-activated CD4 T cells are more efficient than CD8 T cells and B cells at migrating through a monolayer of cytokine-activated rat brain vascular endothelial cells. B cells and CD8 cells adhere more avidly, but CD4 cells migrate better, especially after they or the endothelial cells are activated (174). In models (1) and (2), anti-VLA-4 antibodies would selectively decrease CD4 cell penetration.

(3) In a model using human cells, supernatants from microglia and astrocytes encourage firm junction formation in adult human brain endothelial cells. Monocytes and B cells migrate more easily than T cells (31). B cells use VLA-4/ICAM, but not VCAM (06). Monocytes have the most high-affinity VLA-4 molecules (Table 1), but also use CCL-2 and metalloproteases for penetration. Once through the blood-brain barrier, monocytes increase its permeability to soluble molecules and to migrating T cells (253).

Th2 lymphocytes migrate twice as well as Th1 cells (31). Th2 cells express more VLA-4 and LFA-1 (Antel J 2005, personal communication), and natalizumab inhibits Th2 migration more than Th1 migration. Only Th2 cells express the chemokine receptor, CCR2, which recognizes the chemokine, MCP1, secreted by endothelial cells (31). PSGL1/CLA on Th1 cells, not on Th2 cells, binds to E- and P-selectins on endothelial cells (215). Because Th2 cells migrate better, this suggests that VLA-4 and CCR2 are critical molecules at the blood-brain barrier, and the selectins are not. Supernatants from Th1 cells, but not Th2 cells, upregulate VCAM-1 and ICAM-1 on endothelial cells; this demonstrates the dynamic interaction between Th1 and endothelial cells. Conversely, preincubation of these human brain endothelial cell cultures with interferon beta reduces migration of Th1 cells, but not Th2 cells (221). This implies that even though interferon beta may not cause a Th1 to Th2 shift in the peripheral immune compartment, it can enhance immune suppression in the CNS by favoring Th2 over Th1 migration. CD8 suppressor cell entry has not been studied.

(4) In experimental allergic encephalomyelitis, myelin oligodendrocyte glycoprotein (MOG)-specific Th1 cells express high levels of alpha4 integrin and are dependent on VLA-4 for entry into the spinal cord. Th17 cells have low levels of alpha4 integrin (but more alpha4,beta7, another natalizumab target). Antigen-specific Th17 cells, even in mice whose immune cells do not express VLA-4, are able to enter the brain by using LFA-1 and alphaL,beta2 (237).

(5) Concomitant medications can alter these dynamics. Glatiramer therapy induces IL-4, a Th2 cytokine that inhibits VLA-4 expression on CD8 cells (246). Glatiramer in combination with natalizumab could prevent anti-JC virus CD8 cells from reaching the brain.

Natalizumab blocks immune cell entry into the CSF in multiple sclerosis (268). With therapy, CD4 T cells, CD19 B cells, and CD138 plasma cells are essentially gone from the CSF; CD8 T cells fall to one fifth of pretherapy levels. The ratio of CD4/CD8 falls disproportionately, perhaps because CD8 cells express more VLA-4 than CD4 cells, even after downregulation of expression by natalizumab. Th17 cells are less affected than Th1 cells in mice. Entry of CD4+FoxP3+ regulatory T cells (Tregs) is also blocked, comparable to other subsets; their ineffective function in multiple sclerosis does not change during natalizumab therapy (265). B cells in the CSF are reduced, as are the IgG index and possibly oligoclonal IgG bands.

Topics 7 and 8

(7) Brain immune regulation is unique and differs from immunity in blood.

The brain is an “immunologically privileged” site. A tight blood-brain barrier and high levels of P-glycoprotein transport by the endothelium block serum proteins from leaking into the surrounding tissue. There is reduced immune activation in the CNS – less leukocyte trafficking, significant endogenous immunosuppression (glial production of neurosteroids, prostaglandins, TGF-beta, and IL-10), microglia that are poor at presenting antigens, low MHC class I and minimal or no MHC class II expression on CNS cells, and restricted lymphatic flow out of the brain (230). There is directed lymphatic flow from CNS perivascular spaces (glymphatic drainage; glial plus lymphatic) to cribriform plate, to nasal lymphatics, and then to cervical lymph nodes. This induces a Th2 immune bias to CNS antigens in the periphery. Less inflammation against CNS antigens should ameliorate multiple sclerosis and possibly neurodegenerative diseases.

In multiple sclerosis, endothelial cells and microglia are activated, astrocytes are activated and hypertrophied, and inflammatory cytokines disrupt oligodendroglia and neurons. The brain loses its immune privilege, and there is intermittent immune activation and brain cell death.

Adhesion molecule expression on normal brain endothelial cells and within the brain parenchyma is usually lower than in other sites. Even during experimental allergic encephalomyelitis, VLA-4 and LFA-1 are downregulated on T cells after they enter the CNS environment (233). VCAM-1 is undetectable on normal human endothelial cells, but it appears on microglia and monocytes at sites of inflammation in multiple sclerosis (213). Low-level expression of VCAM-1 on blood vessels and microglial cells and of VLA-4 on immune cells in perivascular cuffs and parenchyma in acute plaques evolves to strong staining in chronic active and chronic silent multiple sclerosis plaques (44). VCAM-1 is the main adhesion molecule on endothelial cells in the CNS and on pericytes in multiple sclerosis plaques; selectins and ICAM have minimal expression (282). This matches the therapeutic failure of anti-LFA-1 (the ICAM partner) in multiple sclerosis (170), compared to the pronounced benefit from natalizumab.

“Pioneer lymphocytes” arrive in the CNS within 2 hours after infusion of anti-myelin T cells in mice (passive transfer experimental allergic encephalomyelitis) (46). These cells penetrate into the CNS by binding to P-selectin and ICAM-1 expressed on undefined cells in the choroid plexus. These cells do not use VLA-4 for adhesion and would not be blocked by anti-VLA-4 therapy. This source of brain-antigen-specific T cells is not sufficient to prevent PML in multiple sclerosis. In multiple sclerosis brains, P-selectin is found on large venules in the meninges and choroid, but not small venules – the site of T cell penetration (146), so P-selectin may not be involved in immune breach of the blood-brain barrier. During experimental allergic encephalomyelitis, choroid plexus epithelial cells upregulate VCAM-1 and ICAM-1, and express MAdCAM-1 de novo (81). These molecules, however, are not seen on the choroid vascular endothelium or on parenchymal endothelium. In contrast, MBP-specific cloned CD4 T cells adhere to VLA-4 on subcortical and periventricular endothelial cells 24 hours after injection (46). The later influx of memory T cells is likely to be blocked by natalizumab. Osteopontin is a matrix glycoprotein named for its stimulatory effects on bone osteoclasts. It is expressed by macrophages and astrocytes in and around multiple sclerosis plaques, and by microglia, astrocytes, neurons, and choroid plexus in experimental allergic encephalomyelitis lesions (50).

Osteopontin is a ligand for the activated form of alpha4beta1 integrin and several other integrins expressed on monocytes, so it could modify responses to anti-VLA-4 Abs (19). The biological effects of osteopontin in multiple sclerosis, including changes in VLA-4 function, are complex.

T cells and monocytes secrete osteopontin. In sarcoidosis, lymphocyte expression of osteopontin correlates with granuloma maturity, T cell chemotaxis, T cell adhesion, and T cell costimulation (205). Osteopontin prevents apoptosis of activated T cells and some monocytes. It is induced by interferon-gamma, and in turn osteopontin induces IL-12 and more interferon-gamma-cytokines that are detrimental in multiple sclerosis. Osteopontin enhances Th1-type immunity, triggers relapses of EAE, and increases CNS inflammation. Osteopontin knockout mice have a shift to Th2 immunity, milder experimental allergic encephalomyelitis, and less neurodegeneration in models of Parkinson disease and stroke (50). Twelve hours after ischemia to the brain, osteopontin appears in microglia and induces astrocyte chemotaxis and activation through integrin signaling (Ellison et al 1998). It also induces proliferation of oligodendrocyte precursors and formation of myelin. In multiple sclerosis, osteopontin serum levels are slightly elevated and correlate with a generalized shift to Th1 immunity. In mononuclear cells from untreated patients, the SPP1 (osteopontin) gene is elevated 200-fold in stable multiple sclerosis and 900-fold during exacerbations (82).

In contrast to osteopontin’s osteoclast stimulating activity, interferon beta inhibits osteoclast differentiation and actually increases bone formation (271). Regulatory T cells secrete TGF-beta plus IL-4 and also block osteoclast differentiation. Interferon beta-1b therapy does not elevate osteopontin levels in relapsing-remitting multiple sclerosis (59). It reduces RNA expression compared to untreated stable multiple sclerosis by 450-fold and to active multiple sclerosis by 1700-fold (82). Thus, interferon therapy may synergize with natalizumab on multiple biological effects.

(8) Anti-VLA-4 therapy for multiple sclerosis.

One dose of natalizumab given during an acute multiple sclerosis relapse had no effect on further relapses or progression but did reduce gadolinium-positive MRI lesions (203; 241). Two doses, given a month apart, reduced MRI lesions (278).

Occasional patients have worsening of multiple sclerosis symptoms after the first 1 or 2 infusions. In a subgroup of highly active patients, lesions active at baseline persisted longer in the first 2 weeks after starting natalizumab (278). There was also a rebound increase in relapses in this subgroup during months 3 to 5 after the second of 2 infusions (38% vs. 9%, p < 0.005) (278). Several small series suggest that occasional patients who had ongoing clinical activity when the drug was started will have exacerbations. Early multiple sclerosis activity might be expected in a treatment that is not 100% effective, or it may be from elevation of pre-B cells (7-fold increase) and mature B cells (3-fold increase) in blood (156), fewer regulatory T cells (225), or activated T cells that produce inflammatory cytokines (interferon-gamma, TNF-alpha, and IL-17) (138; 145).

However, a rebound was not seen in a Danish nationwide study (206) and in a post-hoc analysis of the highly active subgroup (108) of several large studies. In the phase IIb Antegren/natalizumab study, 6 monthly infusions reduced relapses by 57% in 213 relapsing-remitting and secondary progressive (one third of total) patients (183). (Later observational studies also showed benefit on cumulative clinical damage in secondary progressive multiple sclerosis, but there was no benefit in primary progressive multiple sclerosis.) After 6 months of treatment, there was no significant rebound of exacerbations, and the therapeutic effect lasted 2 to 3 months after the final infusion in both forms of multiple sclerosis. Thus, rebounds may be more likely after short-duration therapy, but less likely after longer-duration treatment.

Full responders, with no clinical or MRI activity during natalizumab therapy, have less active disease at baseline (223). Counterintuitively, pre-study clinical multiple sclerosis activity does not predict rebound activity after withdrawal of natalizumab (292; 241). Based on these studies, inactive disease may also rebound after natalizumab cessation. Control groups are essential in these comparisons.

In a 25-month phase III multiple sclerosis trial of 942 relapsing and remitting patients (randomized at 2 natalizumab-treated to 1 placebo; AFFIRM study), monthly infusions of 300 mg natalizumab reduced new and enlarging T2 lesions by 83%, Gd-enhancing MRI lesions by 92%, T1 black hole development by 40% (15% with natalizumab compared to 25% with placebo) (64), and relapse rate by 66% (218). Brain atrophy was greater in year 1 but less in year 2; the fall at 1 year may be from anti-inflammatory effects, as seen with interferon beta and glucocorticoids (184). Progression was slowed by 36% at 1 year and 42% at 2 years (confirmed by neurologic exam scores, 3 months apart). Post-hoc analysis showed that treated patients were more likely than placebos to have “no evidence of disease activity” on clinical (64% free in treated vs. 39% in placebo), radiological (58% vs. 14%), or both measures (37% vs. 7%) (108). Therapy reduced the risk of cognitive worsening by 43% compared to placebo. Neurofilaments in CSF reflect damage to CNS nerves. Natalizumab therapy reduces neurofilaments 3-fold in CSF, down to healthy control levels (105).

Another trial evaluated 1171 patients who had had at least 1 exacerbation in the prior year during therapy with weekly intramuscular interferon beta-1a, ie, they were weighted toward suboptimal interferon beta-1a responses. Patients were randomized into interferon alone or interferon plus natalizumab (SENTINEL study). Natalizumab in combination with interferon reduced new and enlarging T2 lesions by 83%, Gd-enhancing MRI lesions by 89%, relapse rate by 55%, and progression by 24% at 2 years, compared to interferon alone (243). The superiority of natalizumab versus placebo was not studied. Natalizumab also abolished MRI and clinical activity in 5 patients with aggressive multiple sclerosis who had only partial response to cyclophosphamide plus autologous stem cell transplantation (45).

The annual relapse rate is reduced by 0.92 when the expanded disability status scale (EDSS) rating is less than or equal to 3.5, by 0.70 for EDSS of 4 to 6, and by 0.57 for EDSS greater than or equal to 6 (86). In most patients, the onset of benefit is rapid, regardless of baseline disease activity. There is actual improvement in some patients. Therapy restores visual and sensory evoked potentials in one third of patients but does not improve magnetic evoked potentials (182). EDSS scores improve in 69% of patients over 2 years versus placebo; this correlates with quality of life.

Quality of life on physical and mental subscales improved with natalizumab in 2 trials; subjects on placebo or interferon beta-1a alone declined (239). There was a 43% decrease in cognitive decline versus placebo, but the benefit was equivalent to intramuscular interferon beta-1b (289). Natalizumab also reduces loss of visual acuity. Fatigue is reduced in most studies. Pain from multiple sclerosis is reduced in 56% of patients; 15% have more pain with therapy (Foley J 2009, personal communication). Migraine severity decreases from 14 to 10.5 on the migraine disability assessment scale. Cost-benefit analysis shows that this therapy increases productivity by 3.3 hours per week (3216 euros).

Forty-nine patients of African ancestry in 2 large studies had benefit from natalizumab on clinical (60% fewer relapses), cognitive, and MRI measures; benefit was equivalent to that seen in Caucasian patients (62). In case reports, 5 children who failed other therapies improved on natalizumab. In 19 children with multiple sclerosis, the drug improved EDSS and prevented Gd-enhancing lesions (94). Others later found benefit in larger groups of children. Pregnancy registries show no adverse outcomes.

Low serum vitamin D levels are associated with more attacks while on natalizumab. The annualized relapse rate is 0.31 with vitamin D levels of less than 50 nmol/L but improves to 0.10 with levels of 50 nmol/L or higher (252).

Proinflammatory interferon-gamma and TNF-alpha rise in mononuclear cells during natalizumab therapy, perhaps reflecting more activated peripheral immune cells. Circulating B and cytotoxic NK cells increase but CD8+CD28- regulatory T cells do not. However, the few cells that penetrate into the CSF are anti-inflammatory, such as CD4 Treg and IL-10+ PD-1+ CD8 Treg; meanwhile, proinflammatory T-bet+ CXCR3+ B cells are reduced. CSF white cells express more IL-10 and less interferon-gamma, IL-1beta, IL-6, IL-23, osteopontin, chemokines, and matrix metalloprotease-9 (138), and there is a 3-fold reduction of neurofilament light chains to normal levels. Lower osteopontin correlates with cognitive improvement. Rebounds occur in some patients after therapy cessation, perhaps from an influx of cytolytic, T helper, or Th17 cells into the CNS.

The half-life of natalizumab after a single 3 mg/kg dose is 4 to 5 days (242). After monthly infusions of 4.5 mg/kg, the half-life becomes 11 days. The 3 mg/kg dose also shows a half-life increment over time. Natalizumab-VLA-4 complexes are internalized and presumably degraded. Receptors (VLA-4) are over 80% saturated for at least 28 days after a single 300 mg dose; binding falls to 50% at 2 months and 40% at 3 months (242). The drug reduces expression of VLA-4 on B cells twice as much as on T cells; CD4 and CD8 cells have the same drop in VLA-4 expression (225).

Expected loss of drug activity can be based on kinetic markers after drug withdrawal. Serum natalizumab level is less than 1 ug at 84 days, immune cell surface saturation with anti-VLA-4 is less than 20% at 42 days, and white blood cell count falls to normal levels at 98 days. This suggests that a prolonged half-life, plus a slow off-rate, should lead to a therapeutic effect that lasts for several months after drug discontinuation (see extended dosing to prevent PML, section 16). In support, there is persistent depression of the CSF white cell count after natalizumab therapy for at least 6 months (reduced CD4 and CD8 T cells, CD19 B cells, and CD138 plasma cells; CSF changes are greater than changes in the blood; additional therapy was not detailed) (268). A rise in the CSF cell count correlates with anti-JC virus serum antibodies.

However, in another series of patients on no post-natalizumab therapies, the rebound of CSF white blood cells after discontinuation was faster, with low CSF white blood cells at 50 days, but a rebound to normal after 100 days (124). Cells remained low for 100 days after treatment with the combination of natalizumab plus interferon or glatiramer. In the Stuve series, CSF cell counts return to baseline at 14 months in patients treated with other therapies (266). There have been cases of PML 1 to 6 months after the drug is stopped, before CNS immunity is restored.

Despite these kinetic and immune effects, a third of patients have recrudescence of multiple sclerosis symptoms 24 to 28 days after each infusion (“wearing off” phenomenon). This is more common early in the course of treatment, as natalizumab serum levels climb and the disease quiets over time. Natalizumab ameliorates active Crohn disease (274) and possibly rheumatoid arthritis. Case reports show concomitant improvement in psoriasis and Rasmussen encephalitis. Anti-VLA-4 Abs prevent insulitis in spontaneous diabetic NOD mice.

Small-molecule VLA-4 antagonists prolong rat cardiac transplants and inhibit experimental autoimmune uveitis. In multiple sclerosis, firategrast, an oral anti-alpha4beta integrin molecule, reduced Gd-enhancing MRI lesions at high doses, but increased lesions at the low dose. With its 3.5-hour half-life, serum levels were variable, and high doses increased bladder infections and caused vomiting (185).

In summary, anti-VLA-4, an antigen-nonspecific therapy, ameliorated MRI lesions, exacerbations, cognitive loss, and progression and markedly reduced CSF white blood cells in multiple sclerosis. Side effects were minimal, except for PML, and combination with interferon was superior to interferon beta alone. It should not be forgotten that this drug and all other multiple sclerosis therapies are not perfect, and there are still some breakthrough exacerbations and progression.

Topics 9 and 10

(9) Adverse effects of anti-VLA-4 therapy and unexpected benefits.

There are few rare, allergic reactions to natalizumab (4%). In patients with mild to moderate infusion reactions, some infusion centers premedicate with antihistamines (diphenhydramine 50 mg, loratadine 10 mg), acetaminophen (1000 mg), and oral steroids (prednisone 50 mg).

There was no tuberculosis in the pivotal trials, and no statistically significant suggestions of increased cancer, melanoma, CNS lymphoma, or infections. However, there is a risk of PML, which is dependent on prior immunosuppression and anti-JC virus titers. In contrast, VLA-4 blockade is potentially therapeutic in cancer, as is interferon beta. Integrins expressed on cancer cells enhance metastasis and tumor invasion; block of integrins by natalizumab could potentially prevent metastasis.

Neutralizing antibodies to natalizumab increase the chance of an infusion reaction. Hypersensitivity reactions are 5% overall, but with re-exposure after prolonged absence of therapy, are 24% (201). The effect of VLA-4 polymorphisms on neutralizing antibody generation and on the efficacy of natalizumab therapy is unknown. Smokers have a 2.5-fold higher rate of neutralizing antibody formation than nonsmokers, in addition to the reality that smoking promotes exacerbations and progression and cancels the benefit of multiple sclerosis therapies (234; 110).

Natalizumab increases peripheral lymphocyte counts in 50% and causes eosinophilia in 20% (39). Infusion reactions are 10-fold more common in those with high eosinophils. The rise in number of B, NK, and T cells is a marker of better efficacy; those with relapses on therapy are 4 times as likely to have only a minimal rise in lymphocytes (258). Serum antibody titers to cytomegalovirus, Epstein-Barr and JC virus, and myelin-oligodendrocyte basic protein (MOBP) are increased, but viruses other than JC virus are not reactivated (127). Six cases of drug-induced liver injury have been reported.

During embryonic development, VLA-4/VCAM-1 interactions are important in chorioallantoic fusion, skeletal development, and formation of the septum between the aorta and pulmonary trunk (63), as well as sympathetic cardiac innervation. Fibronectin and integrins, including VLA-4, are important in early angiogenesis in the CNS (188). Embryonic defects appear in murine VLA-4 knockouts and with small molecule antagonists of VLA-4. A small molecule alpha4 integrin antagonist, oral CDP323, had no clinical benefit in multiple sclerosis, but pregnancies were not reported. Importantly, in multiple teratogenicity studies of anti-VLA-4 antibodies, defects have not been seen because antibodies are too large to cross the placenta.

A pregnancy registry for natalizumab has been established. Of 35 German women who received natalizumab during early pregnancy, 28 had normal infants, one had hexadactyly, and the others had miscarriages or elective termination, suggesting there was no adverse effect during pregnancy (112). There was no rebound in multiple sclerosis activity with drug withdrawal in these pregnancy studies, but another study shows there is a rebound (219).

There are rare reports of immune thrombocytopenic purpura (ITP), anemia, primary CNS lymphoma, herpes encephalitis, neuroborreliosis, cryptococcal meningitis, and ocular toxoplasmosis. However, there is no apparent increase in these diseases over background levels in over 142,000 treated patients (July 2015), and latent tuberculosis is not activated.

Rebound of multiple sclerosis after discontinuation of natalizumab is discussed in topics 8 and 16.

In neuromyelitis optica, 7 months of natalizumab slightly increased the exacerbation rate and significantly increased the EDSS score in 5 patients (149).

Possible unexpected benefits of natalizumab. GM-CSF and natalizumab both trigger release of cells from the bone marrow and are potentially useful in harvesting cells for immune cell transplants (topic 2).

Treatment does not change immunoglobulin G responses to vaccination with tetanus toxoid but reduces responses to other antigens to 70% of normal. It also reduces the expanded T-cell receptor repertoire of multiple sclerosis in blood, and more so in CSF. There is re-expansion of the repertoire, perhaps virus-specific, during natalizumab withdrawal while treating PML.

Antibodies to VLA-4, and small molecules that block VLA-4, will block VLA-4 binding to fibronectin. In rats with tendon injuries, anti-VLA-4 improves healing by preventing adhesion formation, a major impediment to healing (129).

Multiple myeloma cells bind to cells and stroma and are activated by vascular endothelial growth factor (VEGF). Natalizumab disrupts myeloma binding to stroma and inhibits stroma- and VEGF-induced signaling and tumor growth; it sensitizes the myeloma to proteasome inhibitors (217). Similarly, anti-VLA-4 disrupts immune escape of B-cell lymphoma cells by preventing their bind to bone marrow stroma (195).

(10) Type I interferons potentiate the effects of natalizumab but can also prevent JC virus infection.

The combination of interferon plus natalizumab in the pivotal trials was linked to 2 cases of PML. The apparent synergy in disease induction between interferon beta-1a and natalizumab, compared to monotherapy, is not statistically significant, with only 2 PML cases. However, the 2 agents could theoretically synergize in JC virus pathology in multiple ways.

Interferon beta induces the immunosuppressive Th2 cytokine, IL-10, in T cells (85; 107). Interferon beta also generates CD8 suppressor T cells in vitro and in vivo (200). This cell subpopulation is responsible for some of the therapeutic benefit of interferons and glatiramer acetate but could potentially suppress CNS immunity. VLA-4 blockade of trafficking by CD8+CD28- regulatory cells has not been studied.

Anti-VLA-4 antibodies and interferon beta therapy reduce adhesion and impair migration through brain vascular endothelial cells. Anti-VLA-4 antibodies selectively decrease CD4 cell penetration of the blood-brain barrier (174; 17), likely defusing the antiviral response in the CNS. It is also likely that interferon therapy causes a shift in the immune subpopulations that cross the blood-brain barrier, and this could synergize with a block of blood-brain barrier adhesion by natalizumab. Preincubation of human brain endothelial cell cultures with interferon beta reduces migration of Th1 cells, but not of Th2 cells (221), favoring (inhibitory) Th2 over Th1 migration.

Interferon beta therapy reduces expression of VLA-4, ICAM-1, and VCAM-1, and other adhesion molecules on endothelial cells (229; 159) and on T cells (180; 41) and monocytes in vitro (260). Interferon therapy reduces VLA-4 mRNA in mononuclear cells from patients who respond clinically to interferon beta but has no effect on nonresponders (41; 196). Second, during interferon beta therapy, VCAM-1 is shed from endothelial cells (likely peripheral endothelial cells, but possibly activated CNS endothelial cells) (41). Serum VCAM could block VLA-4 on immune cells and, in turn, block T cell-endothelial cell interactions. Nevertheless, the binding affinity of soluble VCAM is 100 to 1000 times less than membrane-bound VCAM, so circulating VCAM may have little impact.

Interferon beta reduces barrier permeability in co-cultures of brain endothelial and astrocyte cultures through a direct effect on endothelial cells, without any change in tight junction proteins (155). It downregulates interferon gamma-induced adhesion molecules on endothelial cells and counteracts the increased microvascular permeability induced by lipopolysaccharide (229; 159). Passage of T cells into the CNS through a transcellular route (emperipolesis), and not through tight junctions, predominates in experimental allergic encephalomyelitis (295). Passage through cells here is dependent on interaction between LFA-1 and ICAM-1 (102). Type I interferons also increase CD73 on endothelial cells (197). This protein converts extracellular AMP into adenosine, a potent anti-inflammatory molecule that reduces permeability through umbilical vein endothelial monolayers. Type I and II interferons impair endothelial cell pinocytosis (109), already low in blood-brain barrier endothelial cells, and this may further block positive feedback between endothelial and immune cells.

Finally, matrix metalloproteases, reduced by interferons, can alter the surface charge of endothelial cells and allow passage directly through endothelial cells into the perivascular space and then facilitate passage through the basement membrane into the brain.

Immune cells in multiple sclerosis secrete excessive levels of matrix metalloproteases, which degrade tight junctions. Activation of lymphocytes via VLA-4 induces production of matrix metalloproteases. However, interferon beta markedly reduces matrix metalloprotease secretion by T cells, inhibiting penetration of monocytes and lymphocytes through the blood-brain barrier.

Interferon beta could indirectly affect serotonin 5HT2A receptor expression and increase susceptibility to JC virus infection. Interferon beta induces the enzyme indoleamine 2,3-dioxygenase, which decreases tryptophan and serotonin. When tryptophan is reduced, there is a compensatory increase in 5HT receptor expression. More serotonin 2A receptors should enhance JC virus entry into neurons, astrocytes, and oligodendrocytes because all these cells express 5HT2A receptors. Interferon beta has not been associated with the development of PML. However, Sjögren syndrome and lupus, which have high endogenous type I interferon levels, are linked to JC virus infection (83).

Interferon beta therapy could have other consequences when used in conjunction with anti-VLA-4 therapy. Type I interferons block secretion of chemokines, which induce G proteins to generate the activated conformation of VLA-4. The relative effects of 6 MU weekly interferon beta versus more frequent, high-dose interferon beta on interactions with natalizumab are unknown. High-dose interferon alpha interferes with neuronal function, but interferon beta does not (see MedLink Neurology article “Multiple sclerosis”). Finally, inhibition of beta1 integrin signaling may inhibit myelination (23).

Type I interferons enhance T cell responses against polyoma viruses (103). JC virus induces interferon-stimulated genes in human fetal glial cells. Toll-like receptors (TLR) on immune cells recognize single-stranded (TLR7) and double-stranded RNA viruses (TLR3) and DNA viruses (TLR9). Activation of these receptors causes B cell proliferation (TLR7, TLR9) and induces type I interferon secretion by plasmacytoid dendritic cells. In a report, however, interferon beta therapy seemed to diminish responses through TLR9, the receptor for unmethylated DNA on plasmacytoid dendritic cells (14). This would allow JC virus, a double-stranded DNA virus, to escape from this arm of innate immunity. However, interferon alpha has no effect on TLR9 in resting plasmacytoid dendritic cells (21), and type I interferons actually upregulate their own interferon responses (84).

In several reports, interferon beta had only suggestive benefit in human PML. Interferon’s antiviral actions should prevent PML but may not be curative because little interferon is able to cross the blood-brain barrier. It is believed that only 0.1% of serum interferon beta enters the normal CNS (229). Despite this, a strong interferon signature in mouse brain 6 hours after interferon beta injection indicates there is biologically significant penetration into the CNS (97). CSF interferon beta levels have not been studied in multiple sclerosis, but there is blood-brain barrier leakage in multiple sclerosis from CNS inflammation.

In contrast to the arguments above, interferon beta inhibits JC virus infection and spread in cultured SV40-transformed human fetal glial cells (204). In vivo, interferons reduce levels of serum JC virus DNA. In a study with much higher serum DNA levels than in other reports, serum is positive in 29% of healthy controls, 46% of untreated multiple sclerosis patients, and 14% of interferon beta-treated patients (66).

Topics 11 and 12

(11) Other therapies may potentiate the clinical benefit of natalizumab.

Statins interfere with leukocyte-endothelial cell adhesion by reducing the inflammatory response generated through lipid mediators in venules (144). Statins also reduce CD11b on monocytes and VCAM-1 expression on endothelial cells (228).

Links to PML vary among other multiple sclerosis disease-modifying drugs but are quite rare through 2020: 37 fingolimod, 11 dimethyl fumarate, 0 teriflunomide, 1 ocrelizumab, 1 interferon beta, and 0 glatiramer acetate.

FTY720/fingolimod causes lymphocytes to be retained in lymph nodes and prevents them from reaching the CNS. Lymphocyte egress from lymph nodes is blocked when FTY720 binds to the sphingosine 1-phosphate receptor-1 (S1P1). Interferon beta also causes lymphocyte retention in lymph nodes by inducing the lectin CD69, which in turn binds to and negatively regulates S1P1 (256). FTY720 and interferon beta would be expected to synergize by reducing lymphocyte penetration of the blood-brain barrier. FTY720 could enhance antiviral activity by retaining virus and immune cells within secondary immune organs (222). Conversely, it could prevent cytolytic CD8 cells from entering the CNS. Dimethyl fumarate lowers the lymphocyte count; low counts are linked to PML cases.

Glatiramer does not cause PML. It causes a Th1 to Th2 shift that may cause CNS immune suppression in some patients. IL-4, a Th2 cytokine, inhibits VLA-4 expression on CD8 cells (246). Synergy with natalizumab has not been tested.

In other life-threatening disorders, PML is associated with many drugs used as multiple sclerosis treatments, including glucocorticoids, autologous bone marrow transplants, cyclophosphamide, mycophenolic acid, azathioprine, methotrexate, intravenous immunoglobulin, mitoxantrone, and monoclonal antibodies (adalimumab, 1 case; alemtuzumab, 14 cases; bevacizumab, 3 cases; cetuximab, 1 case; efalizumab, 8 cases (1/55; this drug has a black box warning for PML); radioactive ibritumomab tiuxetan, 5 cases; infliximab, 4 cases; and rituximab, 114 cases (1/30,000)), in WHO, Canadian, and PubMed databases (137; 26). Efalizumab, anti-CD11a (LFA-1) used to treat psoriasis, caused PML in 1 in 500 patients and was withdrawn from the market because the risk-to-benefit was too high. In non-multiple sclerosis patients, at least 114 cases of PML have occurred with rituximab, an anti-CD20 MAb. The incidence of PML is 1 out of 100,000 with rituximab used to treat rheumatoid arthritis, but 1 out of 5000 in systemic lupus erythematosus. There are no cases of PML in multiple sclerosis patients treated with rituximab. Many of these immunosuppressive drugs cause the release of bone marrow cells. Rituximab is often used in combination with immunosuppressants to treat connective tissue disease. “Demyelinating disease” after anti-TNF antibody infusions could be confused with PML (192).

Chemotherapy suppresses immune responses. The combination of natalizumab with other drugs, such as fludarabine and cladribine, and also rituximab, leflunomide, statins, glatiramer, and even beta-adrenergic agonists and pentoxifylline, poses theoretical risks (28; 47; 171).

(12) PML pathology and clinical profile: JC virus, target cells, and why PML develops with natalizumab therapy.

Before 2004, when natalizumab was approved to treat multiple sclerosis, PML had never been seen in untreated multiple sclerosis patients. Why did PML appear during natalizumab therapy, without any increase in other infections or CNS cancers? What is the relative risk of PML with natalizumab alone, when combined with interferon beta, or with immune suppression?

In the pre-AIDS era, PML caused visual deficits (40%; typically retrochiasmal homonymous hemianopsia or cortical blindness, not from optic nerve lesions), motor weakness and hemiparesis (30%), dysarthria and myoclonic seizures (20%), and change in mentation and personality (33%), plus occasional apraxia, abulia, and aphasia. Onset is usually insidious over weeks to months but is sometimes rapid and is occasionally indolent over 1 year. In contrast to multiple sclerosis, PML is subacute and symptoms continue to progress; PML is associated with neurobehavioral symptoms, aphasia, and higher cortical symptoms. PML rarely involves the optic nerves or spinal cord (172; 101; 37). PML in multiple sclerosis patients on natalizumab therapy presents with cognitive problems (48%), motor abnormalities (37%), language disturbance (31%), and visual deficits (26%) (25). Approximately 25% of multiple sclerosis patients with PML die from PML; 75% survive with noticeable or severe disability.

A third of patients with PML have seizures; 90% of these PML lesions have a cortical MRI signal, compared to 50% without seizures (140). EEG may show slowing before MRI becomes diagnostic (150).

MRI is almost always positive during PML (Table 2). Lesions are more common in frontal and parieto-occipital areas. The lesions are often large, monofocal, asymmetric, and diffuse in the white matter; usually have no mass effect; and seldom regress. They show hypointense T1 and high-signal T2 (“ground glass”), FLAIR, and DWI lesions. Lesions enhance in approximately 40% of patients (vs. 10% in HIV), diffusely or with a punctate, perilesional “Milky Way” appearance. Lesions are often in the centrum semiovale with a sharp border and sometimes extend up to the subcortical U fibers, forming a scalloped, often ill-defined border there. PML lesions are somewhat more likely to appear at the gray-white junction than in periventricular spaces, suggesting local tropism in affected areas of the CNS or differences in local CNS immune regulation. Adjacent to the subcortical lesion, 60% have hyperintense cortical signal. Lesions also occur in the basal ganglia, thalamus, and brainstem (“across the pons”). Crescentic cerebellar lesions are seen in one fourth of PML cases, but not in multiple sclerosis. Widespread MRI lesions are associated with fatality. Transient punctate gadolinium-positive areas in the parietal, frontal, and thalamic areas can precede T2 lesions by 6 months (179) and are present in 90% of patients with PML. Rare reports show insidious PML with no MRI changes (151).

PML lesions differ from multiple sclerosis plaques. New MRI lesions are smaller; are in periventricular regions, optic nerve, and cord; and usually enhance with gadolinium. Cortical lesions are not subpial in PML, as commonly seen in multiple sclerosis. Giant multiple sclerosis plaques can be confused with PML. Lesions in spinal cord and optic nerve are exceedingly rare and are uncommon in the brainstem in PML, unlike in multiple sclerosis. The relative balance for lesions in PML versus multiple sclerosis is as follows: large confluent T2 (74%/2%:PML/multiple sclerosis), deep gray matter (31/7), and crescentic cerebellar (23/0) (37). Elevated myoinositol and lipid/creatine and low N-acetylaspartate (NAA) on magnetic resonance spectroscopy also indicate inflammation.

Table 2. MRI Changes in PML, Multiple Sclerosis, and Stroke

JC virus–associated granule cell neuronopathy causes cerebellar symptoms and cerebellar atrophy that are visible on MRI. Two cases have been reported with natalizumab therapy of multiple sclerosis and more with HIV, CD40 ligand deficiency, sarcoidosis ± immune therapy, and following immune therapy for non-Hodgkin lymphoma. JC virus encephalopathy from infected cortical pyramidal cells and JC virus meningitis from infected leptomeningeal cells have been reported in non-multiple sclerosis patients.

The histopathologic triad in PML is multifocal demyelination, oligodendroglia with enlarged hyperchromatic nuclei, and bizarre enlarged astrocytes with hyperchromatic lobulated nuclei. Demyelination is from JC virus–induced dysfunction and cytolytic destruction of oligodendroglia (lysis is necrotic, not apoptotic), with secondary loss of myelin. Oligodendrocyte nuclei become large, hyperchromatic, and basophilic with intranuclear inclusions comprised of crystalline arrays of virus particles (“fried egg” appearance). Virus can spread within oligodendrocyte extensions or along myelin in white matter, but not grey matter, axons (Wharton et al 2016). As oligodendrocytes die, virions are slowly released. Budding extracellular vesicles may also spread the virus. Astrocytes in areas of demyelination are hypertrophied (giant, bizarre, ballooned), with multiple mitotic figures that can be confused with astrocytomas. In mice transgenic for the JC virus large T antigen, there is dysmyelination with minimal expression of RNA for myelin basic protein and hyperproliferating astrocytes (100).

Lesions in AIDS-associated PML have more extended foci, more often involve gray matter and infratentorial regions, have fewer atypical astrocytes, and more often have perivascular infiltrates than in non-AIDS forms.

In a treated patient with multiple sclerosis who developed PML, autopsy showed few cells in the cerebral perivascular/Virchow-Robin spaces, no CD4 cells, rare dendritic cells and macrophages (elevated without natalizumab), but some CD8 cells and increased MHC class I (267). In relapsing-remitting multiple sclerosis with natalizumab-associated PML with an IRIS reaction, JC virus was present in white matter and neocortex, and lesions had sharp borders containing CD4 and CD8 T cells (298). In contrast, the older multiple sclerosis lesions in this patient were separate and distinct from adjacent PML lesions and were hypocellular, with myelin loss, axonal preservation, and gliosis but no JC virus.

The CSF of PML cases in AIDS and multiple sclerosis patients is usually nearly normal, with occasional mild pleocytosis (fewer cells during immunotherapy or AIDS), slightly increased protein, and, rarely, increased IgG. Brain biopsy with histopathology and electron microscopy, using in situ hybridization or DNA polymerase chain reaction, is the gold standard but is not perfect. Laboratories vary in ability to detect CSF JC virus by polymerase chain reaction. The limit of detection is 500 virus DNA copies in many hospital labs (in which only approximately 75% of multiple sclerosis PML cases are positive), 50 copies in commercial tests, and 5 copies in the E Major NIH lab (99% detection).

Virus infections (257), and perhaps cancer, are less prevalent than expected in multiple sclerosis, possibly from hyperactive immunity (see MedLink Neurology article Multiple sclerosis: neuroimmunology). PML was usually not suspected in multiple sclerosis and had not been reported before natalizumab-associated cases, although white matter lesions may have been misdiagnosed (148). There are no cases of PML during 2 million patient years of interferon beta therapy for multiple sclerosis, often in combination with steroids and chemotherapy. However, 2 cases appeared in combination with natalizumab in the pivotal trial.

Pathogenesis of PML. PML arises when there is a defect in CNS immune surveillance, ie, few T cells in the brain, possibly in conjunction with peripheral replication and activation of the JC virus. It is likely that natalizumab (1) interacts with interferons or immunosuppressive drugs to affect CNS immune surveillance, (2) induces activation of the JC virus, or (3) causes redistribution of cells infected with the JC virus. Other immune-privileged sites (testes, ovaries, and eye) are apparently unaffected. The dearth of other infections during therapy with natalizumab, plus the usual absence of PML in multiple sclerosis, even after prolonged glucocorticoids, chemotherapy, and stem cell transplantation, suggests a unique phenomenon and that abrogation of immune surveillance is not the only cause.

PML was first detailed in association with leukemia and lymphoma (11), then in autoimmune disorders, transplants, sarcoidosis, idiopathic CD4 lymphocytopenia, connective tissue disease, and more recently in AIDS. Earlier reports go back to Hallervorden in 1930 (101). Approximately 5% of cases were "primary” PML, with no associated immunosuppression. PML appears after solid organ and bone marrow transplants. The rate is 1.24 per 1000 per year after heart or lung transplants (177). PML patients are usually immunosuppressed, and most PML patients have impaired cell-mediated immunity (Herndon R 2005, personal communication). It is currently assumed that the increased incidence in connective tissue disease is from chronic immunosuppressant therapy. However, 40% of PML cases in systemic lupus erythematosus had minimal or no prior immunosuppressive treatments. Other factors in systemic lupus erythematosus and perhaps neuromyelitis optica, such as a strong “interferon signature” (83), could contribute to predisposition if high levels of interferons synergize with other factors (see “Type I interferons can potentiate the effects of natalizumab”) or alter immune regulation. There has been a large shift in the epidemiology of this formerly rare disease with the spread of HIV. Five percent of untreated AIDS patients developed PML. The percentage fell with widespread use of highly active antiretroviral therapy (HAART) for HIV, but AIDS is still the most common cause of PML.

JC virus and its target cells. PML is caused by the JC virus. This nonenveloped polyomavirus has a small 5000 bp circular double-stranded DNA genome. It is closely related to human BK virus and to primate SV40, a neurotrophic polyomavirus that causes CNS tumors. Primary infection and the chronic quiescent state are asymptomatic.

Evidence of prior infection with JC virus is present in 10% of 5 year olds and up to 80% of elderly patients. In the multiple sclerosis age range, serum antibodies to JC virus are present in approximately 55% of all people. The incidence of infection increases by approximately 1% per year (125). The incidence is high in Korea, greater in Europe than Japan, and lowest in the United States (296). Males are affected slightly more than females. There are 14 different JC virus phenotypes based on variation in the coding region. Types 1 and 4 predominate in Europe, and types 3 and 6 in Africa, and they can be used as markers of human migration. Infections with 2 different virus genotypes can occur. Genotypes have not yet been linked to virulence.

Transmission and the portal of entry are unknown but may be oral (oropharyngeal--tonsil stromal cells, gastrointestinal epithelial cells) or respiratory. The virus is typically passed from parents to children. Tonsillar stroma, bone marrow, and B cells contain JC virus DNA. The JC virus is often latent in renal tubular epithelial cells and is excreted in urine, especially in urine of people greater than 40 years old (168) and may contribute to transmission. Hands must be washed. The prevalence in semen and urine doubles in infertile males compared to fertile males. There are up to 1000 viral particles per ml in sewage water from divergent geographical areas (33). JC presence in public and private swimming pools is likely.

CD34 precursor B cells in the bone marrow contain latent JC virus. In culture, JC virus binds to all B cells but few T cells (288). Surface-bound JC virus seldom leads to infection or JC virus mRNA transcription (< 1% of B cells). Nonetheless, low-level productive infection is sustained by a reservoir of B cells in spleen, tonsils, and bone marrow.

JC virus is an extremely cell-associated virus, although serum occasionally contains free virus. It is not shed at high levels, suggesting cell-to-cell interactions are important in spread of the virus. JC virus has restricted tropism. It infects oligodendroglia (reducing myelin gene expression), astrocytes (at low levels), tonsillar stromal cells, B cells, granulocytes (a frequent reservoir), CD34+ stem cells in bone marrow and blood, and possibly choroid plexus cells. The archetype JC virus predominates in kidney uroepithelial cells: 30% of the population has JC virus in urine. One lab found roughly equivalent JC virus DNA levels in CD34, B cells, monocytes, T cells, PMN, and NK cells; all cells contained much more virus DNA than that present in the serum (53). Another lab found that JC virus DNA is absent in CD3+ cells and is restricted to CD19+ and CD34+ cells, which is relevant because natalizumab triples the number of CD34+ cells in blood (212). In vitro, JC virus infects vascular endothelial cells, kidney epithelial cells, amnion cells, and the Jurkat T cell leukemia line, but not primary T cells (172; 193). JC virus is usually latent and quiescent in B cells, bone marrow cells, the genitourinary tract, and occasionally in brain.

In the Virchow-Robin spaces and germinal center-like areas within the meninges, B cells with surface-bound JC virus may be held in check by T cells. In the brain, B cells in inflammatory multiple sclerosis lesions do not harbor replicating JC virus (03). JC virus DNA fragments present in normal brain oligodendrocytes and astrocytes indicate that the archetype and the pathogenic JC virus–Mad form both have full access to all regions of the brain (211). Brain presence of JC virus is altered little by immunosuppression. This indicates that the normal, and even the partially compromised, immune system efficiently and inevitably clears all forms of this virus.

Peripheral infection or reactivation of the virus is followed by hematogenous spread to the brain. B cells may act as “Trojan horses” to carry virus into the brain. Infected B cells can transmit the virus to human fetal glial cells in culture. Fetal glial cells and B cells share common DNA binding factors that bind to the JC virus regulatory sequence and allow more efficient replication of the virus than mature B cells. JC virus infection of astrocytes enhances transcription of hundreds of genes and suppresses a third of that number. These regulated genes affect cell proliferation, signaling, and inflammatory responses.

The virus becomes activated and neurotrophic in white blood cells during natalizumab therapy. B cells contain recombinases that could rearrange the virus genome. The JC virus archetype is present in normal brains at low levels, and the pathogenic mutation may arise in several brain areas from separate JC variants based on rearranged noncoding control regions (NCCRs). In order to become pathogenic, the NCCR of the archetype virus rearranges to the neurotropic Mad strain, with multiple mutant variations. The NCCR is a bidirectional element that contains promoter and enhancer elements and the origin of viral replication. Viral replication in oligodendrocytes and astrocytes is normally inhibited by the SF2/ASF splicing factor (SRSF1 gene), which blocks JC virus RNA splicing and suppresses the NCCR. SF2 is unable to bind to some Mad mutations. The VP1 coding region also frequently mutates in PML, allowing escape from immune control by anti-VP1 T cells and antibodies. Natalizumab-treated and non-treated patient groups have similar rearrangements.

Natalizumab therapy induces many B cell activation and differentiation markers (167), so it is possible that high levels of some of these transcription factors (POU domain family, Spi-B, NF-1X, Oct1, and Oct2) will enhance JC virus replication (115; 171); levels are highest in natalizumab-treated patients who develop PML. NF-1X and Spi-B are upregulated in glial, hematopoietic progenitor, and B cells. Spi-B binds to TATA boxes in the JC virus Mad-1 and -4 mutants, but not to this site in the (non-PML) archetype. After the Mad mutations appear, susceptibility is not at the level of receptor binding but rather is molecular, at the level of these intracellular transcription factors (Major E 2009, personal communication).

Viruses related to JC virus bind to several cell surface proteins for internalization. The VP1 protein of human BK virus binds to gangliosides GT1b and GD1b with alpha-(2,8)- and alpha-(2,3)-linked sialic acids, but not to VLA-4. Murine polyoma virus binds to alpha-(2,3)-linked sialic acid, followed by virus binding to VLA-4, and then retention on the cell surface (288) or internalization through non-clathrin coated vesicles (48). Conformational changes in VLA-4 could affect virus binding.

The virus attaches to the sialylated glycoprotein and can then use the serotonin 5HT2A, B, and C receptors for rapid clathrin-dependent entry and internalization. 5HT2 receptors are present on neurons and are also highly expressed on brain microvasculature and on the abutting astrocytes at the blood-brain barrier. There are also 5HT2 receptors in the area postrema and the choroid plexus (79; 106). Interferon beta could indirectly affect serotonin 5HT2A receptor expression by inducing indoleamine 2,3-dioxygenase, an enzyme that decreases tryptophan and serotonin. Low tryptophan levels cause a compensatory increase in 5HT receptor expression. This potentially enhances JC virus binding to neurons, astrocytes, and lymphocytes, as all express 5HT2A receptors.

Internalized virus activates the MAP kinases, ERK1, and ERK2 (79). B cells, some T cells, and oligodendroglia and astrocytes express alpha-(2,6) sialic acids on glycans (75). A later study found sialic acid sites on microglia, monocytes, choroid plexus epithelium, meninges, and renal distal tubules and collecting ducts but none on astrocytes and oligodendrocytes (106). This argues that oligodendrocytes and astrocytes internalize JC virus through an unknown pathway. Once internalized, JC virus could spread to other cells through extensive astrocyte syncytia and gap junctions.

The JC virus large T antigen is detected in oligodendrocytes, astrocytes, and selected neurons such as cerebellar granule cells, but is seldom detected in kidney (20) or microglia. JC virus T antigens hijack cell cycle machinery, block p53 and Rb, and force cells into an S-phase-like state (22). Infected cells are maintained in an immature, virus-producing state, avoiding apoptosis. Large amounts of T antigen are a marker for viral reactivation and lytic infection. Certain CNS cells have high concentrations of the DNA-binding proteins that are necessary for lytic infection. The transcription factor Tst-1 (aka Oct-6) is found only in oligodendrocytes and Schwann cells; it stimulates transcription of early and late virus genes. NF-1X must be in excess for virus replication. Other cells that are somewhat susceptible to lysis are astrocytes and CD34+ and CD19+ B cells.

There are 2 types of JC virus regulatory region: the nonpathogenic archetype found in kidney cells and the rearranged neurotrophic form with repeated activation sequences. The kidney-resident archetype regulatory sequence has only one 98 bp sequence plus 23 and 68 bp insertions. The archetype is linked to nonlytic latency and possibly dissemination, is present in B cells and kidney cells, and is associated with a better clinical outcome (244). Inside the cell, c-Jun and NF-1 class X protein, NF-1X (high in tonsillar stromal cells and glia, sites of JC virus infection) bind to adjacent sites on the JC virus regulatory region. In the prototype Mad-1 variant, the 200 base-pair JC virus regulatory region contains a 98 bp tandem repeat with 1 c-Jun and 2 NF-1 binding sites and 2 TATA sequences. Spi-B also binds only to Mad-1 and Mad-4 variants. The neurotropic Mad variants appear in infected brain or tonsil tissue, and their presence correlates with poor clinical outcome. Transfection of NF-1X into neurons makes the neurons susceptible to JC virus infections. Spi-B binding sites in the tandem repeats are increased in all virus isolates from cases of PML. Spi-B is upregulated in B and CD34+ B cells of multiple sclerosis patients treated with natalizumab, before they enter the CNS (173). Spi-B is also increased in CD4 and CD8 cells, which do not harbor the virus.

Anti-VLA-4 therapy mobilizes JC virus–infected B cells and progenitor cells in large numbers and provokes release into the peripheral blood. Activation of JC virus by anti-VLA-4 antibodies is also likely (ibid). Mobilization is followed by preferential ingress of activated virus to the CNS or is simply an effect of mass action from large numbers of JC virus–containing cells. Reduced penetration of anti-JC virus cytolytic and memory T cells into the CNS during VLA-4 blockade may set the stage for PML. The cells that do penetrate the BBB are anti-inflammatory (138).

Activation of the JC virus by HIV and other viruses. Five percent of patients with AIDS develop PML (24). There are approximately 7500 AIDS/PML cases per year in the United States. This incidence is higher than in other disorders with immunosuppression, suggesting that HIV has unique interactions with JC virus.

HIV could provoke PML through (1) the paucity of T cells in the blood and CSF from severe AIDS immunosuppression, (2) an increase in activated circulating (JC virus positive) B cells, CD34+ stem cells, and granulocytes, (3) erosion of the integrity of the blood-brain barrier, (4) activated CNS perivascular macrophages, which could increase adhesion molecule expression on adjacent endothelial cells, (5) upregulation of secreted cytokines that affect JC virus genes (not yet demonstrated), (6) HIV tat protein transactivation of the JC virus (24; 245), and (7) HIV and natalizumab both binding alpha4beta7 gut integrins, potentially affecting barriers and immune cells.

(1) Lymphocytes in the CSF decline markedly during natalizumab therapy and during HIV infection. JC virus–infected circulating B cells appear in 40% of immunosuppressed patients. During AIDS, immune surveillance by T cells is lost in the brain, possibly paralleled by a rise in the number of JC virus–infected cells in blood and brain. Bone marrow cells are positive for JC virus in 47% of HIV-positive subjects, but only 13% of HIV patients (272). Moreover, infection with 1 virus changes the ecology of the CNS and may facilitate passage of other viruses into the brain (07).

(2) VLA-4/VCAM-1 adhesion helps retain immune cells in the bone marrow (207). Hematopoietic progenitors are mobilized by blocking VLA-4/VCAM-adhesion. JC virus–positive cells could increase in the blood if there is more release of cells from the bone marrow in HIV, or in multiple sclerosis after natalizumab.

(3) Serum proteins accumulate in white matter glia and neurons, indicating blood-brain barrier breakdown in AIDS (220). There is seldom gadolinium MRI enhancement in AIDS PML. This suggests that endothelial cells are not activated during HIV infection, and that endothelial cell inflammation would not affect leukocyte migration.

(4) The perivascular space in multiple sclerosis contains large numbers of activated macrophages, which could affect migration and adhesion of other immune cells to endothelium. In PML lesions, perivascular B cells and surrounding astrocytes express JC virus DNA and protein (116), suggesting hematogenous spread is followed by infection of astrocytes. The exact sequence and interacting molecules are unknown.

(5) CNS inflammation in multiple sclerosis should enhance antiviral immune mechanisms yet could also modify the JC virus life cycle.

(6) In glial cells, the HIV tat protein directly transactivates the tat-responsive cis-element in the JC virus regulatory region (270). In AIDS/PML, this noncoding control region loses sites for the SP1 transcription factor, but there is duplication of the up-TAR site that binds the HIV tat protein (190). HIV and JC virus generally infect separate cell lineages. Tat activation of the JC virus genome is not relevant in HIV-negative multiple sclerosis, unless endogenous retroviruses are a component of MS pathogenesis.

(7) The gp120 HIV envelope protein binds to integrin alpha4beta7 on T cells (10). This activates LFA-1 on T cells and facilitates cell-to-cell spread of HIV-1. Natalizumab, however, does not block replication of HIV in cultured T cells (210).

During AIDS-associated PML, JC virus is found in CSF (100% positive for JC virus DNA by nested polymerase chain reaction; 100% show active JC virus VP1 mRNA transcription, in some labs), urine (56% JC virus DNA/38% mRNA – ie, weak predictive value), blood plasma (24%/0%), B cells (20%/0%), and non-B cells (12%/ unknown%) (160). However, JC virus DNA detection in other labs is less sensitive (76% with real-time polymerase chain reaction) (36).

The SV40 large T antigen is a tumor promoter. The related large T antigen of JC virus is potentially tumorigenic. Expression of the JC virus early promoter is higher in glial cells than in nonglial cells (245), suggesting that tumors could arise from glial cells. JC virus is tumorigenic in primates and hamsters, causing astrocytomas, glioblastomas, medulloblastomas and abdominal neuroblastomas, primitive tumors, and pinealomas. JC virus has been associated with human oligodendroglioma, glioma, oligoastrocytoma, colon cancer, and medulloblastoma (245). But there is no solid evidence for a causal role of JC virus in human tumorigenesis. Tumor development is countered by the immune system when polyoma virus VP1 binds to sialic acid groups on immune cells and induces toll-like receptors to produce IL-12 and activate T cells.

Infections with viruses such as Epstein-Barr virus could affect the JC virus. Epstein-Barr virus resides in B cells, activates B cells, and is linked to the development of multiple sclerosis. Synergy between JC virus and Epstein-Barr virus is possible. Sixteen of 23 patients with primary CNS lymphoma expressed DNA for the Epstein-Barr virus-transforming protein, LMP1; 5 also expressed JC virus proteins, including T antigen (68). In multiple sclerosis B lymphoblasts, recombination between JC virus and EBV can take place in the JC virus archetype noncoding control region (297). Cytomegalovirus infections activate BK virus (related to JC virus) and the JC virus (294). HHV-6 and cytomegalovirus can increase levels of JC virus in JC virus–infected glial cells (299). Finally, HTLV-I tax protein activates the JC virus promoter.

Are other viruses perturbed by anti-VLA-4 agents? In 36 multiple sclerosis patients, BK virus in urine was reactivated from 8% at baseline to 22% on natalizumab therapy (169). JC virus also rose in urine during therapy, from 19% to 63% in 19 patients (52). In contrast, JC virus DNA in blood (28%) and urine (25%) was the same as in interferon beta-treated multiple sclerosis patients and healthy controls (53). With natalizumab treatment, however, the rate of becoming JC virus–positive increases 10-fold, suggesting there is more virus available to generate an immune response. Non-polyoma viruses are also affected by natalizumab. Human herpesvirus-6 is reactivated during natalizumab therapy, leading to elevated anti-HHV-6 antibodies in serum and occasionally HHV-6 DNA in CSF (299). Anti-VLA-4 antibodies (GG5/3), given 14 and 18 days after Borna disease virus inoculation, prevent most inflammation (238). However, viral titer and Borna disease virus protein distribution in the brain are not affected. This suggests that anti-VLA-4 reduces T-cell migration to prevent inflammation. In Semliki forest virus-mediated demyelination in mice, anti-VLA-4 antibodies also decrease inflammation but do not change virus clearance (259). Finally, rotavirus, which causes childhood diarrhea, has viral capsid proteins that bind VLA-4 (61). This observation suggests a role of rotavirus in Crohn disease-related PML. However, there is no clear link between rotavirus infection and Crohn disease, and natalizumab did not provoke significant gastrointestinal symptoms in Crohn disease trials.

PML and multiple sclerosis therapies are also extensively reviewed elsewhere (187; 60).

Topics 13 and 14

(13) Direct detection and monitoring of JC virus and other polyoma viruses.

Detection of JC virus in blood and brain is difficult, and its success is dependent on the technique used. Detection frequency for JC virus DNA in white blood cells is 25% with standard polymerase chain reaction, 55% with Southern blots, and nearly 100% with nested polymerase chain reaction (71). In this study, detection of the presence of JC virus DNA was equivalent between healthy controls and patients with HIV and PML, but the amounts of viral genome were higher in PML. Lack of utility of JC virus DNA as a predictor of PML was confirmed in 13,000 samples from 1400 patients in natalizumab clinical trials, which showed 12 patients positive for JC virus in plasma, 0 positive in mononuclear blood cells, and 25% positive in urine before and after therapy (240). Twenty-five percent of adults will shed virus in their urine during natalizumab therapy. JC virus DNA is found in genitourinary tissues in 75% of donors with high JC virus antibody levels, 13.3% with low levels, and 0% of those seronegative. It is detected in 45% of other tissues in those with high JC virus titers, 2.2% with low JC virus serostatus, and in 0 seronegative persons (29).

A positive CSF polymerase chain reaction for JC virus DNA strongly suggests the diagnosis of PML in the right clinical setting. Larger PML lesions yield more JC virus DNA, although the anatomical location is not linked to a positive test, and 16% of large PML lesions on MRI test negative for CSF JC virus DNA. Half of the patients with negative CSF testing will become positive within a month. Conversely, JC virus can be detected, absent PML; 9% of multiple sclerosis CSF samples were positive for JC virus DNA, yet no noninflammatory and inflammatory CSF controls were positive (87). Other authors have failed to find JC virus DNA in multiple sclerosis CSF (34), even during PML. During HAART for AIDS, PML stabilizes when CSF JC virus levels fall (36). Antibodies to JC virus are not elevated in untreated multiple sclerosis CSF or blood in the majority of reports. JC virus infections can generate oligoclonal bands specific for the JC virus VP1 capsid antigen and will elevate the CSF antibody index. CSF IgG to VP1 is present in 75% of PML cases, but only in 3% of controls. There is intrathecal synthesis of anti-VP1 antibody in 76% of HIV+ patients with PML, but only in 3% of other neurologic disease controls. This antigen is also a target for cytolytic T cells.

Serum antibodies to the virus indicate prior infection. Antibodies to the VP1 capsid protein are used for epidemiologic studies. The 2-step STRATIFY JC virus antibody test is described below.

DNA monitoring is better developed for BK virus than for JC virus. BK virus is a closely related polyomavirus that infects and is latent in the kidney. It also infects bladder, prostate, cervix, vulva, lung, eye, liver, and brain. BK virus infects 90% of the healthy population at early ages; JC virus is acquired at about 1% per year. If a person is seropositive for BK virus serotype 1, he or she is less likely to be positive for JC virus, indicating that immunity against BK virus is partially protective against JC virus. BK virus is found in some CNS tumors but does not cause PML. BK virus DNA increases in the urine during HIV infection; JC virus does not. BK virus is increased in lupus; JC virus is not. BK virus causes nephropathy following renal transplants and immunosuppression (polyomavirus-associated nephropathy, PVAN). After kidney transplantation, BK viremia is the first sign of active virus replication and is present in 45% of renal transplant patients at 1 year. Viremia (15%) progresses to nephropathy (5% to 10%) and loss of the transplant (5% of all patients; 70% of patients with nephropathy) (38). Polymerase chain reaction is sensitive in monitoring for viruria, but a renal biopsy is the definitive test. Drugs used to prevent transplant rejection suppress the immune response to BK virus, allowing viremia (Table 3). Lowering the dose of immunosuppressive drugs clears viremia in 95%, with an average time to clearance of 54 days.

BK virus studies suggest that reversing immunosuppression in PML could also clear JC virus. For instance, removal of natalizumab with plasma exchange leads to an immune response to JC virus (see IRIS, below).

Serum samples during natalizumab therapy were positive for JC virus DNA in 2.3%, a rate comparable to that in normal controls (153). Complementing this, the anti-JC virus response by cytolytic CD8 cells in multiple sclerosis patients is much stronger than normal, even though virus DNA is not detectable in blood cells, plasma, and CSF (72).

Table 3. Immunosuppressants Allow BK Virus Replication

BK virus DNA can be used as a “best case” model for detection of JC virus genomic DNA. Unfortunately, variability in assays for JC virus (antigen source, methods, polymerase chain reaction primers, and conditions) leads to differences in epidemiology, infection in multiple sclerosis, and prognostic indicators between studies. Moreover, there is no perfect primate or rodent model of JC virus infection.

There is no early warning assay that detects JC virus DNA in the development of PML, unlike BK virus expression in urine. JC viremia, present in up to 2% of healthy subjects, is only weakly associated with PML, but it does indicate there has been exposure to the virus. In most studies, natalizumab does not increase urine JC virus levels, but those with high anti-JC virus index do have higher titers (291). Anti-JC virus titers may reflect intermittent free virus, but not the complete antiviral picture. With help from CD4 cells, cytolytic CD8 T cells easily clear JC virus–infected cells in most people. Some clinicians routinely test for JC virus in serum, or measure the CD4/CD8 ratio, but these tests can’t be recommended for general use. The most reasonable peripheral tissue for JC virus detection may be white blood cells to capture putative JC virus–infected B cells and CD34+ stem cells. With PCR of flow cytometry-purified cells for JC virus T protein, 50% of CD34+ cells were JC virus positive in multiple sclerosis and 20% were positive in healthy controls (212). Urine JC virus assays could be complimentary to the more sensitive serum STRATIFY-2 assay (below). Virus in urine in antibody-negative patients, however, suggests that there is no immune response to systemic JC virus and that the brain is not yet at risk. Monitoring for JC virus and PML during natalizumab therapy is discussed in topic 16.

(14) PML in multiple sclerosis and calculation of risk of PML based on prior treatments and anti-JC virus antibody. American and European guidelines for monitoring.

PML was not reported in multiple sclerosis until natalizumab therapy was introduced. The closest description was in a case of a woman with optic neuritis that was followed 12 years later by PML—after she developed AIDS (73). The optic neuritis, described as rapid onset, 1 week duration, with resolution over 4 days, was not followed by multiple sclerosis. Of interest, histopathology showed no optic nerve lesions, paralleling the case of PML that followed clinical optic neuritis in the pivotal trial of natalizumab therapy (148). MRI detection of PML is discussed in topic 16. Later reports include PML in a patient with common variable immunodeficiency and multiple sclerosis during treatment with IM interferon beta-1a.

Approximately 25% of natalizumab-treated multiple sclerosis patients will die. The remainder is evenly split between severe, moderate, and mild neurologic sequelae. Good prognostic factors in HIV-associated PML are high CD4 cell count, cytolytic cells specific for JC virus T or VP1 antigens, Gd enhancement on MRI, and low JC virus load in the CSF.

Investigators and patients were aware of a potential risk of CNS lymphoma and systemic infections with this experimental drug, but concern had been allayed because there were minimal side effects over 5 years of human trials. After the PML cases arose, some neurologists said this was expected, echoing the 2600-year-old philosophy of Epimenides of Crete, who declared that it was easy to predict past events. Administration of the drug was put on hold, as were important trials of natalizumab in primary progressive multiple sclerosis, Crohn disease, and rheumatoid arthritis, as well as other multiple sclerosis trials with small protein blockers of VLA-4/VCAM interactions. The drug was reintroduced in 2006 with a careful monitoring program (TOUCH program in the United States).

There have been 853 cases of PML during natalizumab therapy as of February 1, 2021. In the pivotal trials, 2 cases were on concomitant interferon beta-1a (after 28 and 37 natalizumab infusions). The increased risk was not statistically significant, but the combination is now avoided. (Possible synergy between these 2 therapies is described in section 10.)

PML during natalizumab therapy is not unique to multiple sclerosis. It developed in a Crohn disease patient treated with natalizumab in early trials, and there have been 3 more cases since. The first patient had chronically deficient hematopoiesis and had received years of treatment with immunosuppressive drugs, some contemporaneous with 3 natalizumab infusions. After a temporary stop in therapy, 5 later natalizumab infusions coincided with a rise in JC virus April genome and the development of PML. This infection in Crohn disease suggests that PML arose from chronic immunosuppression, possibly enhanced by natalizumab. Antibodies to alpha4,beta7 integrin, such as vedolizumab, are not associated with PML.

The risk of PML was roughly 1 out of 1000 in all natalizumab-treated multiple sclerosis patients based on early data, but with longer observation, the risk has increased. As of February 2021, there were 850 cases of PML in 220,700 natalizumab-treated multiple sclerosis patients (882,000 patient years): 489 cases in Europe, 235 in the United States, and 129 in the rest of the world (Biogen Idec MedInfo). The average global risk is 3.86 per 1000 patients (1/259) in all treated patients, weighted for shorter duration of infusion because the greatest number received 0 to 12 months. There are exceedingly few cases of PML in the first year of therapy: 0.05 per 1000 (1/20,000) (Biogen Idec MedInfo). However, the rate rises to 0.62 out of 1000 during 13 to 24 months of infusion therapy (cumulative 0.67/1000), and it is 1.22 for 25 to 36 months (cumulative 1.89/1000), 1.63 for 37 to 48 months (cumulative 3.52/1000), 1.60 for 49 to 60 months (cumulative 5.12/1000), and 1.35 for 61 to 72 months (cumulative 6.47/1000 or approximately 1/155). JC virus–negative patients have no or very low risk after prolonged therapy. In 2015, European patients had more prior to chemotherapy (24% vs. 14%), but the frequency of PML in those with no prior chemotherapy was also increased 4-fold in Europe. Provocative chemotherapy was predominantly mitoxantrone, and to a lesser extent, azathioprine, cyclophosphamide, and methotrexate (17:5:5:5 cases, respectively, for these agents). Thus, given a PML incidence of 4.03 out of 1000 treated per year (1 out of 248) after 7 years of therapy, the risk of PML is 1 out of 77 in U.S. patients with prior chemotherapy and 1 out of 167 in chemotherapy-naive patients. The PML rate initially increased over time due to a slightly higher number of patients with prior chemotherapy and a greater proportion of patients on more than 2 years of therapy (see risk calculations, below) but has now declined with use of the JC virus index.

Antibodies to JC virus reflect the risk of developing PML during natalizumab therapy. Approximately 55% of the multiple sclerosis population has antibodies to the JC virus; those who do not carry the virus can’t develop PML. If a patient has a positive titer, risk is doubled to approximately 1 in 141 with prior chemotherapy but is 1 in 704 in chemotherapy-naïve patients. Antibody-negative patients have no risk unless tests are false-negative (2.2% chance with the new 2013 second-generation ELISA; seen in perhaps 1% of cases with PML, but clinical data are vague; 5 known in the 850 cases above) or they convert to positive. The rate of conversion is approximately 1% per year in untreated patients, but 10% per year in natalizumab-treated multiple sclerosis patients (09; 226; 125; 250). All risks calculated above are biased downward by the bulk of patients who have lower time of exposure. The estimates above are based on average risk that includes the first nearly safe year; the risk after 24 months of therapy rises from 2.84 (1 in 352) to 5.02 (1 in 199).

Anti-JC virus antibody titers in serum further stratify the risk of developing PML. High titers increase the chance of developing PML and are not a marker of protection from the virus. High anti-JC virus titers suggest an antibody response to ongoing virus exposure, perhaps to large numbers of the virus. Virus levels could be high from ineffective cell-mediated immunity or from direct activation of the virus by natalizumab.

The 2-step STRATIFY JC virus antibody test is more sensitive than hemagglutination assay and has 2.2% false negatives, an improvement over 2.8% in the prior version (216). Titers of 0.2 to 0.4 are indeterminate, and this JC virus index must be confirmed with a second assay. Titers above 0.4 are positive. High anti-JC virus titers increase the risk of developing PML by 10-fold compared to low titers. In patients never exposed to immunosuppressants at the 2-year point, positive anti-JC virus titers below an index value of 0.9 connote minimal increased risk (1 out of 3000). From 0.9 to 1.5, the overall risk is 1 out of 1000, and above 1.5 the risk is 1 out of 120 per year (topic 13). Those with positive titers prior to immunosuppressant therapy, and for longer than 2 years of therapy, have a risk of 1 out of 90. Titers in patients with prior immunosuppressant use were similar between patients with and without PML, suggesting different mechanisms in patients naive to or exposed to immunosuppressants (216). Another analysis estimates that after 2 years of natalizumab therapy, the risk with negative titers is 1 out of 125 per year, but with positive anti-JC virus titers is 1 out of 44 per year (27). Because the risk in Europe is 2.5-fold greater than in the United States, an index threshold of 0.6 has been suggested for Europe. The index does not change during pregnancy. Putting these numbers into perspective, daily life also has risks (125). The U.S. National Highway Traffic Administration states that car accidents happen to 1 out of 62 people per year and cause death to 1 out of 7575 per year.

Risk per year is cumulative over time. For instance, a risk of 1 out of 1000 over 5 years will become a risk of 5 out of 1000 (1 out of 200). Thus, calculations must be based on PML incidence by “treatment epoch.”

Cumulative risk estimates, published in MedLink for the prior 5 years, were confirmed with titer stratification (114). After 6 years of therapy, the risk was 1.6 out of 1000 with a titer index of less than 0.9; was 8.5 for an index of 0.9 to 1.5; was 28 for an index of greater than 1.5; and was 27 for patients with prior immunosuppressant use.

The risk of PML is twice as high in Europe and the rest of world versus the United States, perhaps because of more frequent use of immunosuppressants, different virus strains, or more virus exposure. Patients in the United States were 7 years older than those in Europe. Patients more than 50 years old are also at a higher risk, likely from immune senescence (224). The TOUCH monitoring program in the United States could aid in the early detection of PML symptoms. Risk stratification with anti-JC virus titers may lower the risk by removing many “at risk” patients from therapy, but this effect may be obscured by a higher incidence of PML from prolonged duration of natalizumab therapy (27).

Patients are more likely than physicians to accept high risks of PML (111). However, there is a broad spectrum of patient and doctor behavior, which is also influenced by different cultures.

The influence of glucocorticoids is not included in the risk calculations for multiple sclerosis. This commonly used intervention for exacerbations impairs JC virus–specific T cell responses (08). PML may be more frequent in sarcoidosis with glucocorticoid therapy (123). Steroids should not be used lightly for multiple sclerosis exacerbations during natalizumab therapy.

PML is occasionally seen in connective tissue diseases, often on a background of prior chemotherapy, and in patients with lymphoid cancers or transient lymphopenia. Drugs linked to the development of PML are natalizumab (reporting odds ratio [OR] of 35, likely accurate because of the TOUCH program), efalizumab/anti-LFA-1 (OR 27, no longer on the market), rituximab/anti-CD20 (OR 23, but more likely to be underreported), cyclophosphamide (8), azathioprine (6), tacrolimus (4), mycophenolate (3), methotrexate (2) (249), eculizumab/anti-C5a (1), and brentuximab vedotin/anti-CD30 and anti-mitotic, ruxolitinib/JAK1/2 inhibitor. In patients never on natalizumab, dimethyl fumarate therapy is linked to PML in psoriasis, and there are several cases in multiple sclerosis patients. There are also several cases with fingolimod therapy. There is a possible case with teriflunomide, after natalizumab therapy. PML risk is difficult to calculate because few patients have been on long-term therapy with these drugs. Older patients are more likely to develop PML after switching from natalizumab to another therapy. There are 2 cases of PML with ocrelizumab. One followed a switch from fingolimod, with a preexisting PML lesion then steroid therapy for “MS worsening.” The other was a therapy-naive 78-year-old man with progressive multiple sclerosis for 30 years, JC virus index of 2.5, and lymphopenia who developed PML 2 years after starting ocrelizumab (209).

Guidelines for PML monitoring. The American Academy of Neurology recommendations with respect to diagnosing PML are as follows. A diagnosis of PML can be made by brain biopsy or by a combination of clinical observations of a progressive course of new symptoms, PML-consistent white matter lesions on MRI, and JC virus in spinal fluid. Progressive neurologic symptoms should occur on a background of therapy with certain disease-modifying drugs. MRI lesions are high-signal intensity white matter lesions usually near the gray-white junction on T2 or FLAIR ± enhancement ± mild mass effect. The CSF JC virus DNA is detected with sensitive polymerase chain reaction assays.

In Europe, recommendations are to test for antibodies to the JC virus before therapy and every 6 months on therapy (179). MRI scans should be performed every 12 months when titers are negative, every 6 months when titers are less than 1.5, and every 3 to 4 months when titers are greater than 1.5. Additionally, MRI scans should be performed at the end of natalizumab therapy and 3 months after treatment. MRI costs are much lower in Europe than in the United States, and cost and time in Europe are further reduced with a simple screening MRI with FLAIR and diffusion-weighted imaging. If PML is suspected, CSF JC virus polymerase chain reaction should be obtained, and it should be repeated if negative.

Serum neurofilament levels often rise with PML, on average 10-fold higher than baseline, and sometimes appear before clinical symptoms. There is a strong correlation between serum neurofilament levels and the size of the PML lesions. Serum neurofilament levels are a convenient and early marker of PML and, possibly, prognosis. Predictive algorithms are being generated. Plasma MMP-9 levels also increase in parallel to a rise in anti-JC virus antibodies. Natalizumab decreases intrathecal anti-JC virus antibodies, but not other antiviral antibodies; relation to PML is unknown.

The risk of natalizumab therapy must be balanced with its significant clinical benefit.

Topics 15 and 16

(15) Treatment of PML and immune reconstitution inflammatory syndrome (IRIS).

Immune reconstitution inflammatory syndrome (IRIS) as CNS immune function is restored. Progression of PML in patients with AIDS is slowed as the immune system is gradually reconstituted by highly active antiretroviral therapy (HAART). Young patients with high CD4 counts and low JC virus load who start HAART at the onset of PML do best. During AIDS-associated PML, gadolinium enhancement on MRI suggests there is activation of T cell-endothelial cell interactions and T cell penetration of the blood-brain barrier. Gd-enhancement correlates with improved prognosis. Twenty percent of AIDS patients develop PML after HAART is started (56). Even though the MRI shows enhancement and symptoms temporarily worsen, these patients do better than AIDS patients who do not respond to HAART. Mechanisms of this paradoxical “immune reconstitution inflammatory syndrome” (IRIS) include (1) a blossoming immune rebound by cytotoxic CD8 cells against preexisting subclinical PML (also seen with reactions against mycobacteria; reduced with temporary steroid therapy), (2) activation of the virus by cytokines, or (3) changes in cell trafficking with a bolus of new precursor cells from the bone marrow. Enhancing MRI lesions are seen in half of AIDS and PML patients responding to HAART and in HIV-negative patients, but only rarely in HAART nonresponders (117).

Reconstitution of CNS immunity should also clear the virus in multiple sclerosis patients with PML. When natalizumab is stopped, serum natalizumab levels fall by 80% at 42 days and to less than 1 ug/ml serum at 84 days. Because some patients describe a “wearing off” of natalizumab benefit at 3 weeks past the last infusion, it can be assumed that immune cells begin to penetrate the CNS at that time. However, complete reconstitution of CNS immune surveillance is likely to take several months after discontinuing anti-VLA-4 therapy (242; 126; 268), unless the effects of natalizumab are reversed.

IRIS is seen in nearly all cases of PML in multiple sclerosis patients (57). It appears spontaneously in nearly half of the cases. After removal of the antibody with plasma exchange in the rest, IRIS blossoms at approximately 4 weeks (range: 11 days to 8 weeks). IRIS leads to abrupt clinical changes, reflecting damage at multiple sclerosis lesion sites or in new areas, and it sometimes causes seizures and fever. There is an extensive CD8 T cell, plasma cell, and macrophage infiltrate in lesions and the surrounding white and gray matter in 4 of 5 biopsies (181). Others find more CD4 T cells. CD4 T cells produce IFN-gamma, and occasionally both IFN-gamma and IL-4, and use a broad spectrum of strategies to aid the CD8 cytolytic cells in destroying JC virus-infected cells. Cytolytic CD8 cells are present in 91% of PML survivors, but in no PML progressors (194). MRI does not easily discriminate between ongoing PML and IRIS. More than half of multiple sclerosis PML IRIS cases show Gd enhancement on MRI (vs. 30% in HIV IRIS) (57).

Treatment of PML in multiple sclerosis. Multiple sclerosis patients actually have fewer viral infections and possibly less cancer than normal, possibly because of an overactive immune system and altered interferon regulation. Viral CNS infections are rare in multiple sclerosis patients, despite treatment with interferons, glatiramer, and intense chemotherapy.

Natalizumab is the one of most effective current therapies for relapsing-remitting multiple sclerosis. However, it carries a risk of PML. Long duration of therapy, prior immunosuppression or hematologic cancer, or concurrent interferon therapy could increase the risk of PML. Blood and CSF screening for the virus as well as CD4/CD8 ratios have not been helpful in predicting PML. If PML develops, natalizumab therapy should be discontinued, serum natalizumab should be immediately removed (debated, next paragraph), and the virus infection should be treated. Modest enhancement of CNS immunity may be very effective.

Plasmapheresis (plasma exchange, PLEX) is often used, but of questionable efficacy. Typically, there are three 1.5–blood volume sessions over 5 to 8 days; natalizumab concentrations are reduced by 92%, and alpha-4 integrin on immune cells is desaturated (139). PLEX rapidly removes free antibody and allows the return of functional VLA-4 to the lymphocyte surface, enabling the immune system to resume its immunoprotective function in the CNS. As proof of concept, after HAART is begun in AIDS, there is a rebound in the T-cell count and partial improvement in immune function, and PML resolves to variable degrees. In multiple sclerosis, immune function is normal or above normal. Thus, complete removal of natalizumab should lead to protective (or even excessive) immune function in multiple sclerosis/PML, namely, IRIS. In an early case, soon after natalizumab was discontinued, the MRI began to enhance in a multiple sclerosis/PML patient who later improved (162). In PML during natalizumab therapy, plasmapheresis halves the time of onset of IRIS (279) and doubles the duration of IRIS (248). Moreover, PLEX removes anti-JC virus antibodies (Ko M 2018, personal communication). However, in 219 natalizumab PML cases, PLEX had no benefit (161; 264). This indicates that alternative or additional therapies are essential.

Leukocytapheresis could remove peripheral lymphocytes with bound anti-VLA-4 antibodies and mobilize bone marrow cells. This procedure needs to be tested for safety, monitoring whether it mobilizes bone marrow cells that contain JC virus.

Symptoms of PML decrease with high-dose steroids (1 g IV for 3 to 5 days, with slow oral taper), and this reduces some disability.

Antiviral therapy is often not effective in HIV-related cases of PML. In multiple sclerosis, blood-brain barrier disruption should allow penetration of some antiviral agents. There was a dramatic response to Ara C in 1 case of PML in multiple sclerosis (162). Of 5 immunosuppressed dermatomyositis patients who developed PML, 4 were treated with reduction of the immunosuppression plus Ara-C, and 2 survived (286). One patient with nearly complete recovery was also treated simultaneously with 30 mg of mirtazapine, a 5HT2 receptor antagonist, but this drug has not been effective in other patients. Cidofovir and interferon alpha have anecdotal benefit.

Teriflunomide, a multiple sclerosis therapy, strongly interferes with BK replication (Wilson J 2011, personal communication). It is less effective against JC virus, but actions against the mutated variants in multiple sclerosis CNS have not been studied. Leflunomide, the parent compound of teriflunomide, reduces BK-virus induced renal transplant rejection (158). Cidofovir, a treatment for cytomegalovirus, had only occasional benefit in PML.

Interferon beta effects against PML in multiple sclerosis remain uncertain, but it had a suggestive benefit in human PML in several reports. On the negative side, interferon beta prevents migration of antiviral T cells into the CSF. On the plus side, it is antiviral and prevents 95% of JC virus replication in human fetal glial cells (58). Type I interferons may not be curative because only 1/1000 of serum interferon beta enters the normal CNS (229). However, a strong interferon signature in mouse brain, 6 hours after interferon beta injection, indicates there is significant penetration into the CNS (97). CSF interferon beta levels have not been studied in multiple sclerosis, but CNS inflammation in multiple sclerosis may lead to blood-brain barrier leakage. Anti-interferon antibodies enhance polyoma virus infections after virus is inoculated peripherally in mice (103). In vivo, interferons reduce levels of serum JC virus DNA. In a study with much higher virus DNA levels in serum than in other reports, serum was positive in 29% of healthy controls and in 46% of untreated multiple sclerosis patients, and it fell to 14% in interferon beta–treated multiple sclerosis patients (66).

Systemic interferon alpha, 5 MU 3 times per week, and interferon beta have modest benefit in HIV/PML (118), but the benefit is erratic. Intrathecal interferon beta stopped decline and led to modest improvement in 1 case.

5HT2A receptor blockers are designed to cross the blood-brain barrier to treat depression. Risperidone at 4 mg twice daily, combined with paroxetine 40 mg twice daily to slow risperidone metabolism, may block JC virus entry into CNS cells and reversed PML in 1 patient (134). Olanzapine, ziprasidone, clomipramine, chlorpromazine, cyproheptadine, and mirtazapine (147) could also potentially block virus binding to brain cells. Studies of relative blocking ability are in progress (95; 26). Mirtazapine had some benefit in 4 cases of PML in HIV infections (49). Cyproheptadine may also be a blocker (79). All these compounds bind the 5HT receptors used for JC virus internalization (79). Reported benefits are underwhelming.

Maraviroc blocks chemokine receptor type 5 (CCR5) and is used to prevent HIV binding to immune cells. CCR5 controls cell migration, and maraviroc could, therefore, affect cell migration. There are case reports of improvement of IRIS with maraviroc, but others see no benefit (Stefoski D 2015, personal communication).

Mefloquine, an antimalarial agent, has in vitro activity against the JC virus. However, a trial in multiple sclerosis patients with PML was unsuccessful.

Ciprofloxacin decreases the load of the related BK polyomavirus in urine (164). It is unknown if this is a direct mechanism or is from this antibiotic’s ability to enhance immunity—a potential concern in multiple sclerosis, as floxins can trigger exacerbations.

To be effective against PML, agents should penetrate into the CNS, yet not suppress antiviral immunity. Steroids have no benefit.

One third of the cases of non-AIDS PML were initially reported to respond to cytosine arabinoside (Ara-C) (02). Later studies were not as positive (18); those who improved had residual deficits, and MRI improvement took more than 6 weeks. Topoisomerase I inhibitors (camptothecin and topotecan) are potentially helpful.

Chlorpromazine and clozapine synergistically block clathrin-dependent endocytosis of JC virus into glial cells (18); mianserin, ritanserin, and ketanserin also block infection of glial cells (204).

Mefloquine, an antimalarial, inhibits JC virus infectivity in vitro. A trial in PML showed no benefit. This drug has significant neuropsychiatric adverse effects.

Fusidic acid (Fucidin) has reduced viruria and stabilized allograft function in a renal transplant patient with JC virus–associated nephropathy.

Filgrastim (G-CSF) activates immunity and may have benefit in PML, without causing multiple sclerosis exacerbations; PLEX had no benefit (264).

IL-2 therapy led to sustained recovery in cases of PML with myelodysplastic syndrome (157) and Hodgkin lymphoma (40). IL-2, followed by pembrolizumab, a PD-1 blocker, has a possible additive effect.

IL-7 stimulates T-cell reconstitution. It may have had benefit in a patient with idiopathic low CD4 counts (05).

IL-7 plus vaccination with the VP1 JC virus capsid protein, plus imiquimod adjuvant is safe and had surprising benefit in 2 patients (262).

JC virus vaccines, small interfering RNA, and small molecules that would disrupt natalizumab-ligand interactions are in the conceptual stage. CRISPR/Cas9 can be used to induce mutations in the JC virus T antigen and to suppress viral replication.

Autologous or allogenic T cells specific for the JC virus controlled PML in 6 of 9 patients with hematologic cancers, and 5 were in good clinical condition (30). Anti-BK virus T cells may also treat PML.

Rarely, patients improve spontaneously (69).

(16) Can PML be predicted during natalizumab therapy, and can PML be prevented by natalizumab “drug holidays” or extended dosing intervals?

Therapy with natalizumab requires careful weighing of the risk–benefit ratio by well-informed patients and then close monitoring. This includes clinical vigilance, longitudinal neurologic exams by a multiple sclerosis expert, and a low threshold for any clinical signs of PML, plus a baseline MRI for comparison with later putative PML. In the U.S. TOUCH program, patients are quizzed for symptoms of PML before every natalizumab infusion.

PML is not predicted by monitoring peripheral white blood cells, CD4/CD8 cell ratios, or JC virus genome in blood, urine, or CSF. In untreated HIV patients, high JC virus DNA levels correlate with a worse prognosis (36). In HIV/PML, disease activity should be based on clinical, radiological, and virological criteria (CSF JC virus genome quantitation every 3 months) (55). Quantitative JC virus titers in multiple sclerosis do not correlate with severity of PML and are not useful. However, the titer of serum antibodies to the virus does determine the risk of developing PML.

Antibodies to JC virus reflect immune recognition of virus protein or nucleic acid. Elevated or rising titers suggest there is more virus present; this rise is missed by serum and CSF DNA assays. In patients never exposed to immunosuppressants after 2 years of natalizumab therapy, an index value less than 0.9 connotes minimal risk (1/3000), but from 0.9 to 1.5 the risk is 1/1000, and above 1.5 the risk is 1/120 per year.

There is a 10-fold increase in seroconversion from JC virus–negative to JC virus–positive during natalizumab therapy, in contrast to the usual 1% per year in untreated patients (277; 09; 250).

Markers that are not in wide use also predict the risk of PML.

Some neurologists perform yearly MRIs to detect PML before clinical symptoms appear. Perhaps 90 of 678 (13.3%) cases of PML in multiple sclerosis were picked up by MRI before neurologic symptoms (Biogen Idec MedInfo) (13; 70). In many of these asymptomatic cases, however, intensity of clinical monitoring and the diagnosis of PML itself is not clear (162; 166; 281; 32; 147; 214; 178). Conversely, a patient with at least one year of PML (or progressive multiple sclerosis) symptoms had no brain or spine MRI changes, and other cases have been reported with clinical symptoms and negative MRI (151). Nonetheless, MRI is essential to differentiate between multiple sclerosis and PML when symptoms appear. Patients with asymptomatic PML are more likely to recover after natalizumab is discontinued.

Monitoring for PML is arguably not cost-effective. An MRI ± gadolinium at academic medical centers in the United States is approximately $7000, although FLAIR + DWI fast screening protocols of PML could be equally sensitive and much cheaper. Thus, the cost of detecting one case of PML by MRI a few weeks earlier than by clinical symptoms is $7000 x 1000 (1/1000 risk per year) x 12.5 (8% detected by MRI before clinical symptoms) = $87,500,000.

The posology or dosing of therapy could theoretically be adjusted to reduce the risk (not abrogate risk) of PML by less frequent infusions, pulsed therapy, or drug holidays. Some suggest that a drug holiday after 1 to 2 years might reduce the risk of PML in multiple sclerosis if the drug is later restarted. However, this introduces the risk of exacerbations at 3 months. Extended dosing, from 28 to approximately 36 or 55 days was safe in 300 or 600 patients (302). In this study, there was no increase in exacerbations and likely no excess of MRI lesions. Four cases of PML occurred; all were in the every-28-day group, with none in the extended-dosing cohort. Abstracts show a reduction in risk of one half to one third with 35- to 45-day intervals versus 30- to 31-day intervals (Foley J 2018, personal communication).

It is important to remember that these patients still have multiple sclerosis and that removal of effective therapy will allow active multiple sclerosis to return. The same arguments apply to patients who have serum antibodies to JC virus. Those who do not have antibodies to the virus are at very low risk, and drug holidays in them are not logical. Low pretherapy disease activity, a fall in CSF B cells and immunoglobulin G and M synthesis during treatment (284), and higher natalizumab-induced blood lymphocyte counts, controlled by Akt-regulated apoptosis (236), predict there will be fewer flares after the drug is stopped. Importantly, discontinuation of natalizumab to prevent the 1:100 to 1:000 chance of developing PML will almost certainly cause more multiple sclerosis activity unless another effective agent is started.

When natalizumab is discontinued, some physicians delay the start of therapies that reduce CSF immune cells (fingolimod) or suppress immunity (chemotherapy, glucocorticoids, possibly teriflunomide and fumarate). However, the risk of rebound multiple sclerosis exacerbations becomes more likely the longer a second therapy is delayed (discussed in topic 8). The risk of developing PML declines after cessation, and PML is very rare 6 months after the last dose.

A washout of other therapy before starting natalizumab must also balance the risk of transiently combining natalizumab with immunosuppression or altered immune regulation (interferon beta, glatiramer, oral agents), versus the risk of removing all therapy from a patient experiencing disease activity. A washout period of 3 to 6 months after chemotherapy and less than 6 weeks after interferon beta was suggested by an expert panel (98; 130). However, many patients contemplate a switch from other therapies because of disease activity and are actually partially treated. This author has seen rapid clinical worsening within 2 weeks of withdrawing interferon beta and glatiramer and recommends only several days or no gap in therapy. Immune compromise should be ruled out before starting natalizumab, with a complete blood count and evaluation for opportunistic infections.

Combination of natalizumab with interferon beta or chemotherapy, or even glucocorticoids, may be unsafe. The 2 PML cases in the pivotal trials were in the interferon beta-1a/natalizumab combination group, but numbers are too small for confidence. Past chemotherapy is a significant risk for PML; concurrent therapy has potential risks. Glucocorticosteroids are immunosuppressive, but there is no obvious danger signal at this point in combination therapy.

(17) Natalizumab withdrawal and other strategies to prevent multiple sclerosis reactivation

Rebounds of multiple sclerosis after stopping natalizumab are possible. Angry immune cells are kept out of the brain by natalizumab, and activated white blood cells increase during therapy (topic 5) (90). Could CNS inflammation from multiple sclerosis rebound when the drug is discontinued? There was no rebound in the large pivotal drug trials. Clinical and MRI activity returned 3 to 6 months after drug was discontinued, but only to near pre-natalizumab rates (202). Controlled drug holiday studies showed the same result, but patients taken off drug had much more multiple sclerosis activity (28%) than those who remained on therapy (0%) (292). When the drug is withdrawn, cognition declines back to baseline by 1 year. CSF immune cell levels recover in 3 to 6 months (126; 268), and disease activity sometimes reappears at 6 weeks (104). Nonetheless, series by experienced clinicians from single sites, however, suggest a subset of patients will have severe multiple sclerosis symptoms after withdrawal (189). Many patients are highly active before therapy, so when they return to pretreatment levels of activity, a subset will seem do poorly. MRI Gd-enhancing lesions also increase after withdrawal, absent excess clinical activity (189). Perhaps endothelial cells are activated and immune cells infiltrate the Virchow-Robin spaces, but do not penetrate into the brain parenchyma. T1 black hole data would be of interest.

Switching to another drug prevents some rebound of multiple sclerosis activity. In the RESTORE trial, 175 patients who were stable on natalizumab were randomly divided into those who stayed on the drug (25%) and those who switched to placebo (25%) or other therapies (50%) (88). Patients assigned to other therapies chose among the immediate start of glatiramer acetate (19%) or intramuscular interferon beta-1a (19%), or a 3-month delay before starting monthly intravenous methylprednisolone (61%). Patients then switched back to natalizumab at week 28 or if they had an exacerbation. Those who remained on natalizumab were stable over the next 24 weeks (0% MRI lesions; 4% relapses). Those who switched had more activity: placebo (46%; 17%), methylprednisolone (40%; 15%), glatiramer acetate (53%; 27%), and intramuscular interferon beta-1a (7%; 29%).

In the year after switching, another study found exacerbations in 22% of patients after subcutaneous interferon beta and in 38% of patients after glatiramer acetate; 93% of recurrences occurred after 2 months. In a retrospective Swedish study, rituximab was more effective than fingolimod in preventing relapses in multiple sclerosis patients who switched from natalizumab due to positive JC virus antibody. Within 1.5 years of cessation of natalizumab, 1.8% (rituximab) and 17.6% (fingolimod) of patients experienced a clinical relapse (hazard ratio for rituximab = 0.10, 95% confidence interval [CI] = 0.02-0.43). Furthermore, contrast-enhancing lesions were found in 1.4% (rituximab) versus 24.2% (fingolimod) (04). Twelve percent of switches to ocrelizumab have early clinical or MRI activity. Lastly, dimethyl fumarate may be effective in preventing multiple sclerosis rebound after discontinuing natalizumab. In a retrospective study, 39 patients were switched to dimethyl fumarate. There was no immediate rebound effect, and at 2 years there were clinical relapses in 5 patients and MRI activity in 8 (42).

Thus, (1) natalizumab is highly effective; (2) relapses recur 3 to 6 months after discontinuation, but on average are no more severe than the highly active multiple sclerosis state that preceded natalizumab therapy; (3) MRI lesions and relapses recur after switching to placebo, monthly methylprednisolone, subcutaneous interferon beta, and glatiramer acetate (which has a slow onset of action), and to a lesser extent, intramuscular interferon beta and ocrelizumab; and (4) the early start of a second drug is safer than a late start. Interferon beta and glatiramer acetate may have a delayed therapeutic onset compared to fingolimod; these kinetics complicate prevention of disease rebounds, and residual symptoms after exacerbations may also differ between drugs. All told, the optimum time to start interferon beta, glatiramer acetate, or fingolimod is within 0 to 2 months. In 1 study that measured time-dependent rebound of activity, exacerbations doubled with a 16-week wait compared to an 8- to 12-week wait before starting fingolimod (131). If feasible, when natalizumab is tapered down rather than abruptly, the multiple sclerosis relapse rate is reduced (290). Restarting natalizumab has also stopped rebound disease activity.

Summary

Understanding the mechanism of the development of PML in multiple sclerosis will lead to better understanding of the JC virus life cycle, CNS immune surveillance by T cells on patrol, synergy between interferons and anti-VLA-4 agents, and the importance of VLA-4 in peripheral organs, the brain, and other immunologically privileged sites. A vaccine to BK and JC viruses before renal transplant or natalizumab therapy could prevent renal and brain complications.

Natalizumab appears to be more effective in the treatment of multiple sclerosis than other available agents. There are no more infections than expected (tuberculosis, fungal, bacterial, and viral), with the exception of PML. It is unknown if only the combination of natalizumab and interferon beta predisposes to PML. Most cases of PML have developed after prolonged therapy with natalizumab alone.

The dangers of natalizumab must be put in perspective. Discussion of other lifetime risks is important. There is excess mortality from multiple sclerosis (99) and from airplane crashes (1/20,000), fire (1/1000), and automobile accidents (1/100) (89). The author’s yearly risk of an accident while driving to work in Illinois is 1/10,000 per year (125). Anti-tumor necrosis factor therapy, anti-B cell therapy, and methotrexate therapy for rheumatic disease induce demyelinating disease, tuberculosis, lymphomas, and PML. In multiple sclerosis, mitoxantrone causes cardiac degeneration and lymphomas, and glucocorticoids have a plethora of side effects--without long-term benefit. As with many medicines, there is a balance between risk and benefit.

Reasonable suggestions using this therapy are as follows: (1) JC virus antibody-negative patients can be safely treated, with 6-monthly reassessment of their antibody status. (2) JC virus–positive patients must weigh the risks described above. Perhaps those with prior immunosuppressive therapy and with high anti–JC virus titers should not be treated, and the others should be reassessed based on their responses to 1 to 2 years of natalizumab therapy (261).