Central nervous system macrophages in progressive multiple sclerosis: relationship to neurodegeneration and therapeutics

MS results in motor, sensory and cognitive deficits and is one of the leading causes of neurological disability in young adults [1, 17]. The MS disease course is heterogenous, with each person with MS experiencing different symptoms of various severities. MS is diagnosed with the criteria of clinical attacks, dissemination of lesions in space and time and the presence of cerebrospinal fluid (CSF)-specific oligoclonal bands. The clinical course of MS falls into four main categories: clinically isolated syndrome (CIS), RRMS, SPMS and PPMS [18, 19]. The RRMS, SPMS and PPMS phenotypes can be further classified based on two disease modifiers that assess disease activity and progression [20].

MS often presents with a monophasic clinical episode known as CIS, as many as 60–80% of those with MRI detected lesions go on to be clinically diagnosed with MS [21]. Roughly 85% of people with MS (PwMS) present with RRMS, characterized by disease activity of clinical relapses and MRI activity of gadolinium (Gd)-enhancing lesions or new or enlarging T2 lesions, followed by total recovery or partial recovery with periods of stability in between relapses [20, 22, 23]. People with RRMS who have disease activity consisting of relapses or MRI activity are considered ‘RRMS-active’, while those without disease activity are considered ‘RRMS-not active’ [20]. In 50–80% of people diagnosed with RRMS, there is a transition into SPMS within 10–20 years from initial diagnosis [4‐6]. The average age of onset for RRMS is 30 while it is between 40 and 50 for progressive MS [24, 25]. SPMS is typically diagnosed retrospectively as there is no clear criteria to determine when RRMS transitions to SPMS [20, 26]. People with SPMS can be considered SPMS-active or SPMS-not active based on the presence of clinical attacks and MRI activity. In addition to disease activity, people with SPMS can be considered ‘SPMS-progressing’ or ‘SPMS-not progressing’ based on clinical evidence of disease progression or confirmed accumulation of disability independent of relapses [20]. While women are typically affected with RRMS and SPMS more than men, with an approximate ratio of 3:1, PPMS affects both sexes equally [27, 28]. About 15% of PwMS are diagnosed with PPMS, presenting with progressive disability from the onset with no clear relapses and minimal to no recovery [22, 23, 29]. Similar to SPMS, people with PPMS can be considered ‘PPMS-active’, ‘PPMS-not active’, ‘PPMS-progressing’ or ‘PPMS-not progressing’ based on assessment of disease activity and progression [20]. Taken all together, people with progressive MS can be classified as ‘active with progression’, ‘active without progression’, ‘not active but with progression’, and finally, ‘not active and without progression’ [20]. Due to recent irregularities on how these terms are used by regulatory authorities, it is recommended that while ‘progressing’ is used to describe accumulated disability independent of relapses, ‘worsening’ should be used to describe any resulting increase in disability or impairment from relapses or increase in disability in the progressive phase [19].

The classification of the MS disease course is dynamic and is constantly revised, particularly to address the confusion that arises with the approval of new DMDs and updated regulations [19]. There are certain issues yet to be addressed with the current classification of MS disease courses. For example, RRMS and SPMS are widely used in clinical practice and research, yet there is no clear distinction between them as the conversion to SPMS is a process that occurs over several years and is only realized retrospectively [23, 26]. Some studies have shown that there is roughly a 3-year period of diagnostic uncertainty during the conversional phase from RRMS to SPMS [30, 31]. In addition, some people with RRMS exhibit progressive features of the disease and some people with progressive forms exhibit relapses and new MRI activity. While relapses predominate in RRMS, clinically silent lesions occur during progressive MS, suggesting that lesion formation is not restricted to RRMS [32, 33]. Therefore, it is unclear if RRMS, SPMS and PPMS are distinct types of MS, or if MS is a disease with a spectrum [23, 24, 34].

The current subtypes of MS are useful for clinical practice and research, but emerging models to classify MS subtypes are being developed to predict disability and relapse rate. These models, which include a topographical model and an MRI abnormality model, could be particularly useful for grouping PwMS during clinical trials. The topographical model uses a real-time simulation software environment to define the dynamic changes of five different factors during the course of MS: lesion localization, relapse frequency, relapse severity, relapse recovery, and baseline brain volume and progression rate [34, 35]. In this model, the CNS contains finite functional reserve, which is lowered during the course of MS. As functional reserve declines, lesions present in clinically silent regions become uncovered. More work is required to identify the neurological substrate of functional reserve and how it is lost in progressive MS.

Eshaghi et al. proposed three new subtypes of MS based on MRI abnormalities including changes in grey matter, normal appearing white matter (NAWM), and lesion load. These subtypes are termed cortex-led, NAWM-led and lesion-led [36]. Compared to cortex or NAWM-led, those categorized as lesion-led were shown to have higher disease activity, greater risk of confirmed disability progression, and improved response to treatment in people with progressive MS. These considerations may be beneficial during clinical trials to stratify participants or to predict disease activity, disability progression, and treatment response. They are also foundations for developing validated models to be used as prognostic and therapeutic guides for individual patients.

MS is a disease largely dominated by inflammation and demyelination that causes tissue damage in the CNS. Although these processes are present in the early and progressive phases of MS, they can vary in severity [25]. The earlier phase of MS is dominated by a peripheral immune response characterized by the infiltration of lymphocytes into the parenchyma—the CNS tissue—through a disrupted blood–brain barrier (BBB) leading to the formation of new active lesions [24, 25]. In contrast, the progressive phase of MS is characterized by the slow expansion of pre-existing lesions, and by a heightened innate CNS immune response with trapped immune cells behind a closed BBB or CSF brain barrier [7, 37, 38]. The integrity of the BBB can be assessed using MRI to detect uptake of Gd-based contrast agents that are administered into the blood prior to imaging [39]. These Gd-enhancing lesions decline until they can no longer be detected as people enter the progressive stages of MS [40]. At this point inflammation is compartmentalized in the CNS and it is thought that expanding lesions add new cortical demyelination and damage to NAWM and normal appearing grey matter (NAGM) [24, 38], with new demyelination potentially occurring around CNS barriers such as the meninges. In addition to inflammation being more focal in the earlier phase of MS and diffuse in the progressive phases, inflammatory activity is more prominent in SPMS than PPMS as there is higher lesion cellularity and more perivascular cuffs in SPMS [23, 41].

In progressive MS, lymphocytic infiltrates consisting of T and B lymphocytes, plasma cells, and macrophages form lymphoid follicles in the CNS [32, 37, 39, 42]. Lymphoid follicles, which are associated with more severe microglia activation and cortical demyelination, are found in large aggregates in the meninges and the perivascular Virchow–Robin spaces [24, 38]. They are typically found in 40–70% of people with SPMS, but not in people with PPMS; however, increased meningeal inflammation associated with more extensive cortical demyelination and neurite loss is present in PPMS, but without lymphoid follicles [43]. Meningeal-associated cortical lesions, or subpial lesions, are prominent in progressive MS [32, 38]. The formation of subpial lesions, often within layers I–IV [41, 44], are thought to be linked to meningeal inflammation based on their close proximity [42, 44]. This pattern is also reproduced by in vivo MRI data and correlates with present and future disability [45]. PwMS with more cortical demyelination and lymphoid follicles have a more rapid MS onset, progression, and death [44, 46, 47].

While demyelination increases with disease duration, remyelination in PwMS is highly variable and declines with increase in age and disease duration [48]. For PwMS, a higher rate of remyelination is associated with lower disability progression [23, 49‐51]. Remyelinated lesions in MS typically have thin myelin sheaths and short internodes, making them paler than normal, which is why they are referred to as shadow plaques [23, 52, 53]. In RRMS, shadow plaques are elevated in people younger than 55 or within the first 10 years of diagnosis [48, 51] compared to progressive MS where they are relatively sparse [23, 52, 54, 55].

The normal adult brain contains four types of resident mononuclear phagocytes collectively called CNS macrophages. CNS macrophages include microglia that are located throughout the brain parenchyma and three types of CNS border-associated macrophages (BAMs) that are located at the interface between the CNS and BBB [70]. BAMs consist of perivascular macrophages in the perivascular space between the endothelial and parenchymal basement membranes, meningeal macrophages that line the meninges and its vasculature, and choroid plexus macrophages within the choroid plexus [70‐72]. It should be noted that some research groups refer to BAMs as CNS-associated macrophages (CAMs) [73]. Despite their anatomical differences, there are similarities in the development of microglia and BAMs [74]. Fate-mapping studies in mice have found microglia and BAMs largely originate from embryonic progenitors in the prenatal yolk sac that migrate to the brain for maturation [70, 75‐77]. Postnatal microglia and BAMs in mice are long-lived and self-renewing cells. Microglia, meningeal macrophages, and perivascular macrophages do not rely on circulating bone marrow-derived hematopoietic progenitors to replenish their population [70, 75, 78]. Only choroid plexus macrophages are partially replenished by monocytes [70, 77].

Among the CNS macrophages, the roles of microglia are the most well-characterized. In a surveilling state in the healthy CNS, microglia have small cell bodies with a complex highly branched ramified morphology. In a neuroinflammatory environment such as MS, microglia become reactive, a process that includes transcriptional, biochemical, and metabolic remodeling to take on new inflammatory functions [79‐81]. Reactive microglia swell and develop rounder cell bodies, taking on a simpler branching pattern with shorter and thicker cell processes [82, 83]. Microglia reactivity includes downregulation of many homeostatic genes, suggesting that there is an impairment and loss of critical homeostatic functions during MS that may worsen neurodegeneration during the pathophysiology of progressive MS [79, 84‐86]. There are several key genes that regulate microglial function. For example, Trem2 is a regulator of phagocytosis and chemotaxis, which are important defensive responses to inflammation and injury in MS [86, 87]. TREM2 is highly expressed by myelin-rich phagocytes in lesions of people with MS, and TREM2 agonists promote myelin debris clearance and enhance remyelination in an animal model of demyelination, suggesting TREM2 activation may be a promising therapeutic avenue in progressive MS [88]. Related to Trem2 is Apoe, a gene critical for lipid metabolism following microglial uptake of myelin lipid debris. Apoe expression is upregulated in microglia isolated from mice with EAE and is correlated with disease progression [79]. Activation of the TREM2–APOE pathway prevents the ability of microglia to regulate CNS homeostasis in EAE [79]. Expression of Cx3cr1, a homeostatic gene that encodes fractalkine receptor on microglia involved in synaptic pruning and modeling, is also lost in EAE [79]. Blocking the CX3CR1–fractalkine interaction induces microglial production of tumor necrosis factor (TNF)-α, a pro-inflammatory cytokine, and 8-isoprotane, a marker of oxidative stress [89]. Overall, while it is likely true that reactive microglia and macrophages with more pro-inflammatory phenotypes have large contributions to neurodegenerative processes during MS, the failure of microglia and macrophages to maintain reparative and homeostatic functions is also likely important to perpetuate neurodegeneration during progressive disease. Reactive microglia and macrophages express very similar molecular markers, and in many experimental conditions they are indistinguishable [90]. For this reason, the blanket term microglia and macrophage is used, which could also include BAMs.

Because microglia are highly dynamic, they transition through many intermediate morphological forms as they carry out reparative or pathogenic functions [82]. The response of microglia and macrophages to CNS disruption is diverse, but present during virtually all neurological conditions [91]. Upon sensing CNS disruptions such as tissue damage or pathogenic infiltrations, microglia become rapidly reactive. When the disruption is localized, they extend processes to sites of damage [61, 92]. The extension of microglia towards a focal injury can occur in a P2Y12 receptor-dependent manner, and it is in part regulated by extracellular adenosine triphosphate that is enriched after tissue injury [93, 94]. In this neuroprotective response, cell processes rapidly branch out to create a shield surrounding the injury site where they phagocytose pathogens and apoptotic or damaged cell debris [95, 96].

Microglia express over a thousand receptors, and likely respond to hundreds if not thousands of molecules that may drive their reactivity [97]. For example, microglia can also become reactive after toll-like receptor (TLR) or nucleotide-binding oligomerization domain-like receptor-mediated recognition of damage-associated molecular patterns that are released following cellular injury and death in the CNS [98, 99]. This reactivity helps induce the inflammatory cascades that may perpetuate progressive neurodegenerative mechanisms in MS [100]. Under certain circumstances, microglia and macrophages may also contribute to progressive neurodegeneration in MS by producing glutamate, proteases, and reactive oxygen and reactive nitrogen species (ROS/RNS) that may lead to demyelination, neuronal loss, and axonal and mitochondrial damage in lesions (discussed below) [101‐105]. Reactive microglia and macrophages also produce inflammatory cytokines such as TNF-α, interleukin (IL)-6, IL-1β, and IL-23 that leads to further immune activation [104, 106‐109]. People with progressive MS have elevated levels of these cytokines within blood serum, CSF, and CNS lesions compared to people with RRMS or healthy controls [110‐112].

Microglia and macrophages are also necessary for remyelination and may have several other neuroprotective roles [12, 14, 68]. Microglia and macrophages help recruit and maintain oligodendrocyte progenitor cells (OPCs) and their ability to myelinate neurons [66, 67, 113]. They secrete neurotrophic factors including insulin-like growth factor-1 that are required for OPC survival and differentiation into oligodendrocytes [114‐117]. In mice with EAE, microglia can also phagocytose and kill CNS-infiltrating Th17 cells [118].

The dichotomous functions of microglia and macrophages have complicated the literature for decades. Protective and toxic functions were initially ascribed to different polarization states of microglia and macrophages, the pro-inflammatory M1 and immunoregulatory M2 states [119]. Indeed, using transgenic mice with Nos2 based pro-inflammatory states and Arg1 immunoregulatory states, Locatelli et al. examined these states during EAE. They found that cells could preferentially express either Nos2 or Arg1, with the most Nos2-enriched cells found prior to and at onset of disease inflammation [120]. Given that studies found microglia and macrophages can be polarized in culture with specific cytokines, this suggested these states may be important to disease [12, 119‐121]. However, both Arg1 and Nos2 are rarely expressed by microglia in MS models and do little to predict the microglia phenotype [80, 90, 120]. Monocyte-derived macrophages can express these markers differentially, but also take on several other cellular phenotypes suggesting greater diversity [122]. How the function of microglia and macrophages relate to their cellular phenotypes during MS remains an open question.

The identity of BAM-specific markers can be used to distinguish BAMs from microglia. Like microglia, BAMs are highly plastic and change their expression patterns during disease, so they may be confused with reactive microglia [123, 124]. Only recently are new markers becoming available, so the exact roles of BAMs are not as well characterized [71, 72, 125]. Studies demonstrated that BAMs are involved in recruiting immune cells to sites of inflammation and scavenging debris [126, 127]. BAMs are located at the CNS border and help regulate immune cell entry and neuroinflammatory responses within the leptomeningeal space [128]. BAMs during neuroinflammation upregulate expression of molecules such as MHC class II and CD40 that are involved in antigen presentation or co-stimulation, suggesting that they may regulate the activity of lymphocytes in the leptomeningeal space, perivascular space, or choroid plexus [84, 129, 130]. However, experiments that eliminated the MHC class II antigen presenting capabilities of microglia and BAMs found that these cells are not required for T cell-mediated EAE pathogenesis [124, 131‐133]. Instead, it may be the antigen presenting capabilities of CNS-infiltrating myeloid cells, namely dendritic cells, that are involved in reactivating myelin-reactive encephalitogenic T cells towards a pro-inflammatory phenotype, thus inducing pathology in EAE. Overall, because most neuroimmunology studies have not made a distinction between BAMs, microglia, and infiltrating peripheral monocyte-derived macrophages, the exact contribution of BAMs to immune homeostasis of the normal CNS or MS pathophysiology is unclear.

Microglia and macrophages are the major cell types in MS lesions [58, 134]. While they predominantly have a pro-inflammatory phenotype in active lesions indicated by expression of inflammatory markers such as CD40, approximately 70% also express more homeostatic markers [135]. In both human MS and mouse models, microglia and macrophages in early CNS lesions produce increased amounts of inflammatory factors such as nitric oxide (NO), TNF-α, and IL-1β. As they continue to phagocytose, process, and clear cholesterol-rich myelin debris, they acquire a lipid-laden foamy phenotype that generally have more reparative anti-inflammatory characteristics [120, 121, 136]. Myelin breakdown in phagocytes is thought to generate lipid and cholesterol metabolites that bind and activate liver X receptors (LXR), which modulates inflammation and helps facilitate the reverse cholesterol transport system in microglia and macrophages [137‐139]. Effective LXR-mediated cholesterol efflux also increases production of immunomodulatory factors such as IL-10 [136]. The ability of microglia and macrophages to efflux cholesterol-containing myelin debris decreases with age in mice [140]. The resulting cholesterol build-up forms crystals, leading to rupture of the lysosomes and activation of the inflammasome. Consequently, heightened microglia and macrophage inflammasome activity impairs remyelination in the CNS. Given that heighted inflammasome activity promotes neurodegeneration in other neurodegenerative conditions, the impaired efflux of cholesterol and subsequent inflammasome activation in microglia and macrophages may contribute to disability progression in PwMS [136, 140‐142].

Even though lesions are the major pathological hallmark of MS, PwMS also have alterations in the NAWM and NAGM compared to controls. The NAWM and NAGM is not associated with a greater density of microglia and macrophages in people with progressive MS compared to normal age-matched controls [37, 84]. However, the brains of people with progressive MS have significantly increased microglia reactivity in the NAWM and NAGM based on reduced expression of homeostatic microglia markers including P2RY12 compared to normal controls; microglia reactivity increases with disease duration. Increased microglia reactivity in MS is associated with more diffuse NAWM injury including myelin loss and axonal damage [41, 43, 58, 84]. Clusters of reactive microglia termed microglia nodules may be present in over half of PwMS, and they are found in areas around plaques and in NAWM in progressive MS, but not normal controls [41, 151, 153]. These nodules have been associated with Wallerian degeneration of axons in early MS lesions, and may even occur prior to demyelination [153, 154].

Recent advancements in positron emission tomography (PET) imaging allow the visualization of radioligand binding to 18 kDa translocator protein (TSPO), which is a protein that is thought to be upregulated in reactive microglia and macrophages [155, 156]. PET studies have found increased binding of the first generation radioligand ¹¹C-PK11195 to TSPO in the NAWM, thalamus, and cortical grey matter of PwMS, particularly SPMS, compared to healthy controls [156‐159]. The increase in total ¹¹C-PK11195 binding in the cortex correlated with increased expanded disability status scale (EDSS) disability scores and progression in SPMS but not RRMS [156, 157]. Second generation radioligands such as ¹¹C-PBR28 with higher binding specificity and affinity for TSPO showed similar increases in signal, particularly in SPMS, that was associated with increased neurological disability and progression [160, 161]. People with SPMS also have greater overall TSPO expression throughout the grey matter compared to those with RRMS [161]. One limitation of TSPO-PET imaging is that TSPO may not be specific only to reactive microglia as it is also expressed in some reactive astrocytes and vascular endothelial cells [155, 162]. TSPO expression also does not differentiate between human microglia stimulated in vitro under pro-inflammatory or anti-inflammatory conditions [163]. As a result, there is a need for more microglia-specific TSPO-PET radioligands that can distinguish between inflammatory and homeostatic microglia phenotypes [39].

Based on TSPO-PET findings, microglia and macrophage reactivity occurs early in the disease course of MS and escalates during progressive MS. TSPO-PET based microglia and macrophage reactivity predicts disease progression suggesting that these cells have important roles in neurodegenerative processes, particularly during progressive disease.

Characteristic features of progressive MS pathophysiology, including iron deposition, involvement of fibrinogen, and meningeal inflammation are likely important contributors to the reactivity of microglia and macrophages, potentially contributing to neurodegeneration (Fig. 2).

In progressive MS there is increased microglia and macrophages reactivity into a pro-inflammatory phenotype, which may contribute to neurodegeneration via a number of potential mechanisms. Pro-inflammatory microglia activation results in a loss of immunosuppressive factors including CX3CR1 and CD200 and increased secretion of the pro-inflammatory cytokines IL-1β, IL-6, and TNF-α among other neurotoxic factors [230]. Given that microglia and macrophages express and release many cytokines in MS, individual or combinations of cytokines may induce neuronal toxicity.

While there are over 15 DMDs approved for the treatment of RRMS over the last two decades, there are only limited therapeutic options currently available to treat progressive MS [4]. A range of beneficial and detrimental effects of microglia and macrophages have been reported in the models of MS [11, 15, 16], but drugs that solely target these innate immune cells are lacking. The recent treatment successes for progressive MS include sphingosine-1-phosphate (S1P) receptor modulator (siponimod), B-cell targeted therapy (ocrelizumab), and selective immune reconstitution therapy (cladribine) [256], while mitoxantrone and beta-interferon were the first few drugs used for treatment of SPMS. In this section, we described several approved DMDs for the treatment of progressive MS (Fig. 4).

There are several therapies being evaluated as a potential treatment option for progressive MS, such as immunomodulatory therapies that can target myeloid cells (e.g., dimethyl fumarate, ibudilast, lipoic acid) or as a potential neuroprotective agent (e.g., simvastatin). Other examples of drugs currently being assessed in phase 2/3 trials as potential treatments for progressive MS are listed in Fig. 4.

Several other drugs have also been investigated for the treatment of progressive MS via neuroprotective and experimental approaches (e.g., amiloride, fluoxetine, riluzole) or remyelination promoting approaches (e.g., biotin, opicinumab), but showed negative results (Table 1). For example, although the effects of high-dose biotin (MD1003) for the treatment of progressive MS shown in previous studies (pilot study and a phase 2 clinical trial) were encouraging [321, 322], the most recent phase 3 clinical trial failed to demonstrate any significant differences in disease progression between biotin-treated and placebo-treated arms [323]. High-dose biotin is a co-factor for essential carboxylases, which may help support myelin repair by enhancing the production of energy in neuron and it is generally well tolerated [323, 324]. Some studies suggested that the exposure to biotin was associated with new disease activity [325, 326]. Similarly, in an adequately powered multi-centre, multi-arm, phase 2b clinical trial (MS-SMART) based in the United Kingdom published recently, no significant differences in annual percent change brain volume were found among people with SPMS treated with placebo compared with each of the three active treatments (amiloride, fluoxetine and riluzole) [327], despite the positive effects shown in earlier work (i.e., in animals and pilot studies) [328‐333]. Other recent examples of drugs that had been studied but showed negative results are listed in Table 1.