Environmental and genetic risk factors for MS: an integrated review

Abstract Recent findings have provided a molecular basis for the combined contributions of multifaceted risk factors for the onset of multiple sclerosis (MS). MS appears to start as a chronic dysregulation of immune homeostasis resulting from complex interactions between genetic predispositions, infectious exposures, and factors that lead to pro‐inflammatory states, including smoking, obesity, and low sun exposure. This is supported by the discovery of gene–environment (GxE) interactions and epigenetic alterations triggered by environmental exposures in individuals with particular genetic make‐ups. It is notable that several of these pro‐inflammatory factors have not emerged as strong prognostic indicators. Biological processes at play during the relapsing phase of the disease may result from initial inflammatory‐mediated injury, while risk factors for the later phase of MS, which is weighted toward neurodegeneration, are not yet well defined. This integrated review of current evidence guides recommendations for clinical practice and highlights research gaps.


Introduction
Multiple sclerosis (MS) is a chronic inflammatory disease of the central nervous system (CNS). Whether this inflammation is in response to a chronic viral infection, primary neurodegenerative processes, or a reflection of a dysfunctional immune system was once hotly debated. 1 In the past 5 years, consolidation of the evidence for environmental risk factors, 2 identification of new genetic factors, 3 and increased appreciation of the importance of gene-environment (GxE) interactions 4 have led to a greater consensus. This new knowledge can provide clues to the causal biological pathways, and disease prevention for both populations and at-risk individuals. The investigation of factors associated with disease progression and severity is similarly essential to guide the exploration of pathological pathways and, potentially, lifestyle recommendations to slow progression.
In this review, we discuss recent findings regarding genetic, epigenetic, and environmental risk factors for MS risk and prognosis, and how they impact our understanding of molecular processes at play. We also discuss new methods of analysis, proof-of-concept trials, and recommendations for at-risk individuals and patients with the disease, as well as areas of future research needs.

Literature Search Methods
We searched PubMed for papers published between 1 January 2013 and 1 April 2019, without any language restrictions, with the following terms: "environmental factors", specific risk factors such as "EBV", "vitamin D", "sun exposure", "genetic factors", "gene-environment interactions", "epigenetics", "multiple sclerosis", "susceptibility", "disease modification", "modelling". Additional references were gathered from retrieved publications, books, and conference reports. Original research articles and, when appropriate, high impact reviews were included. The final reference

Risk factors for multiple sclerosis onset
Genes That some people are genetically susceptible to developing MS and that the main susceptibility allele is a human leukocyte antigen (HLA) Class II gene have been known for over 50 years. Increasingly sophisticated genetic platforms have expanded our understanding by identifying over 200 non-HLA single-nucleotide polymorphisms (SNPs) and at least one protective HLA allele. 3 Many of the non-HLA SNPs are located near genes involved in adaptive or innate immunity underscoring that MS is a disorder of immune homeostasis. All of the SNPs so far identified are normal gene variants (i.e. not disease genes). Around 20% of the heritability risk is attributable to common genetic variants 3 but low-frequency and rare coding variation also contribute (about 5%). 3 While several HLA Class II and I alleles contribute to MS susceptibility, 8,9 the main one is carriage of the HLA DRB15:01 haplotype (odds ratio (OR) of~3). This gene variant is carried by 25-30% of the population in northern Europe and the USA. The reasons for this association are still unclear. Hypothetically, the DRB15:01-binding pocket may allow optimal binding and presentation of CNS-related autoantigens, which drives T cells to attack the CNS. Autoproliferation of Th1 lymphocytes appears to be elevated in HLA-DR15-positive individuals and mediated by memory B cells in a HLA-DR-dependent manner. 10 Interestingly, the DRB15:01 gene is epigenetically regulated, leading to increased expression and antigen presentation. 8 The second strongest MS gene variant is HLA A02, which is protective (OR~0. 6). 11 The reasons for Class I-associated protective effects are also unclear. The alleles of HLA A02 may be important in the elimination of viruses connected to MS, such as EBV (see later). In addition, binding of a peptide from a protein commonly expressed by many cell types (glycerolphosphatidylcholine phosphodiesterase 1) to HLA-A*02-01 may cross-stimulate a myelin-reactive T-cell receptor, possibly leading to increased CD8 + activity. 12 For the many non-HLA genetic loci associated with MS risk, 3,13 the magnitude of the association is much smaller (e.g. ORs up to 1.3) than for HLA. Furthermore, most of the SNPs are located outside exomes, suggesting that they, or nearby polymorphisms, are affecting the regulation of gene expression rather than the function of a gene product. Only a small number of non-HLA influences in MS have so far been functionally deciphered: a genetically determined increase in the levels of soluble tumor necrosis factor (TNF) receptors blocks a protective action of TNF; 14 a genetic polymorphism resulting in higher IL22 binding may neutralize a protective effect of IL22; 15 and tyrosine kinase 2 exomic variation decreases pro-inflammatory signaling. 16 Many genetic variants are shared across several autoimmune diseases suggesting some common pathogenesis. 17 Although most of the genetic work in MS has been conducted in populations with European ancestry, recent studies in African Americans have shown a significant overlap with MS variants reported in Whites. 18 Environmental factors (see also Table 1)

Life-stage of risk of exposure
Based on migration studies, 19 key environmental exposures associated with adult MS onset occur before age 15. New evidence suggests that risk exposures may occur in utero and in neonates. [20][21][22][23] For example, the risk of pediatric-onset MS is increased in association with maternal illness other than diabetes and preeclampsia during pregnancy, and a father with an agricultural profession during pregnancy. 20 Post-winter birth is more common, and post-summer birth less common, in people with MS, 24 although there is some debate whether this finding may be due to residual confounding. 25,26 Lower vitamin D status during pregnancy or infancy is associated with increased risk of MS in Whites. 22,23,27 Studies of cesarean section delivery and having been breastfed, 20,28-31 both of which have profound effects on the microbiome, have yielded mixed results. Both deserve further study as these are modifiable factors. It is conceivable that some early life exposures may not contribute alone to MS onset, but are markers of risk that also act throughout childhood and teenage years.

Past viral infection and/or reactivation
The mechanisms through which past viral infections, particularly with herpes viruses, and viral reactivation may contribute to MS onset remain hypothetical. 32 Epstein-Barr virus: there is strong, consistent, evidence across multiple racial/ethnic groups that past EBV infection, including infectious mononucleosis, positivity for EBV nuclear antigen (EBNA)-1 IgG, or higher EBNA-1 titers, is associated with an increased risk of MS. 33 The high prevalence of EBV seropositivity in the general population suggests that any pathogenic effect of this agent may depend on other risk factors (e.g. GxE interactions). Recent work provides further evidence that EBV is causally involved in MS, possibly acting via EBNA-2-induced alterations in gene transcription in EBV-infected astrocytes, microglia, and B cells (see Box ). An interaction between EBNA positivity and HLA-A02 has been reported. 34,35 In a recent study of MS cases and non-MS controls who were seropositive for EBV, EBV viral load was lowest in HLA-A*A02-positive individuals, and highest in HLA-B*07-positive individuals. These findings support a causal role for EBV in MS, modulated by HLA-Class 1 genotype through alterations in antigen presentation to T cells 36 . An additive interaction of EBV status with DRB1 has also been reported to modulate MS risk in Whites 37 and Blacks. 37 Human Herpes virus 6 (HHV6): past infection with neurotropic HHV-6 has been inconsistently linked with increased MS risk. A recent meta-analysis of seven serological studies and 34 molecular studies that met high quality inclusion criteria showed an increased MS risk in association with evidence of past HHV-6 infection. 43

Cytomegalovirus (CMV):
A few studies have suggested a protective association of prior CMV infection with adult and pediatric MS while several have reported no association. [44][45][46][47] Herpes simplex virus (HSV): Children with prior exposure to HSV-1 have a modestly increased risk of having pediatric MS. This is mostly seen in Whites and DRB1negative individuals. 44 MS associated human endogenous retroviruses (HERVs): Expression of HERVs is curtailed in healthy individuals by epigenetic factors, but HERVs are transcribed at high levels in autoimmune diseases such as rheumatoid arthritis, Sjogren's disease, systemic lupus erythematosus, and MS. 48 HERVs are expressed in the CNS; some seem particularly associated with MS susceptibility, others with disease course. 48 Sun exposure and vitamin D That MS prevalence increases with greater distance from the equator led to the hypothesis that lower exposure to ultraviolet (UV) radiation and subsequently, lower vitamin D status (blood concentration of 25-hydroxyvitamin D (25OHD)), increases the risk of MS. Latitude gradients have reduced over time in some populations, 49,50 but not all; 49,51 this may be due to changes in exposure to latitude-related factors, for example, sun exposure, diet, in some populations, but not others. Higher latitude 52 and lower summer sun exposure in adolescence 53 have been linked to a younger age of onset.
Observational studies similarly show that lower 25OHD levels are associated with increased MS risk in White populations. 22,59,[65][66][67] A similar association is not seen in Hispanics and Blacks. 68 Within White populations, the doseresponse relationship is inconsistent; for example, both threshold (e.g. protective only at 25OHD levels >99 nmol/ L 65 or increased risk only for the lowest quintile of 25OHD 69 ) and linear effects are described. 66 These inconsistent associations may be because few studies have accounted for sun exposure. Where data are available on both sun exposure and 25OHD levels, there are statistically independent benefits of higher levels of either in Whites. 59,68 Both UV radiation and vitamin D have effects on innate and adaptive immune function that would plausibly be beneficial for MS. 70 It can be difficult in observational studies to distinguish between the effects of low sun exposure and low vitamin D. Sun exposure may be a marker of lifetime vitamin D status; equally, 25OHD level may simply be an objective marker of sun exposure. Vitamin D supplementation leads to a rise in 25OHD levels without a co-intervention of sun exposure; thus, supplementation studies can evaluate the specific impact of vitamin D on relevant outcomes. As an illustration, observational data from the Finnish Maternity Cohort showed a 50nmol/L increase in 25OHD was associated with a 39% lower MS risk, with blood samples taken on averagẽ 9 years prior to MS onset. 67 Routine vitamin D supplementation for pregnant women was introduced in 2004, providing a natural trial of vitamin D supplementation and the impact on MS incidence. Incidence rates for MS in southwest Finland were 5.1, 5.2, and 11.6 per 100 000 for the hospital districts of Uusimaa, Vaasa, and Seinajoki, respectively, from 1979 to 1993, 71 and 12.1 per 100 000 for southwest Finland in 2012-2016. 72 This is not the dramatic reduction in MS incidence that might have been expected

Box. EBV and MS -what was known in 2013, and what is new in 2019
What was known?
• EBV antibodies are present in almost 100% of people with adult MS. 38 • Strong and consistent association between history or serological evidence (mainly anti-EBNA IgG) of past EBV infection and increased risk of developing MS. 39 What is the view in 2019?
• Meta-analysis of meta-analyses confirms consistent association of past history of infectious mononucleosis (OR = 2.17, 1.97-2.39) and anti-EBNA IgG positivity (4.46, 3.26-6.09) and increased risk of MS. 33 • In archived brain tissue, 93% of MS and 78% of control brains contained EBV latent membrane protein; 78% of chronic MS lesions contained EBV-related proteins compared to 33% of non-MS brains. About 85% of MS brains had frequent EBERpositive cells; these were uncommon in non-MS brains. 40 Astrocytes and microglia as well as B cells were infected. 41 • EBV encodes microRNAs (miRNAs) that are abundant in latently infected cells. EBV miRNAs inhibit the expression of viral antigens, allowing infected cells to evade the host's immune system. EBV miRNAs interfere with antigen presentation and immune cell activation thus suppressing the host antiviral immunity.
• Using novel computational methods, one study showed that EBNA2 protein occupies multiple loci as a transcription factor in MS, as well as a range of other autoimmune diseases including systemic lupus erythematosus, type 1 diabetes, inflammatory bowel diseases, rheumatoid arthritis, juvenile idiopathic arthritis, and celiac disease, suggesting genetic effects that are dependent on EBNA2. 42 These findings are supported by recent experimental studies where by MS risk SNPs were overrepresented in the target loci of the EBV transcription factor EBNA2, in genes dysregulated between B and LCLs, and as targets for EBV miRNAs.
with the advent of vitamin D supplementation. However, supplementation may have been relatively low (Nordic recommendations are for 400 IU/day).

Diet and supplements
In vitro and animal studies suggest that lower intracellular sodium concentrations and sodium intake may have protective effects on immune function. 73 However, recent case-control studies show no association between dietary sodium intake and risk of MS. 74,75 Polyunsaturated fatty acids (PUFAs) are strong immunomodulators in vitro and in animal models. Consistent with this, most, but not all, 76 studies that measured seafood-based sources or direct micronutrient PUFA intake found a reduced risk of MS with higher fatty fish intake, 77,78 or cod liver oil supplementation. 79 These findings should be interpreted cautiously, with inconsistencies across studies in the timing of the risk exposure, that is, during childhood, adolescence, or through adulthood. In addition, most studies could not fully account for a substitution effect (i.e. it may not be fish that is protective but what is not being eaten instead of fish that is harmful).

Gut microbiota
Dysbiosis in the gut microbiota has emerged as a potential risk factor for MS, following the report of its key role in shaping the immune response. Many factors modulate the gut microbiota including diet, obesity, antibiotic use, and cigarette smoking. Studies of modest size have reported that gut microbiota diversity is overall similar in prevalent MS cases and controls, including between monozygotic twin pairs discordant for MS. [80][81][82][83][84][85][86] Discrete taxonomic enrichments and depletions have been observed in MS cases compared to controls, some of which vary across studies. The discrepancies across studies could be due to differences in inclusion criteria and disease duration, as well as in the analytic platforms used. In general, the reported discrete taxonomic differences point toward a more pro-inflammatory milieu in cases. Whether this is a potential cause, or consequence of MS, treatments, or dietary changes following diagnosis, has not been determined.

Reproductive factors
Sex dimorphism is predominant between puberty and menopause. Before and after these reproductive milestones the sex ratio is 1:1; between them the ratio is 3:1. Earlier age at menarche is associated with an increased risk of MS, 29,87 and the onset of pediatric MS peaks 2 years after menarche. 88 Whether this is mediated by complex interactions between immunological responses to infections (e.g. EBV), direct effects of sex or other hormones, or other factors is unknown. The effect of menopause on disease course and MS susceptibility is unclear and worthy of further study. One cross-sectional referral center study suggested a decrease in female:male ratio in patients with symptom onset after age 50. 89 Whether this is related to a decline in risk of relapsing-onset MS in postmenopausal women or an increase in risk of progressive-onset MS in older men is unclear. Of note, recently updated MS prevalence estimates in the USA show only a slight decrease in female:male ratio over age 55. 90 Two recent studies suggest that women who breastfeed their infants exclusively 87 and for a prolonged period 29 may have a lower subsequent risk of MS. How breastfeeding could protect the mother against an autoimmune disease is unknown. The potentially protective effect of breastfeeding may explain the inconsistent association between parity and MS risk, either protective 87,91 or no association, 29,92 as most studies did not account for subsequent breastfeeding.
Other behavioral reproductive factors including use of hormonal contraceptives (when smoking is adequately accounted for), age at first birth, and total ovulatory and menstrual years do not appear to be associated with MS risk. 91

Air pollution
Whether air pollution increases the risk of MS is unclear. Exposure to particulate matter (PM) triggers an inflammatory response in the lung, with the release of inflammatory cytokines and elevated systemic levels. Long-term exposure to air pollution is associated with neuroinflammation and damage to the blood-brain barrier. Studies examining exposure to PM of aerodynamic diameter <10 µm (PM 10 ) show an increased risk of MS or no association. [93][94][95] A strong association between air pollution (but not water or land pollution) and having pediatric MS 96 has been reported and attributed to exposure to sulphur dioxide, PM 2.5 , carbon monoxide, and lead. 97

Smoking and oral tobacco
Cigarette smoking is a well-established risk factor for MS onset, with a clear dose-response relationship. 98,99 Age at starting smoking does not seem to affect MS risk. 98 However, quitting smoking is associated with a gradual decline in the excess risk of MS, to zero at 10 years post-cessation, regardless of the cumulative dose. 98 Passive exposure to smoking, including water pipe smoking, has also been associated with increased MS risk. 100,101 In contrast, snuff, a form of smokeless tobacco, has been associated with a dosedependent lower MS risk, particularly among smokers. This suggests that exposure to tobacco smoke is due to nonspecific lung irritation and inflammation. Nicotine itself may have neuroprotective effects as reported in experimental models. 102 The harmful effects of tobacco smoke may be conveyed through peribronchial lymphatic tissue in licensing autoreactive T cells or activation of the aryl hydrocarbon receptor, as is suggested in rheumatoid arthritis. 103 Childhood/adolescent obesity Several high quality observational studies have reported that obesity in adolescence and early adulthood is associated with approximately double the risk of developing pediatric and adult MS compared with normal-weight individuals. [104][105][106] The association has been largely confirmed in females, including a dose-effect, 105 while evidence is mixed regarding the association in males. [106][107][108][109]

Mendelian randomization (MR) studies
Three MR studies have shown that genetically lower 25OHD levels are significantly associated with increased MS risk albeit only in White, non-Hispanics. 104,110,111 Critical MR assumptions were carefully considered. 112,113 As confounding by ancestry could occur, 114 the vitamin D MR models were rigorously adjusted for population stratification and sensitivity analyses were performed to assess ancestry's potential impact, since several of the vitamin Drelated variants have different frequencies in different ethnic groups, and variation in risk of MS according to ethnicity is well-described. Genetic risk scores of variants associated with obesity in large (gene wide association study) GWAS studies have confirmed a strong association with pediatric and adult MS, suggesting causality. 104,115 GxE interactions and MS onset Environment interactions with genotype in MS have focused on HLA DRB1*15:01. Table 2 summarizes current knowledge of interactions between environmental factors and carrying HLA DRB1*15:01 and lack of HLA A*02. Thus far, interactions between HLA DRB1 and EBV and HSV infections, smoking and adolescent obesity, but not oral nicotine or CMV, have been demonstrated in Whites. The interaction between EBV serology and HLA DRB1 was recently confirmed in Blacks. 45 Occupational exposure to inhaled organic solvents may also interact with both HLA DRB1*15:01 and lack of HLA A*02. 116 If these findings are confirmed, it would indicate that the general lung irritation could contribute to MS risk.
While our genotype cannot be altered, smoking is a nice example of how MS risk can be compounded or reduced even in those with a genetic susceptibility for MS.
Nonsmokers who carry HLA DRB1*15:01 and lack HLA A*02 have a combined OR of~5; however, among smokers, the combined OR is~14, much higher than the sum of the main effects associated with each factor. 117 Smoking may also interact with the NAT1 gene (encoding N-acetyltransferase 1 enzyme that is important in the metabolism of aromatic amines present in cigarette smoke) to alter MS risk. 118 These findings support the notion that MS onset requires the interaction between multiple pro-inflammatory states and help explain why MS is rare, but carrying HLA DRB1*15:01 and pro-inflammatory exposures, like EBV infection, smoking, and obesity, are common.

Epigenetics and MS onset
Epigenetics refers to heritable changes outside of the DNA sequence that influence which genes are turned on and off and when. Epigenetic changes reflect aging and our body's interaction with the environment. In some instances, these changes can contribute to specific diseases. Little is known about how such changes influence MS susceptibility or prognosis. DNA methylation is probably the best explored of the three main epigenetic processes, and aberrant methylation in gene regulator regions may underlie processes involved in MS onset. 119 While methylation changes have been reported in MS at HLA-DRB1 in CD4 + but not CD8 + T cells, 120,121 in brains, 122 and in relapsing-remitting versus primary progressive MS, 123 it is unclear if these changes are drivers or responses to the disease.
Expression of several microRNAs (miRNAs) appears to be increased in various tissues from patients with MS and may alter the immune response and ultimately contribute to disease onset. 124,125 Increasing our understanding of epigenetic changes, particularly reversible ones, may elucidate pathogenic pathways and guide the development of new therapeutic strategies.

Risk factors for disease progression
Identifying clear-cut MS prognostic factors is challenging because disability accumulates slowly over decades and there is a lack of good surrogates of disability progression. Disease activity (rate of relapses and new MRI lesions) is thought to reflect inflammatory processes, whereas disease progression (worsening of disability and brain and cord atrophy) may better reflect neuronal injury. 6 An additional challenge is that large population-based MS cohorts that represent the broad range of MS prognosis are rarely associated with genetic and environmental exposure data. Most studies discussed herein are highly selected, referral center populations and utilize anatomical correlates of inflammation (T2 and contrast-enhancing lesions) and neurodegeneration (cerebral atrophy) or relapse rates. Treatment decisions are often based on relapse rate or the development of new MRI lesions. However, while relapses in the first 5 years of disease onset are associated with short-term disability progression, they contribute to a lesser extent to long-term disability progression. 126,127 While spinal cord, and to a lesser extent global brain, atrophy measures are correlated with disability progression, they are not yet broadly implemented for treatment decisions. 126 Finally, cord atrophy, although a very promising predictor of disability progression, has not yet been used as a prognostic marker in studies of the effect of genetic or environmental factors, as its development is more recent than brain atrophy.

Genes
There have not been consistent associations of genetic susceptibility factors with cross-sectional disability metrics or standard MRI outcomes. In a study of over 7000 individuals with MS, a genetic burden score was not associated with disability. 128 Other smaller studies have reported associations of several genetic variants and progression of disability but have not yet been replicated. [129][130][131][132] . Notably, no study has been able to identify consistently a significant difference in genetic burden or variants between relapsing and progressive-onset MS. 11, 133 Though overall susceptibility genes do not appear to have large effects on phenotype, there is some evidence that HLA and non-HLA genetic risk factors may alter relapse rates, although overall findings are mixed and weak. Some susceptibility polymorphisms have been associated with attack or MRI lesion location and severity. [134][135][136][137] The strongest genetic factor (HLA-DRB1*15:01) has been associated with earlier age of onset and possibly with greater deep gray matter atrophy. 138 The MS susceptibility polymorphism within the Abelson Helper Integration site 1 gene (AHI1) has been associated with higher relapse rate in two separate cohorts (one of children and another of adults), with similar effect size. 139 In a genome wide search for alleles associated with relapse rate, a SNP within the gene LRP2 was associated with a twofold increase in the hazard to relapse in three independent MS cohorts, including adults and children. 140 LRP2 is located on the cell surface of neurons and oligodendrocytes and participates in axon guidance. In early MS, a genetic variant within myelin basic protein (MBP) was reported to be associated with both relapse hazard and disability progression. 129 Genetic factors within the vitamin D pathway have been associated with both 25OHD levels and relapse rate. 141,142 Environment Sun exposure and vitamin D Few studies have tested the association between sun exposure and relapse rate and/or progression, while it has been hard to rule out reverse causality in the association of lower 25OHD with more active disease course.
In a population-based prospective cohort study, higher sun exposure prior to MS onset and/or increasing sun exposure after the first MS relapse (onset attack) were associated with a longer first-to-second attack interval and lower relapse rate. 143 Ecological studies show higher relapse rates during seasonal low ambient UV radiation, 144 and a stronger seasonal pattern (peak in early spring) at higher latitudes. 145 A recent comprehensive review of vitamin D and MS concludes that a low serum 25OHD level is associated with increased disease activity in MS, but the findings are typically stronger for MRI than clinical outcomes. 66 Most (but not all) observational studies show lower relapse rates and (Expanded Disability Status Scale) EDSS in association with higher 25OHD level, although this may be disease-induced rather than causal. 66 Randomized controlled trials (RCTs) of vitamin D supplementation have not shown the hoped-for benefits. A recent meta-analysis of RCTs of vitamin D supplementation in people with MS suggests no therapeutic effect on either EDSS score or annualized relapse rate. 146 Indeed, a previous meta-analysis suggested that higher doses of vitamin D supplementation may be associated with worse outcomes. 147 It is important to note that many of the RCTs of vitamin D supplementation for people with MS have had limitations of small sample size (low power), short duration, and including people who are not vitamin D deficient. Studies in other diseases have shown that benefits of vitamin D supplementation are seen only in those who are severely deficient at baseline. 148 Additional studies of vitamin D supplementation in relapsing-remitting MS and CIS are currently underway.
The possible benefits of ultra violet B phototherapy have recently been tested in a RCT of people with clinically isolated syndrome. Following an 8-week course of phototherapy, there was a 30% reduction in risk of converting to MS at the 12month follow-up. This finding is promising, although it was not statistically significant, likely due to a small sample. 149

Diet
Despite great interest, there are few studies of dietary effects on MS prognosis, all with significant limitations. The potential pro-inflammatory effects of a higher salt intake observed in animal models have yielded mixed results in humans. 73 Higher sodium levels in random urine samples were associated with higher rate of new MRI lesions and relapses, 150 yet 24-hour urine sodium levels, and salt intake from food frequency questionnaires showed no association. 151,152 In pediatric MS, a higher saturated fat intake and lower vegetable intake were independently associated with higher relapse risk. 153 Small diet intervention trials have not shown any effect of low-fat diet on relapse rate or MRI activity, but have suggested a possible benefit on fatigue. 154 Trials of omega-3 supplementation have not shown any effect on disease activity. 155 Randomization to calorie restriction diets was associated with improvement of emotional wellbeing compared to controls, but the study was too short to evaluate risk of relapse. 156

Pregnancy, breastfeeding, and other reproductive factors
There is strong evidence that pregnancy does not worsen long-term prognosis, [157][158][159] despite an increased risk of relapses in the early postpartum period. 160 It is unclear if pregnancy improves MS outcome. Exclusive breastfeeding (no supplemental feedings for at least 2 months) reduces this risk of postpartum relapses, whereas nonexclusive breastfeeding appears to have no effect. 161 Whether breastfeeding influences long-term MS prognosis is unknown.
It is unclear whether hormonal contraceptive use or age at menarche affect long-term prognosis as the sparse data inadequately controlled for other factors. In girls with MS, there may be a small increase in the risk of relapse during the peri-menarche year. 88 Air pollution A few studies have shown an association between PM 10 levels and risk of MS relapse or MRI activity. [162][163][164][165] A deeper look at various pollutants suggested an increased risk of MS relapse following recent exposure to nitrogen dioxide (particularly during the cold season) or ozone (particularly for the hot season). 164 Benzene and carbon monoxide were not associated with MS relapses. Finally, worse air pollution in Iran was associated with poorer recovery from a first MS event. 166 Obesity Whether obesity is associated with a higher risk of MS-related disability is unclear. Obesity is a low-grade inflammatory state 167 and could act as a pro-inflammatory cofactor resulting in an earlier age at onset and more aggressive inflammation. However, the evidence is sparse, mixed, and weak. Obesity prior to disease onset is associated with younger age at onset only in women 168 and earlier age at conversion to secondary progressive MS only in smokers, 169 both with marginal effects and inadequate statistical analyses. Studies in highly selected MS populations found that higher body mass index (BMI) at diagnosis or baseline has either no effect 170 or harmful effects 171 on brain atrophy, yet current BMI is not associated with the current level of disability. 172 In addition to obesity, several studies have focused on the association of comorbidities with MS progression but these will not be reviewed.

Prospects for prevention, prognostication, and intervention
Although the risk of MS is higher in patient's first-degree relatives, most first-degree relatives will never develop the disease. There is currently no genetic testing that is recommended for at-risk individuals as most genetic variants associated with MS are normal genes and most carriers will remain MS free.
The following recommendations are not based on definitive proof-of-concept intervention trials as these are difficult to implement in large cohorts, but are derived from epidemiological findings with the caveat that these studies do not provide precise duration and dose of exposures that may have preventative benefits. Table 1 highlights the gaps in the evidence and the room for future research. Exposure to cigarette smoking, low sun exposure, and childhood obesity are all modifiable risk factors that have been repeatedly and consistently associated with an increased risk of MS. Diets high in fish (and low in saturated fats) are recommended as part of a heart-healthy diet and may also reduce the risk of MS. Thus, recommendations for at-risk individuals are consistent with those for general health benefits and may reduce the risk of MS: avoidance of exposure to first-and second-hand cigarette smoke; using the UV index to guide sun exposure in accordance with WHO guidelines, 173 but ensuring regular sun exposure below a sunburning dose on most days of the week; 174 and preventing childhood obesity by promoting a Mediterranean diet and exercise. Many clinicians also recommend vitamin D supplementation to prevent MS, although the hopes of general health benefits are probably limited aside from preventing rickets or osteoporosis in those individuals with extremely low levels (e.g. <25 nmol/L). While vitamin D 3 supplementation doses of up to 4000 IU/day (as commonly used in MS) do not appear to be harmful, 175,176 this should not be viewed as a replacement for sun exposure, particularly in non-Whites, as sun exposure may provide additional benefits (see Sun exposure and Vitamin D section).
For patients with established MS, some modifications of lifestyle factors can be recommended with reasonable confidence for their likely benefit to overall health, even if MS-specific benefits are unclear. These include smoking cessation, achieving and maintaining normal body weight, and osteoporosis prevention.
Obesity or obesity-related comorbidities have been associated with worse outcomes among people with MS. 177 The health benefits of a Mediterranean-style diet for such comorbidities are well-established. 178 Thus it is reasonable, in the absence of a specific "MS diet" that is shown in well-designed trials to reduce the risk of relapses or disability, to advise patients with MS to adopt a Mediterranean-style diet as a means of reducing obesityrelated comorbidity risk. 179 There is no evidence that pregnancy detrimentally affects long-term MS disability; thus, women with MS who want to have children should be reassured and supported through this life phase. Most women with newborns should be encouraged to breastfeed if they desire. Breastfeeding exclusively may reduce the risk of postpartum MS relapses; however, certain MS treatments should not be resumed while breastfeeding. 180

Conclusion
We have come a long way over the past 50 years in the understanding of risk factors for MS onset and progression. These appear to be extremely complex, and individually heterogeneous. Many genetic variants, most with modest effects, have been reported, but their precise biological role remains to be clarified. Similarly, it is likely that many environmental risk factors are also at play, possibly with modest effects as well. Causality remains challenging to establish for many environmental factors. Trans-ethnic and newer analysis strategies have helped strengthen evidence for EBV, obesity, and sun exposure, and yet point to low vitamin D as a pro-inflammatory cofactor only in Whites. While it is conceivable that some environmental factors may have several biological downstream effects relevant to disease processes (i.e. air pollution could decrease sunlight exposure and activate biological pathways such as the oxidative and aryl hydrocarbon receptor pathways), it is also expected that various environmental exposures could act on a limited number of common efferent pathways (i.e. result in similar biological effects). Nevertheless, despite recent progress, these efferent pathways remain unclear, both for susceptibility (onset) and disease progression. RCTs currently underway, and innovative designs, for example, for dietary interventions, and analysis/modeling, for example, mediation analysis for modeling epigenetic changes as a result of environmental influences, should help to clarify these pathways and beneficial causal influences.

Acknowledgments
We thank Dr Erin Walsh, Research School of Population Health, Australian National University who helped prepare Figure 1.

Authors Contribution
All the authors have performed literature search and review, drafted various sections of the first draft, including tables and figure. The authors have provided edits for the entire manuscript.

Conflict of Interests
E. Waubant has not received any pharmaceutical company honorarium. She is site PI for a Novartis and Roche trial. She has volunteered on an advisory board for a Novartis trial. She is a non-remunerated advisor for clinical trial design to Novartis, Biogen-IDEC, Sanofi, Genentech, Serono and Celgene. She has funding from the NIH, NMSS, PCORI, and the Race to Erase MS. She is the section editor for Annals of Clinical and Translational Neurology, and co-Chief editor for MSARD. R.M. Lucas  instrumental variable (i.e. the genetic polymorphisms) can only affect the outcome by altering the exposure they are an instrument for. 192 That is, the genes involved in the genetic "score" for higher 25OHD levels that is used as an instrumental variable should not affect the risk of MS through any other pathway than causing higher 25OHD levels.
Odds ratio: the odds ratio is an estimate of the increased or decreased risk of disease. A value of >1.0 indicates an increased risk of having the disease; an OR <1 indicates a reduced odds of having the disease (i.e. a protective effect). The value of the OR can be quantitatively interpreted as the percentage increase/decrease in the odds of having the disease for a one-unit increase in the exposure, for example, an OR of 1.20 for ever smoking versus never smoking can be interpreted as a 20% increase in the odds of disease comparing those two categories.
Susceptibility factors: Genetic and environmental factors that alter the risk of developing a disease.
UV irradiance: levels of ambient UV radiation decrease with increasing distance from the Equator.
Vitamin D status: Vitamin D status is measured by the blood concentration of an intermediate metabolite, 25-hydroxyvitamin D (25OHD). Because the 25OHD concentration depends on the ambient UV radiation as well as time in the sun and how much skin is exposed, there may not be a clear latitude gradient (where measurements use a standardized protocol for sample collection and assay). Importantly, the 25OHD concentration is a marker of both vitamin D status and of recent sun exposure, and questionnaire measures of sun exposure are used as a proxy for past vitamin D status.