Sex differences in EAE reveal common and distinct cellular and molecular components

Abstract

Experimental autoimmune encephalomyelitis (EAE) is commonly used as an animal model for evaluating clinical, histological and immunological processes potentially relevant to the human disease multiple sclerosis (MS), for which the mode of disease induction remains largely unknown. An important caveat for interpreting EAE processes in mice is the inflammatory effect of immunization with myelin peptides emulsified in Complete Freund’s Adjuvant (CFA), often followed by additional injections of pertussis toxin (Ptx) in some strains to induce EAE. The current study evaluated clinical, histological, cellular (spleen), and chemokine-driven processes in spinal cords of male vs. female C57BL/6 mice that were immunized with mouse (m)MOG-35–55/CFA/Ptx to induce EAE; immunized with saline/CFA/Ptx only (CFA, no EAE); or were untreated (Naïve, no EAE). Analysis of response curves utilized a rigorous and sophisticated methodology to parse and characterize the effects of EAE and adjuvant alone vs. the Naive baseline responses. The results demonstrated stronger pro-inflammatory responses of immune cells and their associated cytokines, chemokines, and receptors in male vs. female CFA and EAE mice that appeared to be offset partially by increased percentages of male anti-inflammatory, regulatory and checkpoint T cell, B cell, and monocyte/macrophage subsets. These sex differences in peripheral immune responses may explain the reduced cellular infiltration and differing chemokine profiles in the Central Nervous System (CNS) of male vs. female CFA immunized mice and the reduced CNS infiltration and demyelination observed in male vs. female EAE groups of mice that ultimately resulted in the same clinical EAE disease severity in both sexes. Our findings suggest EAE disease severity is governed not only by the degree of CNS infiltration and demyelination, but also by the balance of pro-inflammatory vs. regulatory cell types and their secreted cytokines and chemokines.

1. Introduction

Experimental autoimmune encephalomyelitis (EAE) is an ascending paralytic autoimmune disease of the Central Nervous System (CNS) in animals that predominantly involves myelin-specific T cells and myeloid cells that migrate from the periphery into the CNS. Upon reactivation in spinal cord, these cells attack and damage myelinated axons [1–5]. The EAE model may be useful for understanding the disease process in multiple sclerosis (MS) due to similarities in the formation of demyelinating perivascular white-matter lesions and induction of acute, chronic, and progressive motor disabilities [5–10]. EAE induction involves activation of encephalitogenic T cells upon immunization of susceptible strains of rodents by strain-specific myelin peptides emulsified in complete Freud’s adjuvant (CFA), with an additional requirement in some strains for subsequent injections of Bordetella pertussis toxin (Ptx) [10–15]. Although the cause of MS has defied intense investigation, the assumption is that EAE induction to some extent resembles a similar process in MS that involves induction of pro-inflammatory T and B cells and activation of myeloid cells, perhaps due to CNS damage by microbial infection and/or molecular mimicry of myelin peptides with microbial antigens. In both diseases, emergence of clinical signs likely depends upon the amplification of autoimmune recognition of CNS target antigens [15–19].

It is widely known that MS disproportionately affects females [20–28], although MS disabilities are often more severe in males with progressive disease [28–38]. Yet, few EAE studies have systematically evaluated sex differences in disease induction or severity. Thus, NIH guidelines [39] indicate the need to study both sexes in various disease models.

The goal of this study was to evaluate sex differences as well as the contribution of myelin peptides emulsified in CFA with added Ptx injections to the induction of severe acute EAE in both male and female C57BL/6 mice, with reference in all assays to naïve, untreated control mice as well as mice injected with CFA + Ptx alone. We thus evaluated clinical and histological EAE disease severity, composition of immune cell subtypes in spleen, and mRNA expression levels of a panel of 84 cytokines, chemokines/receptors, and cell process factors in spinal cord tissue from female and male mice in each of the three treatment groups. We found that EAE disease severity is governed not only by the degree of immune cell infiltration into the CNS and demyelination, but also by the induction of pro-inflammatory vs. regulatory splenocyte cell subtypes and spinal cord expression levels of cytokines, chemokines, and receptors, many of which varied in expression in a sex-dependent manner.

2. Materials and methods

2.1. Animals

C57BL/6 male [33] and female [32] mice were obtained from The Jackson Laboratory (Sacramento, CA) at 6–7 weeks of age and used in experiments between 8 and 12 weeks of age. Mice were maintained on a 12-hour light/dark cycle with access to food and water ad libitum in the Animal Resource Facility at the Portland Veterans Affairs Health Care System. This study was carried out in strict accordance with Federal, NIH, and Institutional guidelines using a protocol approved by the Portland VA Animal Care and Use Committee.

2.2. Preparation of the mouse (m)MOG-35–55 antigen and induction of active EAE

Mice of both sexes were categorized in two independent experiments into three groups totaling 10–11 mice/group: Naïve (untreated), CFA (saline, CFA, Ptx), and EAE (mMOG-35–55/CFA/Ptx). The mMOG-35–55 peptide (MEVGWYRSPFSRVVHLYRNGK) was synthesized by NeoMPS (San Diego, CA). EAE mice were immunized in the flanks at four sites with 200 µl total emulsion containing 200 µg of MOG-35–55 peptide and 400 µg of CFA containing 4 mg/ml of heat-killed Mycobacterium tuberculosis [40–41]. Mice were also given injections of 75 ng and 200 ng of Ptx intraperitoneally on days 0 and 2 respectively, relative to immunization [41–42]. Mice in the CFA group received the same immunization schema as EAE mice, with the replacement of saline for mMOG-35–55 peptide as immunogen. The Naïve mice did not receive immunogen or Ptx; however, they were injected with saline and sham-handled in the same manner as treatment groups. The mice were assessed for signs of EAE according to the following scale: 0 = no signs; 1 = limp tail or mild hindlimb weakness; 2 = moderate hindlimb weakness or mild ataxia; 3 = moderately severe hindlimb weakness; 4 = severe hindlimb weakness and mild forelimb weakness or moderate ataxia; 5 = paraplegia with no more than moderate forelimb weakness; 6 = paraplegia with severe forelimb weakness or severe ataxia or moribund condition. Mice were monitored daily for changes in disease score and weight changes until tissues were collected for ex vivo analyses on day 20 post-immunization. The mean sum of daily scores from each mouse in each group from days 8–20 post-immunization was represented as the cumulative disease index (CDI) for the group.

2.3. Histology

On day 20 post-immunization, 4 randomly chosen mice/group were perfused with 1x phosphate-buffered saline (PBS). Spinal columns were collected and placed in 4% paraformaldehyde (PFA) overnight at 4 °C. The following day, spinal cords were extracted and the lumbar spine was sectioned into 10 µm coronal segments. The formalin fixed paraffin embedded (FFPE) sections were stained with either Hematoxylin & Eosin or Luxol Fast Blue [43–44]. Light microscopy with a Zeiss Apo-Tome.2 (Zeiss, Oberkochen, Germany) was used to capture tiled images of the stained slides at 5x zoom followed by ImageJ processing (ImageJ, U. S. National Institutes of Health, Bethesda, Maryland) to assess for infiltration and demyelination. H&E staining was assessed by converting files to RBG Stack images, tracing of white matter, and calibrating the threshold to include only nucleated cells. Percent nucleated cells corresponds to nucleic acid surface area versus white matter cross-sectional area. LFB staining required identification of white matter demyelination by manual outlining. Cross sectional areas of demyelination were identified and recorded as percent of demyelinated area within total white matter.

2.4. Splenocyte preparation and flow cytometry

Spleens were harvested from the remaining 6 mice/group on day-20 post-immunization. Spleens were filtered through 100 µm nylon mesh filters (BD Falcon, Bedford, MA) and suspended in RPMI 1640. Erythrocytes were lysed in 1x red cell lysis buffer (eBioscience, Inc., San Diego, CA) and remaining splenocytes were washed in RPMI 1640. Splenocytes were enumerated using a Cellometer Auto T4 cell counter (Nexcelom, Lawrence, MA) and resuspended in staining buffer (PBS with 0.1% NaN3 and 1% BSA) at 1 × 10⁶ cells/ml. Fc receptors of the splenocytes were bound by rat anti-mouse CD16/CD32 Mouse BD Fc Block™ (BD Bioscience, San Jose, CA) [43,45]. The cells were subsequently incubated with fluorescent tagged antibodies.

The following antibodies were used: CD1d (1B1), CD19 (1D3), CD122 (TM-β1), CD86 (GL1), CD138 (281–2), CD11b (M1/70), CD69 (H1.2F3), CD206 (CO68C2), CD8 (53–6.7), CD4 (RM 4–5), PD-L2 (TY25), CD45 (30-F11), TNFα (MP6-XT22), LY6G (1A8) (BD Biosciences/BD Pharmagen), PD-1 (RMP1–30), CD44 (1M7), RORγ (AFKJS-9), NOS2 (CXNFT) (eBioscience), PD-L1 (10F9G2), T-bet (4B10), CD5 (53–7.3), CD115 (AFS98), LY6G (HK1.4) (Biolegend) and ARG1 (R&D Systems). 7-amino-actinomycin D (7AAD) was used to assess cell survival (Thermo Fisher Scientific).

Cells stained for intracellular or transcription factor markers were fixed in 4% PFA and stained per manufacturers’ protocols (BD Bioscience and eBioscience, respectively). Intracellular stains were carried out on resuspended cells in permeabilization buffer (BD Bioscience), then incubated with antibodies. Transcription factors were stained by cell fixation and permeabilization with buffers per the manufacturer’s instructions (eBioscience) followed by antibody incubation. All stained cells were analyzed with a BD Accuri™ C6 (BD Bioscience) containing a four-channel fluorescence flow cytometry display (FITC, PE, PerCP Cy5.5, and APC).

2.5. RNA isolation and RT-PCR

Spinal cord tissue was excised from three randomly chosen mice from the Naïve, CFA, and EAE groups, then pooled, gently homogenized, and cleared by centrifugation at ~21,900 rcf at 4 °C, and total RNA was isolated and purified using the RNeasy Mini Kit (Qiagen, Valencia, CA) following the manufacturer’s instructions [43]. After total RNA isolation, the concentration and the 260:280 ratio were determined using the NanoDrop™ One/OneC Microvolume UV–Vis Spectrophotometer (Thermo Fisher, Waltham, MA). All the samples used in the experiment showed a 260:280 ratio between 1.8 and 2.0. After isolation, total RNA was subjected to a genomic DNA elimination step before proceeding to first strand cDNA synthesis using the reverse transcriptase provided with the RT2 HT First Strand kit (Qiagen, Valencia, CA). cDNA samples were loaded onto an RT² Profiler PCR Array (96-well format) Plate for Mouse Chemokines and Receptors (Qiagen Cat. No. 330231 PAMM-022ZA) for further amplification of gene expression. The pooled samples were split into triplicate assays and run as one replicate per card. The panel included 84 mouse-specific chemokines, cytokines, and receptors, as well as five housekeeping genes and several artificial probes for quality control (QC). Amplification was carried out using the Applied Biosystems StepOnePlusTM model operated following the manufacturer’s instructions. Single RT-PCR assays were used to confirm significant results and determine the effects of genes involved in the MIF/CD74 axis per the manufacturer’s protocol.

2.6. Statistical analysis

Prism software 6 (GraphPad Software, La Jolla, CA) was used for data analyses of the daily clinical scores, CDI, spinal cord histology, single-assay qPCR, and flow cytometry. The Mann-Whitney U test was used sequentially by measurement time for determining significance for disease course. Histology, qPCR and flow cytometry data were analyzed using ANOVA with a Fisher’s Least Significant Difference post-hoc test or Student’s t-test where appropriate. A p-value of ≤0.05 was considered significant. All quantification was carried out in blinded fashion.

Stata software (version 16.1; StataCorp, College Station, TX) was used to interrogate the Ct values reported from the per-protocol RT-PCR analysis of the Qiagen PAMM-022ZA array used to measure chemokine expression in the pooled murine spinal cord tissue samples from each experimental group. Data were monitored thoroughly for artifacts and subjected to extensive QC rules based on the performance of control probes on the array to monitor genomic DNA contamination, PCR efficiency, and RT failures. All QC go/no-go criteria recommended by Qiagen [46] were met for each array card. The profile plots showed very strong amplification curves for each of the 84 chemokines on the panel, with no evidence of genomic DNA contamination. After evaluating amplification performance and counts, we excluded 10 chemokines from further analysis for having variant or too-low expression, including Ccl20, Ccl24, Ccr1l1, Cmtm2a, Cxcl1, Cxcl9, Cxcl11, Cxcl15, Cxcr1, and Ifng. Thus, we retained data from 74 chemokines for the study.

2.6.1. Normalization

Two of the housekeeping genes on the array (Actb and Gapdh) were found not to vary by sex or experimental condition and were used as endogenous normalizers. Minor differences between treatments for Actb and Gapdh were regularized (i.e. subtracted) out separately via regression using a fixed-effect absorbing regression adjustment [47] and the adjusted values were averaged to yield a normalizer. Normalization was accomplished by subtracting the subject-specific mean for the normalizer (i.e. the average of Actb and Gapdh after adjustment) from each measured Ct value for the subject and then recentering the entire dataset to the raw Ct median.

2.6.2. Expression analysis

Ct values from RT-PCR may become unreliable at very low input concentrations (generally, at Ct ≥ 35 or so) and cannot be trusted to give fully quantitative information in those ranges. Since the recorded Ct represents an upper boundary on the expression (i.e. the true expression is no greater than the value represented by the observed Ct), the result is actually an interval-censored observation: We know that the true Ct must fall somewhere between the observed value and 40 (when the PCR cycling ends). Thus, we treat Ct = 40 as the zero expression point and for any Ct value x ≥ 35 we say that the true (unobserved) Ct falls within the interval [x,40]. We then use interval-censored regression methods [48] to estimate treatment group means. Specifically, we employed a fully Bayesian estimator of the interval-censored regression coefficients (the ‘bayes: intreg’ command in Stata) for each chemokine individually, with a vague but proper exchangeable zero-centered Student’s t test prior (4 degrees of freedom with scale set to 10) on the design coefficients (fully factorial in sex and treatment condition) and a Cauchy prior (also scale 10) on the log residual standard deviation. Monte Carlo Markov chain (MCMC) sampling was used to estimate the posterior distribution of the coefficients (burn-in run of 2500 simulation samples followed by 10,000 posterior samples), and initial values for the MCMC were taken from maximum-likelihood estimates of the model. Mean expression levels and standard errors for each chemokine for each sex and condition were derived as marginal ergodic summaries of the joint posterior distribution, and similarly for inferential questions such as posterior probabilities of linear combinations of model coefficients (hereafter “effects”) being of a given sign or having magnitude within a specified range. For example, for each sex for each chemokine, we calculated the posterior probability that the EAE effect for the chemokine within that sex was >0.5 Ct (a value chosen based on the standard deviation of Ct values for the stable artificial control probes included on the array). Log₂ fold change estimates and associated standard errors, as well as all ratios or sums of such values, were based on these model results, and inference about significance was based on posterior probabilities of specific outcomes (e.g. that a given effect is positive). The following contrasts were estimated for each gene for each sex: CFA vs Naïve conditions (CFA-vs-N), EAE vs Naïve conditions (EAE-vs-N), EAE vs CFA conditions (EAE-vs-CFA), and EAE vs the average of the Naïve and CFA conditions (EAE-vs-CFA + N). For an expected profile where only the EAE induction is active for chemokine expression, the EAE-vs-CFA + N contrast is the best estimate of the effect of the encephalitogen, mMOG-35–55. In cases where the CFA reagents were also active, comparisons involving the EAE-vs-N and EAE-vs-CFA contrasts were needed. The CFA-vs-N contrast was compared to the others to estimate a response shape.

2.6.3. Clustering analysis

To facilitate an organization of the chemokines based on response patterns, we devised 10 metrics (SUPPLEMENTAL TABLE 1) that reflect qualitative features of the response curves (i.e. the trajectory of response levels across treatment conditions) for each chemokine, including summaries of the shape, pitch, and general clarity of the curves, as well as measures of the concordance or discordance of the male and female response trajectories. These 10 metrics were a highly coherent summary of the data. A principal components analysis followed by multivariate correlation of the 10 metrics to the top four informative principal components showed that they collectively accounted for 82% of the response-shape variance in the assay. We then carried out a clustering analysis using hierarchical Ward’s linkage and Canberra distance to group chemokine profiles [49].

2.6.4. Definition of effect

It was important to clearly define the appropriate metric for examining the “EAE effect” in each kind of response pattern. An essentially linear response pattern points to some confounding of the EAE effect by CFA influences that are already present, so we need to remove those influences before assessing the effect of EAE. But if the EAE and CFA effects are in opposite directions (possibly due to measurement noise), then performing this adjustment has the unintended consequence of exaggerating the EAE effect, so in this case the raw EAE-vs-N contrast is the appropriate metric. Finally, if the response shape is highly nonlinear and the EAE effect is overwhelmingly dominant, then the contrast between the EAE condition and the average of the CFA and N conditions is the most powerful metric. Thus, to measure the effect of EAE, we use the EAE-vs-N contrast if the CFA-vs-N and EAE-vs-N contrasts differ in sign; we use the EAE-vs-CFA contrast if the response shape is monotone and approximately linear or if the ratio of the CFA and EAE effects is large (so that CFA is a large contributor to the total effect); and we use the EAE-vs-CFA + N contrast if the response shape is monotone and strongly nonlinear where the ratio of the CFA and EAE effects is small (indicating low contribution from CFA to the expression change). For each EAE effect estimate presented below, we indicated which contrast was chosen to define the EAE effect. The CFA effects were estimated directly as the CFA-vs-N contrast for each sex. Several examples of curve shape effects for factors important for EAE and CFA in females vs. males are shown in HYPERLINK “sps:refid::app1” SUPPLEMENTAL FIGURE 1.

2.6.5. Ranking and selection

We prioritized confidently positive effects, defining “confidently” differently for CFA and EAE effects (hereafter, ΔCFA and ΔEAE). For CFA effects, we required that both ΔCFA and ΔEAE were strictly positive and present, and that the contribution of CFA to the total effect of EAE (i.e. the ratio |ΔCFA|:|ΔEAE|) was substantial. Specifically, we required that the ratio be >1/3, and the posterior probabilities that ΔEAE > 0.5 Ct and that ΔCFA > 2 Ct were each at least 2/3 (i.e. minimum 2:1 odds in favor of the effect being large). In some cases, |ΔCFA| was larger than |ΔEAE| (possibly due to measurement noise). In these cases we required that the effects were within 1 Ct of one another to be counted as coherent for the chemokine to be considered as a CFA candidate. For EAE effects, we defined significance as having a posterior probability of at least 90% that ΔEAE > 0.5 Ct, <1/3 contribution of CFA to the total effect (i.e. |Δ CFA|:|ΔEAE| <1/3), and some additional coherence criteria (i.e. that |Δ EAE| > |ΔCFA| and that there was a minimum of 0.5 Ct separation between the conditions) to ensure acceptable curve shape.

2.6.6. Evaluation of total (aggregate) chemokine expression

For all included chemokines, we evaluated total expression both cross-sectionally (between sexes) and longitudinally (within sex). For both analyses we calculated the antilog (i.e. exponentiating to the base of 2) of the posterior mean Ct value for each sex and condition for each chemokine, and took this as the expression level. The rationale for this is that Ct values from RT-PCR are theoretically counts of doublings. Although these values are only relative to the endogenous control levels (and thus have no absolute quantitative interpretation), nevertheless ratios of sums of these are a coherent measure of relative expression across groups. For the cross-sectional analysis we simply summed the expression levels for each group and assessed the female:male ratio within each treatment condition. For the longitudinal analysis we calculated average trajectories across chemokines by sex and treatment condition using a nonparametric kernel-based regression method [50] (Epanechnikov kernel) with bootstrapped standard errors (200 bootstrap replicates) and optimal bandwidth chosen by cross-validation. Additionally we ranked the chemokines separately within each sex according to percent contribution of the chemokine’s expression level to the total, for each treatment condition.

3. Results

3.1. Sex differences in EAE induction

To investigate the role of cell types and associated factors in the inflammatory process in EAE induction, we compared clinical and histological signs of disease in Naïve, CFA/Ptx-immunized (CFA) and CFA/Ptx/MOG-35–55 peptide-immunized (EAE) groups of C57BL/6 male and female mice that were 8–12 weeks old at immunization.

3.1.1. Clinical and histological EAE

As shown in Fig. 1A and B, clinical signs at onset of EAE were evident on Day 12 (mean scores of 0.5 for males and 0.7 for females) daily disease severity and cumulative EAE disease scores through Day 20 were the same in males (CDI 31 ± 2) and females (31 ± 3) immunized with CFA/Ptx/MOG-35–55. Clinical signs of EAE were not observed in CFA or Naïve groups as expected. Moreover, spinal cord H&E sections likewise revealed significantly increased cellular infiltration in the EAE and CFA groups vs. the Naïve groups, in the EAE groups vs. the CFA groups and in females vs. males in both the CFA and EAE groups (Fig. 2A). However, spinal cord LFB sections revealed demyelination only in the EAE groups, with females having significantly greater demyelination than males (Fig. 2B). Thus, females had greater cellular infiltration and demyelination than males in spite of essentially identical EAE disease severity. Moreover, both male and female CFA groups experienced cellular infiltration without demyelination.

Fig. 1.

EAE disease course is similar in C57BL/6 wt male and female mice. Naïve (sham handled), CFA (immunized with complete Freund’s adjuvant – CFA – and pertussis toxin – Ptx) and EAE (immunized with mMOG-35–55/CFA/Ptx) were scored over a 20-day period. Mean clinical EAE daily disease scores (top) and CDI scores (bottom) are shown for A) male and B) female mice. Significant differences between groups in daily clinical EAE and CDI scores were determined by sequential testing using the Mann–Whitney U test and unpaired two-tailed t-test respectively (***p ≤ 0.001).

Fig. 2.

Lymphocyte chemotaxis and white-matter demyelination are exacerbated in females with CFA/Ptx-initiating chemotaxis and mMOG-commencing demyelination. Spinal cord histology reveals infiltration with immunization of CFA + Ptx alone in both sexes and demyelination only with the addition of mMOG-35–55, both characteristics more severely affecting females (n = 4, all groups) than males (n = 5, all groups). A) H&E staining represented in percent nucleated cells per area reveals initiation of lymphocyte chemotaxis with the introduction of CFA + Ptx in both males and females compared to naïve mice (p < 0.05 males, p < 0.001 females) and between male and female CFA groups (p < 0.05). Cohorts with the addition of mMOG-35–55 exhibited concentrated areas of white matter infiltrate compared to the CFA groups (p < 0.001 males, p < 0.001 females), with significantly greater lymphocyte chemotaxis in females vs males (p < 0.05). B) LFB staining represented in percent demyelination per area reveals exacerbated axonal damage in female EAE mice compared to males (p < 0.05). All percentages are presented with single-SEM error bars (*p < 0.05, **p < 0.01, ***p < 0.001).

3.2. Sex differences in splenic cell types

Spleens from 6 mice in each group were evaluated individually on Day 20 after EAE induction and total cell numbers were established. As shown in Table 1, Naïve females had significantly higher cell counts per spleen than Naïve males (60 vs. 40 million). Compared to Naïve mice, CFA/Ptx immunization significantly increased splenocyte numbers in both males and females, although the CFA males had significantly higher splenocyte counts than CFA females (90 vs. 70 million). Male and female EAE groups had the same numbers of splenocytes (60 million). This reduction of cell numbers in EAE vs. CFA groups was anticipated due to increased migration of immune cells from the spleen to the CNS during EAE induction.

Table 1.

Sex differences in splenic cell types. With the exception of Total Cell Numbers from spleen, numbers in the Table for CD4+, CD8+, CD19+ and CD11b+ cell types represent percentages of total gated mononuclear cells. Numbers for the remaining cell subsets represent percentage of the major cell types.

Cell Type		Male (Mean±SEM)			Female (Mean+SEM)			Male vs Female (p≤value)
		Naïve	CFA	EAE	Naïve	CFA	EAE	Naïve	CFA	EAE
Splenocytes	Total Cell Number	40±4M	90±1Ma***	60±12Mc*	60±6M	70±1Ma*	60±5Mc*	M<F*	M>F***
T cells	CD4+	12±1	10±1	10±1b*	15±1	11±14a*	13±1	M<F**		M<F*
CD4+ helper	CD4+CD44+	23±1	52±1a***	41±2b***c***	24±2	43±2a***	44±3b***		M>F***
	CD4+CD69+	22±3	23±1	28±2c*	7±1	7±1	9±1	M>F***	M>F***	M>F***
	CD4+T-bet+	18±1	42±2a***	28±3b**c**	23±1	28±2	21±3c*	M<F*	M>F***
	CD4+RoRγ	1±0	5±0a***	4±1	2±1	2±0	1±0		M>F***
T-reg	CD4+D122+	14±1	19±1a**	14±1c*	6±0	9±0a***	10±2b*	M>F***	M>F***
Checkpoint	CD4+PD1+	18±1	23±2a*	17±1c*	9±1	16±0a***	18±2b***	M>F***	M>F**
	CD4+PDL1+	9±1	32±3a***	18±2b**c**	7±1	16±1a***	16±3b*		M>F***
	CD4+PDL2+	11±1	9±1	13±0b*c**	6±1	11±1a***	11±1b**	M>F***
	CD8+	8±1	4±0a***	6±1	10±0	5±0a***	6±1b***	M<F**
T-reg	CD8+CD122+	15±1	20±1a*	19±1b*	6±0	11±0a***	12±1b***	M>F***	M>F***	M>F**
Checkpoint	CD8+PD1+	16±1	15±2	17±2	3±0	8+±0a***	8±2b*	M>F***	M>F***	M>F**
B Cells	CD19+	65±1	35±4a***	32±3b***	63±1	40±3a***	41±3b***
B-reg	CD19+CD5+CD1hi	3±0	8±0a***	7±1b***	2±0	6±2	5±1b**	M>F*
	CD19+CD138+CD44hi	1±0	3±0a***	5±1b**	1±0	2±0a***	2±0b**		M>F**	M>F*
Checkpoint	CD19+PDL1+	8±1	23±1a***	20±2b***	8±1	17±1a***	16±3b**		M>F**
	CD19+PDL2+	7±0	5±0a**	8±0b*c***	4±0	5±0a*	7±1b*	M>F***
Mac/Mono	CD11b+	7±0	23±2a***	40±3b***c**	7±0	30±2a***	27±4b***		M<F*	M>F*
Pro-inflam	CD11b+TNF+	45±3	52±3	52±1b*	70±4	62±5	54±6b*	M<F***
	CD11b+CD86+	41±5	15±2a***	17±2b**	39±2	12±1a***	12±1b***
Anti-inflam	CD11b+Arg1+	46±7	32±3	26±1b*	49±3	30±3a***	20±3b***c*
	CD11b+CD206+	33±4	17±4a*	20±1b*	32±3	28±6	20±4b*
	CD11b+iNOS+	50±4	43±4	49±1	55±4	47±4	43±6
	CD115+	13±1	22±1a***	23±1b***	3±0	12±1a***	11±1b***	M>F***	M>F***	M>F***
	CD115+CD45+LY6C+G	51±1	58±4	34±1b***c**	66±0	59±3a*	61±1	M<F***		M<F***
Checkpoint	CD11b+PDL1+	40±2	30±4a*	13±1b***c**	43±2	17±2a***	19±2b***		M>F***	M<F*
	CD11b+PDL2+	26±1	16±3a**	13±1b***	30±1	13±1a***	17±2b***

Legend: M = Male; F = Female

^a= Naïve vs CFA

^b= Naïve vs EAE

^c= CFA vs EAE

^*= p ≤ 0.05

^**= p ≤ 0.01

^***= p ≤ 0.001.

To evaluate effects on spleen cell subpopulations in Naïve, CFA and EAE groups, we carried out 4-color FACS staining of splenocytes from the 6 individual mice from each treatment group (Table 1). Two major changes were observed in spleen cell populations after immunization. First, there was a reduction of B cells from >60% in Naïve mice to 32–41% of spleen cells in EAE and CFA immunized mice. Although generally reduced after immunization, the splenic B cells showed modest increases in regulatory CD5⁺CD1d^hi and CD138⁺CD44^hi subtypes (M only) as well as strong increases in B cells expressing the checkpoint inhibitor marker PD-L1⁺ (M > F). Second, the reduced percentages of B cells were largely replaced by CD11b⁺ and CD115⁺ monocytes/macrophages that increased from ~7% in Naïve mice to ~25–40% in CFA and EAE mice. This increase after immunization was reflected by stable or increased percentages of monocyte/macrophage subtypes expressing the pro-inflammatory cytokine TNFα (M only) and neuroprotective CD115 (M > F) markers compared to the Naïve group. In contrast, there were marked reductions in EAE spleens in the percentage of CD11b⁺ cells expressing the pro-inflammatory marker CD86, anti-inflammatory markers Arg1 and CD206, the neuroprotective marker CD115⁺CD45⁺LY6C^hiG⁻ (M only) and the checkpoint inhibitor markers, PD-L1 and PD-L2. Interestingly, there were significant differences in the EAE vs. the CFA groups of macrophages/monocytes, including higher percentages of CD11b⁺ spleen cells (40 vs. 23%, M only) and lower percentages of Arg1⁺ (20 vs. 30%, F only), PD-L1⁺ (13 vs. 30%, M only) and CD115⁺CD45⁺LY6C^hiG⁻ cells (34 vs. 58%, M only). Such strong reductions may reflect migration of these splenic monocyte subpopulations into the damaged CNS.

Differences between male vs. female monocyte/macrophage markers were also observed in the Naïve groups, including increased percentages in males of CD115⁺ (13 vs. 3%) and in females of TNFα⁺ (70 vs. 45%) and CD115⁺CD45⁺LY6C^hiG⁻ cells (66 vs. 51%); in the CFA groups, including increased percentages in males vs. females of PD-L1⁺ (30 vs. 17%) and CD115⁺ cells (22 vs. 12%) and in females of CD11b⁺ cells (30 vs. 23%); and in the EAE groups, including higher percentages in males of CD11b⁺ (40 vs. 27%) and CD115⁺ cells (23 vs. 11%), and in females of PD-L1⁺ (19 vs. 13%) and CD115⁺CD45⁺LY6C^hiG⁻ cells (61 vs. 34%).

CD4⁺ and CD8⁺ T cells that together comprised ~20–25% of Naïve splenocytes had relatively smaller percentages in CFA (14–16%) and EAE (16–19%) splenocytes. However, there were many differences in CD4⁺ and CD8⁺ T cell subsets. EAE vs. Naïve mice had significantly increased levels of CD4⁺CD44⁺ (~40 vs ~20%, M = F), CD4⁺T-bet⁺ (28 vs. 18%, M only), CD4⁺PD-1⁺ (18 vs.9%, F only), CD4⁺PD-L1⁺ (~17 vs. ~8%, M = F), CD4⁺PD-L2⁺ (13 vs. 11%, M; & 11 vs. 6%, F), CD8⁺CD122⁺ (19 vs. 15%, M; & 12 vs. 6%, F), and CD8⁺PD-1⁺ (8 vs. 3%, F only) on T cells vs. corresponding values in Naïve mice.

One metric in particular that would appear to distinguish pathogenic vs. non-pathogenic splenic subsets is significantly decreased percentages of inflammatory cell phenotypes in the EAE group vs. the CFA group that likely would reflect their enhanced migration from spleen to injured CNS. Such phenotypes include CD4⁺CD44⁺ (memory T cells: 41 vs. 52%, M only) and CD4⁺T-bet⁺ (Th1 cells: 28 vs. 42%, M; 21 vs. 28%, F) T cell subsets. Additionally, significant decreases were noted in regulatory and checkpoint phenotypes in the EAE vs. the CFA groups, including CD4⁺CD122⁺ (14 vs. 19%, M only), CD4⁺PD-1⁺ (17 vs. 23%, M only) and CD4⁺PD-L1⁺ (18 vs. 32%, M only) T cells, and CD11b⁺Arg1⁺ (20 vs. 30%, F only), CD115⁺CD45⁺LY6C^hiG⁻ (34 vs. 58%, M only), and CD11b⁺PD-L1⁺ (13 vs. 30%, M only) Mono/Mac subtypes, suggesting a selective migration of these cells into the CNS that could counteract the pro-inflammatory subtypes.

In summary, these results suggest that the stronger pro-inflammatory T cell responses observed in male CFA and EAE T cell subsets appeared to be partially offset by increased percentages of male anti-inflammatory, regulatory, and checkpoint T cell, B cell, and monocyte/macrophage subsets. These differences in peripheral immune responses may explain the reduced cellular infiltration in the CNS lesions of male vs. female CFA immunized mice and the reduced CNS infiltration and demyelination observed in male vs. female EAE groups of mice (Fig. 2) that ultimately resulted in the same EAE disease severity in both sexes (Fig. 1). These findings suggest EAE disease severity is governed not only by the degree of CNS infiltration and demyelination, but also by the balance of pro-inflammatory vs. regulatory infiltrating cell types.

3.3. Inflammatory versus regulatory factors in spinal cords

Spinal cord tissues from 3 mice from each experimental group were evaluated for mRNA and chemokine/receptor expression using RT-PCR.

3.3.1. Sex differences in the MIF/CD74 axis

Due to our interest in the role of the MIF/CD74 axis as a disease modifier in EAE, we evaluated RNA expression in spinal cord tissue from five individual mice from each group on Day 20 after immunization. As shown in Fig. 3, there was a striking and highly significant increase in CD74, the common macrophage migration inhibitory factor (MIF) and _D-dopachrome tautomerase (D-DT) receptor, in both EAE males and EAE females, with relative expression values significantly higher in females, and a modest increase in expression of CD74 observed in CFA females. Moreover, there was a modest increase in expression of the CD44 co-receptor in EAE males but not EAE females. Increased expression of MIF was observed in CFA females but not in other groups. Also, a slight increase in expression of D-DT was observed in CFA males (and a slight decrease in CFA females), but a large decrease in D-DT expression was observed in both EAE males and EAE females. These results indicate involvement of CD74 and CD44 in the male EAE group, with only CD74 involvement in the female EAE group. In contrast, the expression of CD74 and MIF was higher in the CFA females than in the CFA males. This pattern suggests that the MIF/CD74 axis could in part account for the increased inflammation and cellular infiltration observed in the CFA female group.

Fig. 3.

Expression of MIF, D-DT and their cognate receptor complex CD74/CD44 genes is sex-dependent. RNA from the spinal cords of representative Naïve, CFA, and EAE-treated mice (n = 5) were pooled, converted to cDNA, and amplified via qPCR to assess MIF-CD74 axis involvement in disease induction. A) While the CD74 gene is expressed at the same levels in Naïve or CFA-treated male and female mice, EAE induced a dramatic upregulation of CD74 gene message (p < 0.001) as evaluated by qPCR, in which females expressed higher levels of CD74 message than the cohort male group (p < 0.001). B) However, CD44 message expression was only significant in males with EAE with respect to Naïve and CFA groups (p < 0.001) but not in females. CD44 expression was slightly downregulated in CFA-treated females (p < 0.05). C) In contrast, MIF gene expression was upregulated in females but not in males under CFA conditions compared to Naïve experimental animals, but MIF message was downregulated in males as well as in females in the EAE group. D) Likewise, D-DT gene expression was shown to be downregulated under EAE experimental conditions when this group was compared to Naïve or CFA groups (p < 0.001) but females showed a deeper downregulation of levels (p < 0.001) than males in the EAE cohort. It is noteworthy that D-DT is slightly upregulated only in CFA-treated males compared to Naive males (p < 0.05) and moderately downregulated in females treated with only CFA. All data were analyzed using Prism software and are presented with single-SEM error bars (*p < 0.05, **p < 0.01, ***p < 0.001).

3.3.2. Inflammatory chemokine array analysis

The reliability of the RT-PCR assays for chemokine/receptor expression appeared to be excellent. Assessing components of variance, we found that 73% of the variance in the Ct values was due to gene differences, and 17% to sex and treatment differences, summing to 90% explained by known experimental factors, with just 10% due to unexplained residual variance. This allowed us to state the following analytical results with reasonably good confidence, despite the limited sample size and exploratory nature of the analysis.

As detailed above (see Methods), we performed a hierarchical clustering analysis of the chemokines ligands and their receptors based on 10 derived features of each gene’s response profile across conditions and sexes. The clustering analysis yielded a very coherent and interpretable classification that we describe in an annotated dendrogram (Fig. 4). Cutting this dendrogram at a distance of ~10 – corresponding to the equivalent of a “1 unit” difference on every single feature — formed 6 very distinct clusters. Numbering from left to right, the first two clusters are noisy examples of positive trends, differing in the similarity between sexes; cluster 1 contains genes where male and female responses are similar, and cluster 2 contains genes where they disagree on some important feature. Cluster 3 consists of a group of genes with no systematic trend and generally low weight of evidence. Overall, clusters 1 through 3 are high-variance-low-expression genes where males and females sometimes have quite different baseline levels and response trajectories. However, cluster 1 contains genes with some of the most extreme EAE responses observed in the experiment (e.g. Ccl8 and II1b). The right fork at the top of the dendrogram also houses three clusters (clusters 4 through 6), and in contrast to the left fork, these clusters are of genes showing generally high expression with low residual variance and a broad similarity between male and female trajectories. Cluster 4 is the largest, containing 25 genes, nearly all of which show a positive trend with little to no effect of CFA. Clusters 5 and 6 contain genes with weak or very uncertain trends where a CFA effect (sometimes in the opposite direction of the EAE effect) is prominent. Cluster 5 contains the most erratic patterns, with little agreement between males and females, while cluster 6 is more coherent and also contains the small subset of unusual genes where there appears to be downregulation of expression under EAE rather than the more common pattern of upregulation.

Fig. 4.

Hierarchical clustering of chemokines reveals biologically relevant groupings. A hierarchical Ward’s linkage clustering based on Canberra distance between the genes with respect to 10 response curve feature metrics (see Supplemental Table) yields a very coherent and interpretable classification of the panel of chemokines into 6 distinct clusters (numbered from left to right) when the dendrogram is cut at a distance of ~10, corresponding to the equivalent of a “1 unit” difference on every feature. The top-level split is between primarily low-expression genes on the left and high-expression genes on the right, and the second-level splits separate clearly trending genes (i.e. those where the gene is upregulated under the EAE condition) from those where a trend is either absent or unclear. Lowest-level splits (separating cluster 2 from 1, and 6 from 5) are mainly based on whether male and female response patterns are similar (F–M) or different (F⊥M). Cluster 4 is the largest, containing 25 genes, nearly all of which show a positive trend with little to no effect of CFA.

Using the dendrogram as a guide, we identified several classes of factors that differ in their contribution to EAE induction. A heatmap showing the direction, intensity, and precision of log₂ fold change estimates for EAE effects is presented in Fig. 5. The genes are ordered by cluster number (from the dendrogram) and then by maximum effect size (across sexes) within each cluster. The size of the bars is proportional to the precision of the estimated effect (i.e. larger bars indicate higher confidence that the effect is in the estimated direction). We see quite clearly in the heatmap that clusters 4 (high confidence) and 1 (low confidence) show similar effects between sexes, while clusters 2 and 3 tend to show effects that differ greatly in magnitude between sexes. Most effects are positive (reddish in color), and the small number of possibly negative effects (bluish in color) tend to be either very close to zero (gray) or very imprecise (shorter and thinner bar) and thus possibly measurement artifacts; the few exceptions (e.g. Ackr1, Slit2, and Cx3cl1) to this may represent actual downregulation of expression under EAE. Clusters 5 and 6 are mostly comprised of stable high-expression genes that do not appear to be affected much by EAE. To identify the most prominent chemokine players in EAE induction, we looked for genes where the CFA effect (ΔCFA) is negligible and where the EAE effect (ΔEAE) is confidently large (see below for precise definitions; also see Methods).

Fig. 5.

Heatmap of EAE effects organized by chemokine cluster. The direction (hue), intensity (saturation), and precision (bar size) of log₂ fold change estimates for EAE effects are shown, ordered by cluster number (from the dendrogram in Fig. 4) and then by maximum effect size across sexes within each cluster. Larger bars indicate higher confidence that the effect is in the estimated direction. Clusters 4 (high confidence) and 1 (low confidence) show similar effects between sexes, while clusters 2 and 3 tend to show effects that differ greatly in magnitude between sexes. Most effects are positive, while the small number of possibly negative effects tend to be either close to zero or imprecise (and thus possibly measurement artifacts). Clusters 5 and 6 are mostly comprised of stable high-expression chemokines that do not appear to be much affected by EAE.

A similar approach was used to determine the contribution of the CFA + Ptx components to the EAE score. The idea was to identify genes where ΔCFA and ΔEAE are both present (i.e. of nonnegligible magnitude), where the effect direction is positive and monotone (i.e. ΔCFA and ΔEAE have positive sign with |ΔCFA| ≾ |ΔEAE|), and where the percent contribution of CFA to the total effect of EAE (i.e. the ratio |Δ CFA|:|ΔEAE|) is substantial. Since we required that ΔCFA and ΔEAE have the same sign, the ratio of the two effects ranges from 0 (no change under CFA + Ptx treatment compared to the Naïve group) to 1 (CFA + Ptx accounted for essentially all changes in the indicated factor observed in the EAE group); in a few genes where |ΔCFA| was nominally larger than |ΔEAE| (e.g. Ppbp for females and Il4 for males), we truncated the ratio at 1 and allowed the gene to be selected as long as ΔCFA was within 1 Ct of ΔEAE (or up to 100% of |ΔEAE|, if |ΔEAE| < 1 Ct). To count a gene as showing substantial contribution of CFA to the EAE effect, we required |ΔCFA|:|ΔEAE| > 1/3, and considered genes with |ΔCFA|:|Δ EAE| < 1/3 as having negligible CFA contribution.

Table 2 shows 40 prominent EAE-associated genes, defined as all those with |ΔCFA|:|ΔEAE| ratio <1/3 and posterior probability of at least 90% that ΔEAE > 0.5 Ct (strictly positive; i.e. is upregulated) for either or both sexes. In addition we required acceptable curve shape (i.e. |ΔEAE| > |ΔCFA|) and separation of at least 0.5 Ct between the CFA and EAE conditions. The ‘Contrast’ column for each sex indicates the most appropriate comparison to assess ΔEAE, based on CFA contribution amount and curve shape, as detailed above (see Methods). Note that information for both sexes is presented in the row for each gene, and significant effects are in bold. Some of the nonsignificant factors for a particular sex in the table are in fact significant CFA contributors for that sex (e.g. Xcl1 for males and Ccl2 for females), and thus appear in both groups. Rows are sorted in descending order of maximum EAE effect size (across sexes). Of the 40 highlighted genes, 13 are important for males only, 4 are important for females only and 23 are important for both sexes. These data are shown in a heatmap illustrating the relative magnitudes and posterior probabilities of all significant EAE effects for each sex [SUPPLEMENTAL FIGURE 2]; note that blanks in the heatmap indicate either that the contribution of CFA to the total effect is at least 1/3, or that ΔCFA or ΔEAE are incoherent or too small or lacking in precision to be significant for the sex in question. The rows are sorted in descending order of maximum probability of the effect (across sexes), and quantitative log₂ estimates of the EAE fold changes are displayed inside the bars, which are proportional in size to the ΔEAE values.

Table 2.

EAE-associated factor expression by dendrogram cluster. Significant effects for each sex are in bold. Rows are sorted in descending order of maximum EAE effect size, ΔEAE (across sexes). Of the 40 highlighted genes, 13 are important for males only, 4 are important for females only and 23 are important for both sexes.

Gene	Cluster	ΔCFA/ΔEAE (F)	ΔCFA/ΔEAE (M)	Contrast (F)	Contrast (M)	ΔEAE (F)	ΔEAE (M)	P(ΔEAE > 0.5) (F)	P(ΔEAE > 0.5)(M)
Ccl8	1	0.0129	0.1734	EAE-vs-CFA + N	EAE-vs-CFA	8.2814	9.9604	0.9972	1.0000
Tnf	1	(sign diff.)	0.2661	EAE-vs-N	EAE-vs-CFA	4.7352	7.4499	0.9725	1.0000
Ccl5	4	(sign diff.)	0.1221	EAE-vs-N	EAE-vs-CFA + N	4.8169	6.5496	1.0000	1.0000
Ccr7	4	(sign diff.)	0.0611	EAE-vs-N	EAE-vs-CFA + N	6.3203	6.0530	0.9961	0.9999
Ccl22	1	(sign diff.)	0.2615	EAE-vs-N	EAE-vs-CFA	6.2480	4.8153	0.9959	0.9961
Xcl1	2	(sign diff.)	0.5298	EAE-vs-N	EAE-vs-CFA	6.1802	2.9540	1.0000	1.0000
Cxcr3	1	0.0453	(sign diff.)	EAE-vs-CFA + N	EAE-vs-N	5.8199	6.1231	0.9552	0.9974
Ccl9	1	0.0088	(sign diff.)	EAE-vs-CFA + N	EAE-vs-N	5.9227	3.7267	0.9933	0.9563
Cxcl10	4	0.2010	0.0980	EAE-vs-CFA + N	EAE-vs-CFA + N	3.4500	5.9045	1.0000	1.0000
Ccl2	2	0.3859	(sign diff.)	EAE-vs-CFA	EAE-vs-N	5.6877	5.7123	0.9980	0.9966
C5ar1	1	0.5925	0.2508	EAE-vs-CFA	EAE-vs-CFA	3.1019	5.6867	0.8674	0.9573
Ccr8	1	0.0074	(sign diff.)	EAE-vs-CFA + N	EAE-vs-N	5.6396	4.6786	0.9744	0.9990
Cxcl16	4	(sign diff.)	(sign diff.)	EAE-vs-N	EAE-vs-N	3.3198	4.9303	1.0000	1.0000
Ccl12	4	(sign diff.)	0.2554	EAE-vs-N	EAE-vs-CFA	4.3269	4.8419	1.0000	1.0000
Ccl7	2	0.6561	(sign diff.)	EAE-vs-CFA	EAE-vs-N	2.1304	4.5384	1.0000	1.0000
Ccl6	4	(sign diff.)	0.0304	EAE-vs-N	EAE-vs-CFA + N	2.6485	4.4385	1.0000	1.0000
Ccl3	4	(sign diff.)	0.2175	EAE-vs-N	EAE-vs-CFA	3.5207	4.2884	1.0000	1.0000
Itgb2	4	0.1990	0.1960	EAE-vs-CFA + N	EAE-vs-CFA + N	3.4452	4.2665	1.0000	1.0000
Ccl17	4	(sign diff.)	(sign diff.)	EAE-vs-N	EAE-vs-N	2.1608	4.2360	1.0000	1.0000
Ccl1	4	0.0814	(sign diff.)	EAE-vs-CFA + N	EAE-vs-N	3.2892	4.0736	0.6832	0.9573
Cxcl3	4	(sign diff.)	0.1646	EAE-vs-N	EAE-vs-CFA + N	3.8766	3.9024	0.9094	0.9889
Ccr1	2	0.6766	0.1279	EAE-vs-CFA	EAE-vs-CFA + N	2.1241	3.5971	0.9612	0.9938
Tlr4	4	0.5656	0.0523	EAE-vs-CFA	EAE-vs-CFA + N	1.3995	3.5929	0.8130	0.9907
Tlr2	4	(sign diff.)	0.2283	EAE-vs-N	EAE-vs-CFA	2.8127	3.4271	1.0000	1.0000
Ccl19	4	0.4762	0.1445	EAE-vs-CFA	EAE-vs-CFA + N	1.7356	3.3981	0.9831	1.0000
Ccl4	2	(sign diff.)	0.3063	EAE-vs-N	EAE-vs-CFA	0.9184	3.2643	0.8029	1.0000
Cxcl2	2	1.0000	0.2244	EAE-vs-CFA	EAE-vs-CFA + N	0.9034	3.0380	0.7117	0.9976
Ccr6	2	(sign diff.)	0.2585	EAE-vs-N	EAE-vs-CFA	0.4080	3.0184	0.4279	1.0000
Itgam	4	0.2665	0.1242	EAE-vs-CFA	EAE-vs-CFA + N	2.0597	2.9343	1.0000	1.0000
Tgfb1	4	0.2745	0.0614	EAE-vs-CFA	EAE-vs-CFA + N	2.0653	2.8405	1.0000	1.0000
Ccr5	4	(sign diff.)	0.1563	EAE-vs-N	EAE-vs-CFA + N	1.9780	2.7427	0.9989	1.0000
Cx3cr1	4	(sign diff.)	0.1662	EAE-vs-N	EAE-vs-CFA + N	1.8243	2.2697	1.0000	1.0000
Cmtm3	4	0.4137	0.1880	EAE-vs-CFA	EAE-vs-CFA + N	1.0686	2.2565	0.9918	1.0000
Cmklr1	4	0.1120	(sign diff.)	EAE-vs-CFA + N	EAE-vs-N	2.1843	2.1540	0.9990	1.0000
Cxcr4	4	(sign diff.)	0.2185	EAE-vs-N	EAE-vs-CFA + N	1.6593	2.1456	0.9989	0.9980
Ccr3	4	(sign diff.)	0.3589	EAE-vs-N	EAE-vs-CFA	1.8240	2.1288	0.9866	0.9883
Ccrl2	4	(sign diff.)	(sign diff.)	EAE-vs-N	EAE-vs-N	1.3218	1.6898	0.8307	0.9130
Il16	2	0.7858	0.1858	EAE-vs-CFA	EAE-vs-CFA + N	1.1177	1.5788	0.9650	0.9954
Cmtm6	4	0.1911	0.2003	EAE-vs-CFA + N	EAE-vs-CFA	1.2183	0.5549	0.9702	0.6036
Gpr17	5	(sign diff.)	0.6473	EAE-vs-N	EAE-vs-CFA	0.8181	0.2678	0.9827	0.0587

Table 3 shows 28 CFA-associated genes (14 for males only, 11 for females only, and 3 in common), specifically those where ΔCFA and ΔEAE are positive and coherent (|ΔCFA| ≾ |ΔEAE|, as defined above), have the same sign, and the |ΔCFA|:|ΔEAE| ratio >1/3. The 11 most significant genes (3 for males only, 5 for females only, and 3 in common) appear in bold. Significant CFA-associated genes were defined as those where the posterior probabilities that ΔEAE > 0.5 Ct and ΔCFA > 2 Ct (both strictly positive) are each at least 2/3 (i.e. minimum 2:1 odds in favor of the CFA effect being large, positive, and contributory to a probably positive EAE effect). A heatmap showing relative effect magnitudes and posterior probabilities of the 11 significant CFA-associated genes is shown in [SUPPLEMENTAL FIGURE 3]. Note that blanks indicate either that the contribution of CFA to the total effect is <1/3, or that ΔCFA or ΔEAE is too small or lacking in precision to be significant for the sex in question. Color intensity in the heatmap is proportional to |Δ CFA|:|ΔEAE|, and bar size is proportional to ΔCFA. Log₂ fold change estimates for the CFA effects are displayed inside the bars. Rows are sorted in descending order of maximum CFA effect size (across sexes).

Table 3.

CFA-associated factor expression by dendrogram cluster. Significant genes for each sex are in bold. Rows in each panel are sorted in descending order of CFA contribution to the total EAE effect (‘ΔCFA/ΔEAE’). Of the 11 highlighted genes, 3 are important for males only, 5 are important for females only, and 3 are important for both sexes.

FEMALES					MALES
Gene	Cluster	ΔCFA	ΔEAE	ΔCFA/ΔEAE	Gene	Cluster	ΔCFA	ΔEAE	ΔCFA/ΔEAE
Ppbp	2	1.9803	1.3757	1.0000	Il4	3	2.3332	2.0450	1.0000
Il6	2	3.7853	3.8589	0.9809	Ccr10	5	1.5695	1.3511	1.0000
Mapk14	6	0.7533	0.8281	0.9097	Pf4	6	1.2759	0.9503	1.0000
Fpr1	2	1.2005	1.5031	0.7987	Cxcr5	3	1.1197	0.5886	1.0000
Il16	2	4.1000	5.2177	0.7858	Ackr1	5	0.5981	0.8056	0.7424
Cxcr6	4	5.0300	6.5716	0.7654	Cxcr6	4	4.4590	6.0358	0.7388
Ccr1	2	4.4435	6.5675	0.6766	Cxcl12	5	0.6580	0.9070	0.7254
Ccl7	2	4.0642	6.1946	0.6561	Ccr9	3	2.8061	4.0231	0.6975
Il1b	1	4.9221	7.6033	0.6474	Ccr4	3	1.8647	2.6814	0.6954
C5ar1	1	4.5095	7.6114	0.5925	Gpr17	5	0.4914	0.7592	0.6473
Tlr4	4	1.8218	3.2213	0.5656	Hif1a	5	0.6226	1.0036	0.6204
Ccl19	4	1.5780	3.3136	0.4762	Xcl1	2	3.3284	6.2824	0.5298
Cmtm3	4	0.7542	1.8228	0.4137	Il6	2	4.4383	8.4131	0.5275
Ccl2	2	3.5746	9.2624	0.3859	Cxcl13	2	2.7322	5.7118	0.4783
					Ccr2	2	2.6134	5.7709	0.4529
					Ccr3	4	1.1918	3.3206	0.3589
					Il1b	1	3.2801	9.4831	0.3459

To summarize, the selection criteria detailed above led to the identification of 40 significant EAE effect genes and 11 significant CFA effect genes. The distribution and overlap of these effects across sex categories are shown in a Venn diagram (Fig. 6), color-coded by category of effect: {CFA vs. EAE} × {male only vs. female only vs. shared}. Genes appearing in each category are listed below. The top panel shows the divisions of counts and totals along each axis, and the bottom panel provides annotation of gene names for each cell of the diagram, organized by color. (Note that a small number of genes appear twice, in separate cells of the Venn diagram, once as a CFA effect for one sex and once as an EAE effect for the other sex. An example is C5ar1, which has a significant effect in EAE males but not EAE females, and a significant effect in CFA females but not CFA males.) As expected, the significant effects identified in our study included most of the known factors implicated previously in EAE induction but here stratified by sex, namely Il1b, Il6, and Ccr2 (M only) for CFA, and Tnf, Tgfb1, Cxcr3, Ccl2 (M only; F important for CFA), and Ccr6 (M only) for EAE, as well as many other highly expressed factors. For validation, increased expression of 23/24 different factors was confirmed by individual RT-PCRs, and the single gene (Ccr1) that failed to achieve statistical significance nevertheless was concordantly increased in EAE relative to Naïve, as expected [SUPPLEMENTAL FIGURE 4].

Fig. 6.

Venn diagram of significant CFA and EAE effects by sex shows large overlap in genes important for EAE, but little overlap in genes important for CFA and many sex-specific effects. The diagram divides the significant chemokines along two axes: sex and whether the chemokine is primarily important for CFA or for EAE. The top panel shows the divisions of counts and totals along each axis, and the bottom panel provides annotation of chemokine names for each cell of the diagram. A few chemokines appear in two cells because their role with respect to CFA and EAE appears to differ by sex (e.g. C5ar1, which is important for CFA in females but only for EAE in males).

It is also of interest to consider how the significant factors were distributed across the clusters of genes we identified as the first stage of the analysis of the chemokine array (Fig. 4). An alternative view of Fig. 4 [SUPPLEMENTAL FIGURE 5] shows the genes appearing in each cluster, but identified according to the color scheme used in Fig. 6 (green = CFA M, yellow = CFA common, orange = CFA F; blue = EAE M, purple = EAE common, red = EAE F; nonsignificant genes are colored gray). This annotation shows quite strikingly that clusters 1 and 4 are wholly significant and mostly comprised of chemokines where the effects are similar across sexes. All of the shared EAE effects (purple) and two of the three shared CFA effects (yellow) appear in clusters 1 and 4, and the shared effects (of either type) dominate both lists. Sex-specific effects (green/orange for CFA, blue/red for EAE) also occur in clusters 1 and 4 but predominantly appear in cluster 2, which has almost no shared effects (only Il6 shared for CFA). Finally, clusters 3, 5, and 6 are almost wholly nonsignificant (only Gpr17 is an exception for females for EAE, but with a rather small effect size). An obvious implication of this tight agreement between the trajectory-based clusters and contrast-based lists of significant effects is that a rigorous visual analysis of the shape, pitch, and clarity of dose–response curves can be an extremely informative initial step when screening large numbers of potential factors of unknown sensitivity to disease or intervention, allowing one to aggressively filter unpromising response shapes and narrow the focus to only the most likely players before conducting any formal statistical hypothesis testing. We recommend this approach as one that we feel may mitigate the risks of false positive and false negative findings that plague traditional statistical cutoff-based analyses of moderate-to-large-scale screening experiments.

Finally, as a metric for assessing total impact of how the chemokine mixture in the system adjusts to EAE induction, we determined the total expression score for all 74 factors included above for each experimental group by sex. Fig. 7A shows the total expression trajectory across treatments for males and females, normalized to the Naïve level in each sex (i.e. showing changes as multiples of the Naïve level). We see that total expression increased more in males than females in the CFA groups (~1.8 vs. ~1.4), with increases accelerating further (to ~4.1 vs. ~2.7) in both EAE groups — a pattern generally consistent with that of most of the factors. Thus, although males appear to react about 50% more to CFA than females, the observed increase from CFA to EAE was proportionally similar for both sexes; that is, the ratios 4.1/1.8 and 2.7/1.4 are both approximately 2, suggesting a roughly 200% change in total expression for EAE over CFA in both sexes. Generally all chemokines increase expression under both CFA and EAE — on average (across all genes) by ~0.9 Ct (roughly 85%) for CFA (z = 8.63) and by ~2.2 Ct (roughly 350%) for EAE (z = 9.00). Fig. 7B presents a cross-sectional view of the total expression for each sex by condition, showing that females have ~25% greater total Naïve expression (z = 3.61), but that under CFA and EAE the males catch up and eventually surpass the females by ~17% (z = 2.18). Note that the curves in Fig. 7A were derived from a nonparametric weighted kernel regression used to average the response curves across genes (afterwards scaling the curves within sex), whereas the bars in Fig. 7B simply sum the total expression level estimates (without accounting for measurement precision). Thus, the top panel presents a more accurate view of the within-sex trajectory for each sex separately (but cannot be used to compare total abundance amounts between sexes), and the bottom panel presents a more accurate cross-sectional view of the relative total abundances between sexes in each condition (but is not appropriate for tracing across conditions).

Fig. 7.

Total chemokine expression profiles reveal sex differences. A) The top panel shows total mean chemokine expression (i.e. posterior means summed over all chemokines) for each sex at each condition, represented as a multiple of the total Naïve level. All chemokines tend to increase expression under both CFA and EAE, but males appear to experience more accelerated increases than females, with approximately a 50% steeper rise in slope (z = 2.22). B) However, the bottom panel indicates that females have ~25% greater total baseline expression (z = 3.61). Under CFA and EAE conditions the males match up and eventually surpass the females by ~17% (z = 2.18). Note that the curves in the top panel were derived from a nonparametric weighted kernel regression used to average the response curves across chemokines, whereas the bars in the bottom panel simply sum the level estimates (without accounting for measurement precision); thus, the top panel presents a more accurate view of the trajectory, and the bottom panel presents a more accurate cross-sectional view of the relative abundances between sexes.

Using this more general metric of total chemokine expression, it was then possible to determine the contribution of each individual factor as a percentage of the total expression of all factors. A panel of plots showing these profiles for males and females is presented in [SUPPLEMENTAL FIGURE 6]. (Note that the actual expression volumes accounted for by the highest- and lowest-expression factors span many orders of magnitude, so the vertical spacing and color intensities in these plots are logit-scaled, i.e. transformed by a logistic function, in order to facilitate visualization. Each vertical tick in the trajectory plot represents a concentration 10-fold smaller than the tick above it, and the color intensity ramps in the heatmap plots behave similarly. In the heatmap panels for each sex, the factors are sorted in descending order of Naïve rank.) This factor-by-factor analysis revealed that for the most part the sexes looked very similar, and most of the factors had a fairly stable ranking of contribution to total expression (as measured by percentage of total) across conditions. Although as a general pattern high-expression factors tended to decline slightly in importance while low-expression factors tended to rise slightly in importance, both high-expression factors and low-expression factors each tended to remain in their respective positions across all conditions. There were a few possible exceptions, mostly involving females. For example, Ccl26 and Ccl28 were switched off almost entirely for females under CFA (and EAE), while Ccl7, Il16, and Xcl1 went from being essentially switched “off” suddenly to “on” at high levels (approximating those of males in the CFA and EAE conditions). Other factors such as Ccl4 and Cxcl13 for females and Ccl7 for males had more irregular patterns caused perhaps by large variance. However, in both male and female groups >50% of the total expression in Naïve, and >65% of the total expression in CFA and EAE conditions, could be attributed to just 11 very-high-expression factors (Ackr1, Cmtm5, Cx3cr1, Cxcr12, Cxcl16, Gpr17, Hif1a, Itgam, Itgb2, Mapk1, and Mapk14), with slight reductions in most of these factors observed in the EAE groups due likely to large increases in expression of many lower-expression factors.

Nearly all of the most dynamically increased factors (Ccl8, Tnf, Ccr7, etc.) from our study had a relatively small contribution to the total expression, with curves that contributed an extremely low percentage in Naïve mice but increased to clearly measurable values in EAE and some CFA groups. Ccl8 showed ~1000-fold changes in expression going from Naïve to EAE, but because of the low baseline, even a 1000x increase in that factor produced only a modest increase in its fraction of the total. It thus appears that during EAE induction leading to severe inflammation in the CNS, many of the high-expression factors like Ackr1 and Itgam that may normally act as gatekeepers were substantially downregulated, thus allowing highly significant increases in many lower-expression factors that contributed meaningfully to the induction of clinical and histological disease.

4. Conclusion

The results of this study provide new perspectives on two major issues regarding the underlying processes involved in the induction of EAE in C57BL/6 mice. First, regarding sex differences, the clinical course and severity of EAE were nearly identical in male and female mice despite increased spinal cord lesions and demyelination potentially enabled by fewer immunoregulatory mechanisms in females vs. males. Understanding such sex differences may allow further evaluation of which components are hormonally or chromosomally regulated and could aid in designing better therapies for both sexes. Second, since the CFA/Ptx components of adjuvant are a requirement for EAE induction in C57BL/6 mice when injected with the mMOG-35–55 peptide, determining the degree of EAE enhancement attributable to prominent adjuvant-induced factors was found to segregate key innate inflammatory factors and otherwise could be valuable for identifying infectious or chemical processes that under the right conditions could trigger onset of human autoimmune diseases.

Evaluation of the splenic cell types and subsets shown in Table 1 revealed higher total numbers in Naïve female vs. Naïve male mice but increased numbers after immunization with CFA/Ptx. However, in EAE mice the numbers were less than in the CFA mice, likely due to selective migration of cells from spleen to CNS during disease induction. Many subset percentages of both pro-inflammatory and regulatory T and B cells in male EAE mice were actually lower than in CFA males, indicating T and B cell chemotaxis, whereas those percentages were essentially the same in females. Increased cell migration from the spleen to the CNS may have been facilitated in both males and females by the strong upregulation of CD74 and modest increases in its signaling co-partner, CD44, in conjunction with continued baseline expression levels of MIF but not D-DT (lower expression levels), as shown in Fig. 3. These data extend our previous observation that MIF and D-DT may play a role in augmenting EAE severity [36]. Males with EAE also had more macrophages/monocytes than females with EAE, with variable percentages of subsets. In total, these data support the conclusion that males had stronger pro-inflammatory responses than females, but these responses appeared to be tempered by increased anti-inflammatory and regulatory immune cell subtypes resulting in less CNS inflammation and demyelination but EAE scores the same as in females.

Analysis of chemokines, cytokines and receptors from spinal cords implicated a total of 40 factors that were significantly altered during the EAE induction process, 13 in males only, 4 in females only and 23 in common (see Fig. 6). Linked analysis of responses from Naïve, CFA, and EAE mouse groups revealed a subgroup of 11 CFA/Ptx-induced genes that contributed significantly to induction of clinical EAE signs, as well as 34 others that were increased in EAE but were relatively unaffected by the adjuvant effect (Fig. 6). To further segregate the most robust factors contributing to EAE induction, we matched inflammatory chemokines with their respective receptors (Table 4), and for both males and females identified factor/receptor combinations with log₂ fold change estimates summing to >10, or individual factors with log₂ fold change estimates >5 (when there were multiple receptors for a single factor, the sum of all receptors for the factor was used). This analysis revealed 11 pathways with ligand/receptor scores >10, four that were female-specific (Ccl2/Ccr2,4; Ccl9/Cxcr3; Ccl22/Ccr4 and Xcl1/Xcr1), five that were male-specific (Ccl3/Ccr1,5, Ccl4/Ccr5,8, Ccl12/Ccr2, Cxcr4; Ccl19/Ccr7,Ccrl2 and Cxcl10/Cxcr3), and two that were common to both sexes (Ccl5/Ccr1,3,4,5 and Ccl8/Ccr1,2,3,5). Additionally, there were two individual factors >5 for females (Ccr7 and Ccr8) and five for males (Ccl2, Ccl7, C5ar1, Il1b, and Tnf).

Table 4.

EAE effects of chemokines and receptors.

Factor/Receptor	Female Effect	Male Effect	Target Cell Type
CCL1/CCR8	3.3/5.6	4.1/4.7	B and T activation, NK, DC chemotaxis, M, MG activation, MG chemotaxis, demyelination
CCL2/CCR2; 4	5.7*/1.3; 3.8 = 5.1	5.7/3.2*; 0.8 = 4.0	M chemoattractant, A, MG activation, N, NSC/NPC chemotaxis and differentiation, T, DC chemotaxis and activation
CCL3/CCR1; 5	3.5/2.1*; 2.0= 4.1	4.3/3.6; 2.7 = 6.3	Inflammatory M, MG, A, N, NSC/NPC microglial chemotaxis, NP chemotaxis and activation
CCL4/CCR5; 8	0.9/2.0; 5.6 = 7.6	3.3/2.7; 4.7 = 7.4	Inflammatory M, MG and T subset chemotaxis, T subset adhesion, T, NK chemotaxis
CCL5/CCR1; 3; 4; 5	4.8/2.1*; 1.8; 3.8; 2.0 = 9.7	6.5/3.6; 2.1; 0.8. 2.7 = 9.2	Adaptive immunity, T expression and secretion, MG chemotaxis, BA, E chemotaxis and activation
CCL6/CCR1	2.6/2.1*	4.4/3.6	M chemoattractant from M, MG, T recruitment, A migration, NP chemotaxis
CCL7/CCR2; 3	2.1*/1.3; 1.8 = 3.1	4.5/3.2*; 2.1 = 5.1	Priming T, M migration across BBB, T chemotaxis, A, ↑ by TNFα
CCL8/CCR1: 2; 3; 5	8.3/2.1*; 1.3; 1.8; 2.0 = 7.2	10/3.6; 3.2*; 2.1; 2.7 = 11.6	M chemoattractant, Mast cells, BA, E, T, NK chemotaxis and activation
CCL9/CXCR3	5.9/5.8	3.7/6.1	M inflammation, DC chemotaxis
CCL12/CCR2; CXCR4	4.3/1.3; 1.7 = 3.0	4.8/3.2; 2.1 = 5.3	M chemoattractant, E, T chemotaxis, stem cell recruitment
CCL17/CCR4	2.2/3.8	4.2/0.8	T chemotaxis, HSC chemoattractant
CCL19/CCR7; CCRL2	1.7/6.3: 1.3 = 7.6	3.4/6.1; 1.7 = 7.8	M inflammation, T, B, DC, infiltrating L, A, MG
CCL22/CCR4	6.2/3.8	4.5/0.8	M, DC, NK chemotaxis, infiltrating L, A, MG
CX3CL1/CX3CR1	−0.2/1.8	−1.3/2.2	MG, M recruitment, Adhesion molecule
CXCL2/(CXCR2)	0.9/NT	3.0/NT	PMN, HSC chemoattractant, M, MG, NP inflammation, Microvascular EC
CXCL3/(CXCR2)	3.9/NT	3.9/NT	M, MG, EC, N Precursors, NP inflammation
CXCL10/CXCR3	3.5/5.8	5.9/6.1	M, B, T, NK, NKT, L recruitment, HSC, filial signaling
CXCL13/CXCR5	1.0/0.1	3.0*/−0.5	B, T, A activation, M, Follicular DC,
CXCL16/CXCR6	3.3/1.5*	4.9/1.6*	B, T, NKT, DC, ↑ by IFN, TNF
XCL1/XCR1	6.2/4.7	3.0*/1.5	T, NK, NKT, CT DC chemoattractant
Itgb2/Itgam	3.4/2.1	4.3/2.9	MG, NKT, M and NP adhesion, demyelination
C5aR1	3.1*	5.7	A, DC, M, MG, Infiltrating T
CmklR1	2.2	2.2	DC, M, NK, MG
CmTm3	1.1	2.3	T, M
IL-1β	2.7*	6.2*	M, MG, NP, Inflammatory M, Infiltrating T
IL-6	0.1*	4.0*	A, M, B, T activation, T and CT differentiation
TGFβ1	2.1	2.8	T, CT, B, M, A, DC
TLR2	2.8	3.4	M, B, T, DC, MG,
TLR4	1.4	3.6	Priming T, M, B, CT, A, MG
TNFα	4.7	7.4	M activation, T, NK, NP, E, N, OGD, Mast cells

Bold: ΔCT score (Log2 units): Single factor >5; Factor + Receptor(s) >10.

^*Also contributed to CFA effect. NT = Not Tested Peripheral and CNS cell types: A (astrocyte); B (B cell); BA (basophil); E (eosinophil); CT (cytotoxic T cell); DC (dendritic cell); EC (endothelial cell); HSC (hematopoietic stem cell); M (monocyte/macrophage); L (Leukocyte); MG (microglia); N (neuron); NP (neutrophil), NSC/NPC (neural stem/progenitor cell); NKT (natural killer T cell); OGD (oligodendrocyte); PMN (neutrophil); T (T cell).

The results of our chemokine array were largely validated when compared to a recent review listing previously published factors deemed to be critical for EAE induction [24]. These included IL-6, IL-1β, IL-23, and TGFβ needed to differentiate autoreactive Th17 cells to an encephalitogenic CCR2⁺, CCR6⁺ phenotype. Upon local reactivation by monocytes and dendritic cells within the CNS, the CCR2⁺ Th17 cells and monocytes release IL-17, GM-CSF, TNFα, and CCL2 that activate microglial cells and further recruit other CCR2, CCR6, and CXCR2 expressing leukocytes across a CCR2⁺ vascular endothelial cell barrier into the CNS where they release a variety of inflammatory factors that cause demyelination, axonal damage, and clinical signs of EAE. Although our array did not include IL-17, IL-23, or GM-CSF, it did clearly implicate IL-6, IL-1β, and a third factor, CXCR6 (CD186, an upstream marker on DC and NK cells), as adjuvant-induced contributors to EAE in both males and females. Our data also implicated the critical-for-EAE CCR2 axis involving CCL2, CCL7, and CCL8 (monocyte chemoattractant proteins — MCP-1, 3, and 2 respectively) demonstrating both adjuvant-assisted and non-adjuvant-associated effects. Of particular interest is CCL8, our highest-ranked factor not yet implicated in EAE that activates mast cells that release vasoactive amines, the key EAE-enhancing components induced by Ptx. [51] Surprisingly, there were strong adjuvant-induced increases in IL-4 expression in males that might limit CNS inflammation, as is noted above, and C5ar1 in females that implicates the complement cascade in EAE induction. Additionally, TNFα and TGFβ1 were strongly upregulated non-adjuvant-induced EAE-associated genes in the CNS of both male and female mice. Of the previously implicated factors evaluated in our current study, only CXCR2 was not implicated in EAE induction, due possibly to strain differences between C57BL/6 mice used in our study and SJL/J mice in which CXCR2 was required for neutrophil recruitment to develop EAE [52]. CXCR2 was barely amplified in our RT-PCR assays, apparently suffering critically low expression under all conditions in both sexes, and because of the large uncertainty in the measured values we did not venture to interpret it in this study.

In addition to CCL2 mentioned above, a number of other chemokines and receptors have been shown to be increased in the CNS during the course of acute EAE [24], including CCL5 (showed a correlation with increased clinical signs in C57BL/6 mice) [53]; macrophage CXCR3 [54]; macrophage CCR4 [55,56]; and macrophage CXCR7 [57]. Parenthetically, CCL3, CXCL10, and CXCR2 (as mentioned above) were found to be more important in the pathogenesis of acute EAE in SJL/J mice, and CCL2 localized on astrocytes was associated with EAE relapses in (SJL × SWR) F1 mice [58]. The critical CCR2-CCL2 (MCP1) axis also involves other CCR2-family ligands strongly upregulated in our screening (Fig. 5), including CCL7 (MCP3) that guides Th17 cells to lymph nodes rather than the CNS, and CCL8 (MCP2), our top candidate in both males and females. It is noteworthy that expression of CCR2 on myeloid cells in the CNS appears to be more important in EAE progression than expression on infiltrating encephalitogenic Th17 cells [59,60].

A particular strength of our approach is the rigor with which we analyzed and parsed the effects of the CFA and EAE treatments on the chemokines and receptors that we measured. First undertaking a thorough exploratory descriptive analysis of the response curves, we identified several distinct visual shape patterns that we characterized using a set of novel summary indices and then categorized using hierarchical clustering analysis. For each type of curve shape we defined appropriate contrasts of curve points to describe the CFA and EAE effects, and quantified these contrasts using a robust Bayesian regression modeling strategy that properly accounted for censoring and uncertainty in the RT-PCR measurements. This approach yielded a set of highly relevant and interpretable metrics that we used to characterize, compare, and group the contributions of each factor to EAE induction. The high degree of consistency we see in our results, both internally in terms of chemokine-receptor co-regulation and externally in terms of factor identifications that have been implicated in other studies, speaks strongly for the merits of sophisticated methodological approaches such as ours over more simplistic conventional analyses (based on p-value sorting) that tend to capitalize on small-sample artifacts and thus carry a high risk of false positive findings.

In summary, our study provides clinical, histological, cellular, and molecular evidence demonstrating underlying mechanistic similarities and differences in components leading to essentially identical clinical EAE severity in male vs. female C57BL/6 mice. Moreover, specific factors have been identified in the CFA group that were significantly enhanced during the induction of EAE along with many non-adjuvant-associated factors that may now be evaluated specifically for targeted reduction by potential MS therapies.

Abbreviations:

7AAD

7-amino-actinomycin D

C57BL/6, C57BL/6

wild type mouse strain

cDNA

Complementary deoxynucleic acid

CDI

Cumulative Disease Index

CNS

Central Nervous System

ELISA

Enzyme linked immunosorbent assay

EAE

Experimental autoimmune encephalomyelitis

FACS

Fluorescence-activated cell sorter

CFA

Complete Freund’s adjuvant

IFNγ

Interferon gamma

i.p.

Intraperitoneal

LN

Lymph nodes

MHC

Major histocompatibility complex

MS

Multiple sclerosis

Mtb

Mycobacterium tuberculosis

MOG

Myelin oligodendrocyte glycoprotein 35–55

PBS

Phosphate-buffered saline

PFA

Paraformaldehyde

Ptx

Pertussis toxin

PD-1

Program death receptor 1

PD-L1

Programmed death ligand 1

PD-L1 and PD-L2

Programmed death ligand 1 and 2

RNA

Ribonucleic acid

SC

Spinal Cord

SEM

Standard error of the mean

TNF

Tumor necrosis factor

VA

Veterans’ Affairs

WT

Wild type

7. Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.