T cells from NOD-PerIg mice target both pancreatic and neuronal tissue

Abstract

It has become increasingly appreciated that autoimmune responses against neuronal components play an important role in type 1 diabetes (T1D⁴) pathogenesis. In fact, a large proportion of islet-infiltrating B-lymphocytes in the NOD mouse model of T1D produce antibodies directed against the neuronal type III intermediate filament protein, peripherin. NOD-PerIg mice are a previously developed BCR-transgenic model in which virtually all B-lymphocytes express the heavy and light chain immunoglobulin molecules from the intra-islet derived anti-peripherin reactive hybridoma H280. NOD-PerIg mice have accelerated T1D development and PerIg B-lymphocytes actively proliferate within islets and expand cognitively interactive pathogenic T-cells from a pool of naïve precursors. We now report adoptively transferred T-cells or whole splenocytes from NOD-PerIg mice expectedly induce T1D in NOD.scid recipients, but depending on the kinetics of disease development, can also elicit a peripheral neuritis (with secondary myositis). This neuritis was predominantly composed of CD4⁺ and CD8⁺ T-cells. Antibody depletion studies showed neuritis still developed in the absence of NOD-PerIg CD8⁺ T-cells, but required CD4⁺ T-cells. Surprisingly, sciatic nerve-infiltrating CD4⁺ cells had an expansion of IFN-γ⁻ and TNF-α⁻ double negative cells compared to those within both islets and spleen. Nerve and islet infiltrating CD4⁺ T-cells also differed by expression patterns of CD95, PD-1 and Tim-3. Further studies found transitory early B-lymphocyte depletion delayed T1D onset in a portion of NOD-PerIg mice allowing them to survive long enough to develop neuritis outside of the transfer setting. Together, this study presents a new model of peripherin-reactive B-lymphocyte dependent autoimmune neuritis.

Introduction

The NOD mouse has greatly expanded our knowledge of the genetics and pathological underpinnings of autoimmune mediated type 1 diabetes (T1D). However, NOD mice are also prone to the development of other autoimmune disorders such as spontaneous salivary and lacrimal gland lesions reminiscent of Sjögren syndrome, spontaneous thyroiditis and parathyroiditis, and aging-associated neuritis/meningitis (1, 2). NOD mice are also susceptible to inducible diseases with clinical similarities to Hashimoto’s thyroiditis, systemic lupus erythematosus, myositis, colitis, and allergic encephalomyelitis (1, 2). Several genetic or pharmacological manipulations of NOD mice have also triggered the development of autoimmune neuritis. B7–2 deficient NOD mice spontaneously developed a peripheral neuritis resulting in peripheral nerve demyelination and hind limb paralysis (3, 4). In this model, CD4⁺ T-cells and IFN-γ were indispensable for disease development (3, 4), with the majority of pathogenic effectors specifically recognizing myelin protein 0 (5). ICAM-1 deficient NOD mice also develop a demyelinating form of neuritis mediated by CD4⁺ T-cells, and although IFN-γ was increased in infiltrating T-cells, IL-17 production determined disease severity (6). NOD-H2^b/b Pdcd1^−/− mice and, to a lesser extent NOD-H2^b/g7 Pdcd1^−/−mice, which are both PD-1 deficient, also develop spontaneous CD4⁺ T-cell mediated neuritis (7). Despite the development of anti-myelin autoantibodies, this model does not require B-lymphocyte help since genetic introduction of an Igμ^−/− mutation did not abrogate neuritis development. Finally, IL-2 blockade in NOD mice, in addition to exacerbating T1D development, triggers the formation of multi-system autoimmunity, including peripheral neuropathy mediated both by CD4⁺ and to a lesser extent, CD4⁻ T-cells (8).

Autoreactive T- and B-lymphocytes recognizing nervous system components are contributors to the natural course of T1D development in NOD mice. The well-known BDC2.5 CD4⁺ T-cell clone originally isolated from the spleen of a diabetic NOD mouse (9) was eventually determined to respond to the neuronal as well as ß-cell expressed protein Chromogranin A (10), and thereafter to modified or hybrid versions thereof (11–13). Antibodies against glutamic acid decarboxylase 65 (14) expressed in both neuronal and ß-cells have long been known as a marker for T1D risk in humans. Autoimmune responses mediating peri-islet neuronal Schwann cell death has been reported to precede β-cell destruction in NOD mice (15, 16). Interestingly, a large number of hybridomas developed from islet-infiltrating B-lymphocytes in NOD mice produce antibodies against neuronal elements (17). The target antigen was later identified as peripherin (18). Antibodies directed against phosphorylated peripherin have also been detected in sera from a majority of tested T1D patients (19).

Peripherin is expressed during early development of β-cells although it is undetectable in adult pancreata (20). Accordingly, in adult NOD mice, anti-peripherin antibodies stained neuronal components and not β-cells (17). However, it has been hypothesized that early pancreatic immune infiltration during the timeframe in which islet expression of neuronal components occurs (21, 22) may cause cytokine mediated increases in peripherin expression (23, 24). This could trigger an autoimmune response against peripherin during the early stages of T1D development (18). As a resource to test these possibilities, we created a NOD mouse stock (25) transgenically expressing the heavy and/or light chain immunoglobulin (Ig) molecules from the peripherin autoreactive hybridoma clone H280 (18, 26). Carriers of the H280 heavy (NOD-PerH) or light chain (NOD-PerL) alone both had accelerated T1D onset compared to standard NOD controls, and disease development was even more rapid in mice expressing both transgenes (NOD-PerIg) (25). NOD-PerIg B-lymphocytes actively proliferate within the pancreatic islets (25). T-cells from NOD.IgHEL.Igμ^null mice (in which all B-lymphocytes recognize the diabetes irrelevant antigen hen egg lysozyme and thus cannot participate in the prior activation of diabetogenic T-cells) transferred T1D more rapidly to NOD.scid-PerIg than standard NOD.scid recipients (25). This further indicated a role of peripherin autoreactive B-lymphocytes in expanding diabetogenic T-cell responses. We initially aimed to learn more about the peripherin-reactive T-cells that are expanded in NOD-PerIg mice. Surprisingly, we discovered that in addition to displaying diabetogenic activity, T-cells from young NOD-PerIg but not standard NOD mice, can transfer an autoimmune neuritis (with secondary myositis) to NOD.scid recipients. This study assessed the pathogenic basis of such neuritis development.

Materials and Methods

Mice

NOD/ShiLtDvs (27) (hereafter NOD) and all other described mouse strains are maintained at The Jackson Laboratory under specific pathogen-free conditions. The NOD.Cg-Emv30^b Prkdc^scid/Dvs (hereafter NOD.scid) strain has been previously described (28). NOD/ShiLtDvs-Tg(IghH280)48Dvs Tg(IgkH280)934Dvs/Dvs (hereafter NOD-PerIg) and NOD.Cg-Emv30^b Prkdc^scid Tg(IghH280)48Dvs Tg(IgkH280)934Dvs/Dvs (hereafter NOD.scid-PerIg) mice transgenically expressing the heavy and light chains from peripherin reactive hybridoma clone H280 (18, 26) have been previously described (25). Female mice were utilized for all experiments (donors and recipients). All mouse work has been approved by The Jackson Laboratory’s Animal Care and Use Committee.

Cellular Enrichment

Total T-cell enrichment was accomplished via negative selection over LD columns (Miltenyi Biotec; Bergisch Gladbach, Germany). Biotinylated B220, CD11b, and CD11c specific antibodies (Tonbo Biosciences; San Diego, CA) were used in conjunction with streptavidin MACS beads (Miltenyi Biotec) to negatively select away unwanted splenocyte populations. For T-cell co-mixtures, additional biotinylated CD4 or CD8 specific antibodies (BD Biosciences; San Jose, CA) were used to negatively select away the T-cell subpopulation of choice.

In vivo cellular depletion

CD8⁺ T-cell depletion was accomplished by injecting 250μg anti-CD8 (53–6.72 originally from ATCC, expanded and purified in house) i.p. every three weeks. CD4⁺ T-cell depletion was accomplished by injecting 250μg anti-CD4 (GK1.5, BE0003–1; BioXCell; West Lebanon, NH) i.p. weekly for the 1^st four injections, followed by injections every two weeks for the remainder of the study. Anti-CD8 and anti-CD4 injections were initiated in NOD.scid recipients on the same day they were engrafted with splenocytes. B-lymphocyte depletion was accomplished by injecting i.p. 250μg of anti-CD20 (MB20–11, IgG2C; Medimmune; Gaithersburg, MD) every two weeks.

Monitoring T1D development

Mice were checked weekly for presence of glucosuria using Ames Diastix (Bayer; Leverkusen, Germany). Two readings of ≥0.25% (corresponding to ≥300mg/dl in blood) on separate days was defined as T1D.

Flow Cytometry

For all flow cytometry experiments, single cell suspensions were prepared and data collected on either a FACSymphony A5, a BD FACSCalibur, or an LSR II SORP (All instruments: BD Biosciences). Data was analyzed using FlowJo Version 9/10 (BD Biosciences). Gey’s buffer was used to lyse red blood cells for all spleen samples (29). For sciatic nerve preparations, nerves were removed from both hind limbs and placed in Dulbecco’s PBS w/calcium & magnesium (ThermoFisher Scientific; Waltham, MA). Nerves were transferred into 1mL digestion mixture containing Hank’s Balanced Salt Solution (MilliporeSigma; St. Louis, MO), 5U dispase (Collaborative Research Inc.; Waltham, MA) and 400U collagenase D (MilliporeSigma). Nerves were mechanically disrupted, incubated for 30 minutes at 37°C, and then triturated up and down. Digestion was inhibited with Hank’s Balanced Salt Solution + 2% fetal bovine serum. The cell mixture was passed through 70μm nytex to remove any remaining clumps. Hand picking of dissociated islets via a collagenase/DNase I digestion procedure has been previously described (30). The following modifications were performed: after the second pick into fresh media, islets were mechanically disrupted by vigorous pipetting.

For splenic engraftment experiments, samples were run on a BD FACSCalibur. No singlet discrimination was performed. Therefore, initial FSC-H vs Propidium Iodide (PI) gating was used followed by gating on either CD4 or CD8 vs TCRβ amongst live cells, or CD19 amongst live cells followed by IgM^a or IgM^b gating. For sciatic nerve cellular subsets analysis assessed on a FACSymphony A5 or LSR II SORP, initial “rough leukocyte” FSC-A vs SSC-A gating was performed to exclude debris and small events with large SSC. This was followed by singlet discrimination performed via both FSC-A vs FSC-H and SSC-A vs SSC-H, gating on the diagonal singlet cell populations. PI gating was performed to identify live cells, followed by gating on CD45.1 (see Supplementary Figure 1). For cytokine and cell surface marker experiments assessed on a FACSymphony A5, initial gating started with the two singlet gates as described in Supplemental Figure 1, followed by gating tightly on either Ghost Dye UV450⁻ (for cytokines; Tonbo Biosciences) or PI⁻ (for surface markers) vs CD45.1⁺ cells. T-cell subset gating followed as in Figure 3.

Figure 3 – Nerve damage and immune cell infiltration.

Whole splenocytes (normalized to contain 1×10⁶ T-cells) from 4–6-week-old NOD-PerIg mice were injected i.v. into NOD.scid recipients. (A) Mice were monitored weekly for T1D and/or neuritis development out to 21 weeks post-transfer. At study initiation 23 recipients of NOD-PerIg splenocytes were subsequently monitored for T1D/neuritis development. Any recipients developing T1D were removed from the study. Controls consisted of 8 NOD.scid mice. Mice from this cohort, in addition to surviving mice from the behavioral studies were analyzed for nerve damage (B, C) and immune cell infiltration (D-J). Mice from this cohort that developed T1D prior to 15 weeks post transfer were only subjected to immune cell infiltration analyses. (B) Quantification of hip compound muscle action potential (CMAP) comparing age-matched controls (unmanipulated or PBS-injected controls from behavioral cohort) to recipients of NOD-PerIg splenocytes. Mice with no detectable electromyography amplitudes have CMAP values of 0mV. (C) 40x images of sciatic nerves comparing age-matched controls (left) to recipients of NOD-PerIg splenocytes with clinical symptoms of neuritis (center) or no-disease symptoms (right). (D) 100x image of a sciatic nerve of a mouse with clinical symptoms of neuritis showing various indicators of pathology. Arrows: demyelinated axons; Squares: myelin ovoids; Circle: onion bulb. (E) Quantification of gated live cells showing percentage of leukocytes (CD45.1⁺) within sciatic nerves. (F and G) Representative flow cytometry plot (F) and quantification (G) showing the percentage of TCRβ⁺ CD90⁺ T-cells amongst CD45.1⁺ cells within sciatic nerves. (H and I) Representative flow cytometry plots (H) and quantification (I) of gated T-cells showing percentage of CD4⁺ vs CD8⁺ T-cells within sciatic nerves. (J and K) Representative flow cytometry plots (J) and quantification (K) of gated leukocytes (CD45.1⁺) showing B220⁺ CD19⁺ B-lymphocytes.

For intracellular cytokine staining, single cell suspensions were plated in RPMI 1640 (ThermoFisher Scientific) supplemented with 1X GlutaMAX (ThermoFisher Scientific), 1mM sodium pyruvate (ThermoFisher Scientific), 1X MEM NEAA (ThermoFisher Scientific), 100 μM 2-mercaptoethanol (Fisher Scientific), 100 U/mL penicillin (MilliporeSigma), 100 μg/mL streptomycin (MilliporeSigma) and 10% (v/v) heat-inactivated HyClone fetal bovine serum (GE Healthcare Life Sciences; Marlborough, MA). Cells were initially stimulated for 5 hours in the presence of 25ng/mL PMA (MilliporeSigma), 1 μg/mL ionomycin (Cayman Chemical; Ann Arbor, MI), and GolgiPlug/Brefeldin A (BD Biosciences) at the manufacturer’s recommended 1/1000 dilution. Extracellular markers were stained as normal. Cells were washed with Dulbecco’s PBS then incubated with Ghost Dye UV450 for live dead discrimination. Cells were then fixed with Cytofix/Cytoperm (BD Biosciences) per the manufacturer’s recommended protocol. Intracellular staining was done in the presence of BD Perm/Wash Buffer (BD Biosciences).

Fluorochrome-conjugated monoclonal antibodies were purchased from the following vendors: BD Biosciences: Ly6g (RB6–8C5; BV421), IgM^b (AF6–78; PE), IgM^a (DS-1; FITC), TCRβ (H57–597; BV711 or FITC), CD8α (53–6.7; BV480 or APC), CD11b (M1/70; BV650), B220 (RA3–6B2; BUV496), CD45.1 (A20; FITC), Ly6c (HK1.4; BV570), IL-4 (11B11; PE), IL-5 (TRFK5; PE), IL-17A (TC11–18H10, BV786), CD154/CD40L (MR1; PE), CD95/Fas (Jo2; FITC), CD69 (H1.2F3; BUV737); Biolegend (San Diego, CA): IgM^a (MA-69; PE), CD90.2 (53.21; PE-Cy7), CD90.2 (30-H12; APC-Cy7), CD4 (GK1.5; BV785), CD4 (RM4–5; BV570), IFN-γ (XMG1.2; APC), TNF-α (MP6-XT22; PE-Cy7), TCRβ (H57–597; PE), CD223/LAG-3 (C9B7W; BV421), CD134/OX-40 (OX-86; PerCP-Cy5.5), CD366/TIM-3 (RMT3–23; PE-Cy7), CD357/GITR (DTA-1; PerCP-Cy5.5), CD278/ICOS (C398.4A; BV605), CD127/IL-7Rα (A7R34; BV650), CD279/PD-1 (29F.1A12; BV711, CD25 (PC61; BV785); Tonbo Biosciences: CD11c (N418; APC), CD19 (ID3; RF710, APC), CD45.1 (A20; APC); ThermoFisher Scientific: CD4 (GK1.5, PE), IL-15 (eBio13A; PE), CD178/FasL (MFL3; Super Bright 600).

Histology & Pathology

Pancreata were fixed in Bouins (Rowley Biochemical; Danvers, MA) overnight and embedded in paraffin blocks. Three levels were cut and stained with H&E and aldehyde fuchsin. For neuritis studies: Cohort 1 – Figure 1A, mice were provided to and processed by The Jackson Laboratory Necropsy Core to determine the cause of the visible hind leg weakness. Subsequently, we focused on analysis of hind limbs for neuritis development. Briefly, hind limbs were fixed in 10% Neutral buffered formalin (StatLab Medical Products; McKinney, TX), then transferred into ImmunoCal (StatLab Medical Products) for 48 hours. Hind limbs provided the following sections: One hind limb provided two femur and tibia cross sections; The other hind limb provided two longitudinal (sagittal) sections. All sections were stained with H&E. Descriptive pathology details were provided by a blinded pathologist (R.D.). In some figures this descriptive pathology was converted to a number scale for purposes of plotting quantitative differences between groups. Specifics are provided in the figure legends. Limb, cranial, and spinal H&E histology images were obtained with an Olympus DP72 microscope digital camera using cellSens Standard 1.5 imaging software (Tokyo, Japan). All non-supplemental H&E histology images were processed in the following manner using GIMP 2.8.16 (The GIMP Development Team; https://www.gimp.org): The images were scaled down to 2.1 inches wide, set at 300 dots per inch (to decrease figure file size), and finally vertically cropped to final dimension of 2.1 × 1.5 inches in height. Cropping was performed to remove the embedded magnification/scale information, which was too small and pixelated to read at final publication size/resolution. Higher resolution scale information was superimposed on top of the cropped images in Inkscape 0.91 (The Inkscape Project; https://inkscape.org/) during figure preparation.

Figure 1 – NOD-PerIg T-cells transfer diabetes as well as neuritis into NOD.scid recipients.

(A) Diabetes incidence from Cohort 1 NOD.scid recipients receiving 1×10⁶ enriched T-cells from 5-week-old NOD or NOD-PerIg donors. Selected mice from Cohort 1 were sent to necropsy for analysis of hind limbs, brain, and spine after hind limb weakness was observed. (B) Representative histology displaying signs of neuritis in a peripheral nerve of the hind limb (Scale bar 200 μm). (C) Representative showing one cranial nerve (Scale bar 500 μm) which showed signs of infiltration. (D) Representative spinal nerve with signs of infiltration (Scale bar 500 μm). In B-D black arrows designate representative area of infiltration in affected tissue. (E) Diabetes incidence in Cohort 2 NOD.scid recipients of 1×10⁶ enriched T-cells from 7-week-old NOD or NOD-PerIg donors. (F) Representative image of hind limb peripheral nerve neuritis in a Cohort 2 recipient of NOD-PerIg T-cells (Scale bar 200 μm). (G) Quantification of mean insulitis scores for non-diabetic mice at 20 weeks post transfer comparing recipients of NOD or NOD-PerIg T-cells.

Nerve histology has been described previously (31). Briefly, sciatic nerves were dissected free and fixed by immersion in 2% paraformaldehyde/2% glutaraldehyde in 0.1mol/L cacodylate buffer. Nerves were then plastic-embedded, sectioned at 0.5μm thickness, and stained with toluidine blue. Images were collected at 40x or 100x magnification on a Nikon Eclipse 600 microscope with DIC-Nomarski optics. Three blinded observers (T.J.H., A.L.D.T., R.W.B.) examined the histology slides to assign the mice to control, clinical symptoms, and no clinical symptom groups based solely on histology.

Insulitis Scoring

Mean insulitis scores (MIS) were determined by a blinded observer (D.V.S.) using a previously described method (32). Briefly, for non-diabetic mice, Aldehyde fuchsin stained islets were scored at end of incidence as follows: no lesions, 0; peri-insular aggregates, 1; <25% islet destruction, 2; >25% islet destruction, 3; >75% islet destruction, 4. The final score was determined by dividing the total score for each pancreas by total islets examined.

Nerve Conduction Velocity

Briefly, mice were anesthetized and placed on a heating pad. Legs were straightened and fixed in place with tape, then the ground recording electrode was placed subcutaneously in the left hind paw. The negative recording electrode was placed between the last two toes of the right hind paw and the positive electrode was placed subcutaneously in the right hind paw. Stimulating electrodes were positioned at the ankle so that the two electrodes flanked the sciatic nerve. Beginning with 1mA, the stimulus was increased until either a maximum compound muscle action potential (CMAP) amplitude, or a stimulus of 3mA was first reached. The location of the stimulating electrodes was marked and the electrodes were moved to the hip to stimulate at the sciatic notch using the same stimulus intensity. The distance between the two stimulation points, along with the latency for each stimulus to elicit an electromyographic response recorded in the foot, was used to determine the conduction velocity (m/s) (33).

Statistical Analyses

All statistics (except for wheel running in Supplemental Figure 3, which utilized the IBM SPSS Statistics 26 package; Armonk, NY) were calculated using Prism 7/8 (GraphPad; San Diego, CA). Scatter dot plots display bars indicating Mean ± SEM where applicable. P-values for neuritis and flow cytometry scatter dot plots comparing two groups are two-tailed Mann-Whitney analyses. For graphs with two groups and two comparisons (Figure 2B and Figure 3H) 2-way ANOVA with Sidak’s multiple comparisons were performed. For graphs with three groups and one comparison, 1-way ANOVA with Tukey’s multiple comparisons were performed. For graphs with three groups and more than two comparisons (Figure 5C, E), a 2-way ANOVA with Tukey’s multiple comparison were performed. P-values for diabetes incidence curves are calculated by Mantel-Cox analysis.

Figure 2 – NOD-PerIg whole splenocytes transfers neuritis into NOD.scid but not NOD.scid-PerIg recipients.

Whole splenocytes (normalized to contain 1×10⁶ T-cells) from 4–6-week-old NOD-PerIg mice were transferred into NOD.scid or NOD.scid-PerIg recipients. (A) T1D incidence in NOD.scid and NOD.scid-PerIg recipients. (B) Quantification of yield of transgenic IgM^a+ or non-transgenic IgM^b+ CD19⁺ B-cells in the spleens of NOD.scid and NOD.scid-PerIg recipients (NOD.scid n=7, NOD.scid-PerIg n=4). (C) Yield of CD8⁺ TCRβ⁺ cells in the spleens of NOD.scid and NOD.scid-PerIg recipients (NOD.scid n=7, NOD.scid-PerIg n=4). (D) Yield of CD4⁺ TCRβ⁺ cells in the spleens of NOD.scid and NOD.scid-PerIg recipients (NOD.scid n=7, NOD.scid-PerIg n=4). (E) Number of hind limbs displaying signs of neuritis (NOD.scid n=7, NOD.scid-PerIg n=10).

Figure 5 – Islet and sciatic nerve infiltrating CD4⁺ T-cells are highly activated.

NOD.scid mice were injected with whole splenocytes (normalized to contain 1×10⁶ T-cells) from 4–6-week-old NOD-PerIg mice. Mice were monitored weekly for T1D and neuritis development. Upon diabetes or neuritis development, mice were removed from incidence and spleen and sciatic nerve-infiltrating CD4⁺ T-cells were examined for cytokine production and surface marker phenotype by flow cytometry. For non-diabetic mice, islet-infiltrating CD4⁺ T-cells were also analyzed. (A) T1D or neuritis incidence in NOD.scid recipients of NOD-PerIg splenocytes (n=38 from three separate cohorts). (B) Representative staining pattern of IFN-γ, IL-4,5,13, IL-17A, and TNF-α on live gated sciatic nerve CD45.1⁺ CD90⁺ TCRβ⁺ CD4⁺ cells. (C) Quantification showing Mean±SEM percent indicated cytokine amongst CD4⁺ T-cells in the indicated organ (Data is combined from 5–6 experiments; spleen n=12 for TNF-α, n=21 for other cytokines; islet n=10 for TNF-α, n=11 for other cytokines; sciatic nerve n=17 for TNF-α, n=26 for other cytokines). (D) Representative staining pattern for IFN-γ vs TNF-α on live gated sciatic nerve CD45.1⁺ CD90⁺ TCRβ⁺ CD4⁺ cells. (E) Quantification of indicated cytokine combination amongst CD4⁺ T-cells in the indicated organ (Data is combined from 3–4 experiments; spleen n=12, islet n=10, sciatic nerve n=17). (F) Quantification of median fluorescence intensity (MFI) of the indicated cell surface marker on gated CD4⁺ T-cells showing Mean±SEM (Data is combined from 4–5 experiments; spleen n=17, islet n=14, sciatic nerve n=21). (G) Based of CD4⁺CD25⁺ phenotype proportion of Tregs in spleens, islets, and sciatic nerves. (Data is combined from 4–5 experiments; spleen n=17, islet n=14, sciatic nerve n=21). (H) Quantification of MFI of the indicated cell surface marker on gated CD4⁺ T-cells showing Mean±SEM (Data is combined from 4–5 experiments; spleen n=17, islet n=14, sciatic nerve n=21). (I) Representative staining pattern for CD4 vs TIM-3 on live gated sciatic nerve CD45.1⁺ CD90⁺ TCRβ⁺ CD4⁺ cells showing gated TIM-3^HI cells. (J) Quantification of the proportion of TIM-3^HI cells amongst CD4⁺ T-cells (Data is combined from 2–3 experiments per organ; spleen n=10, islet n=11, sciatic nerve n=14).

Results

NOD-PerIg T-cells transfer diabetes and neuritis to NOD.scid mice

We initially tested whether a dose of 1×10⁶ purified T-cells from 5-week-old NOD-PerIg mice transferred T1D to NOD.scid recipients at a faster rate than T-cells from NOD controls. Likely due to the young age of the donors, a single recipient of NOD T-cells developed diabetes by 20 weeks post-transfer. There was no T1D in recipients of NOD-PerIg T-cells (Figure 1A). However, during weekly glucosuria testing, it was observed that recipients of NOD-PerIg T-cells began to display a sudden visible hind leg impairment after approximately 15 weeks post-transfer. Mice were sent to The Jackson Laboratory Necropsy Core to determine the cause of this visible impairment. Prosector observation prior to necropsy assessed the phenotype as uncoordinated hind leg movement while walking and an inability to discern presented edges. As detailed in Table I, blinded pathology results indicated control NOD.scid and NOD-PerIg mice had no visible lesions. Recipients of NOD T-cells also had no visible lesions. However, recipients of NOD-PerIg T-cells had severe lymphocytic neuritis in the hind limbs with some evidence of cranial and spinal nerve infiltration (Table I; Figure 1 B, C, D; Supplementary Figure 2). Both mice examined for front limb involvement also had severe lymphocytic neuritis (Table I).

Table I –

Histological Findings In Two Cohorts of NOD-PerIg vs NOD T-cell Transfers

Group	Treatment	Brain	Spine	Hind Limb	Front Limb
Control NOD.scid (n=3)	None	NVLa	NVL	NVL
Control NOD-PerIg (n=3)	None	NVL	NVL	NVL
NOD.scid (n=4b)	NOD T Cells	NVL	NVL	NVL
NOD.scid (n=7b,c)	PerIg T-cells	3/5 mice examined: Multifocal – few nerves affected. Primarily gliosis with few lymphocytes.	4/5 mice examined: Multifocal (a few nerves affected) lymphocytic neuritis	7/7 mice examined: Severe lymphocytic neuritis.	2/2 mice examined: Severe lymphocytic neuritis.
NOD.scid (n=10)d	NOD T Cells			NVL
NOD.scid (n=13)d	PerIg T-cells			2/13 mice examined: Lymphocytic neuritis

^a– NVL=No Visible Lesion

^b– Cohort 1

^c– 6 mice from Cohort 1 (Figure 1A) + 1 mouse that had a thymoma at 15 weeks and was removed from incidence study.

^d– Cohort 2 (Figure 1E, F).

A 2^nd cohort of NOD.scid mice were injected with 1×10⁶ T-cells enriched from 7-week-old NOD or NOD-PerIg donors to determine whether a similar phenotype was observed. This time, diabetes was efficiently transferred by both NOD and NOD-PerIg T-cells, likely due to the slightly older age of the donors (Figure 1E). In Cohort 1, hind limb inflammation was more consistent than cranial and spinal nerve involvement (Table I; Supplementary Figure 2). Thus, in Cohort 2 hind limbs were analyzed for neuritis at diabetes onset or at the end of incidence. None of the recipients of NOD T-cells had hind limb neuritis, while this phenotype was present in 2/13 of the recipients of NOD-PerIg T-cells (Table I). Only three recipients of NOD-PerIg T-cells in Cohort 2 remained T1D free up to 15 weeks post-transfer when neuritis was present in similar Cohort 1 recipients. One of these three mice subsequently developed T1D and had neuritis. Another such recipient also subsequently developed T1D without neuritis. The third recipient remained T1D free to 20 weeks post-transfer and did develop neuritis (Figure 1F). This mouse with neuritis but not T1D at 20 weeks post-transfer had insulitis levels on par with the three recipients of NOD T-cells that remained normoglycemic (Figure 1G). These collective results indicate neuritis is only mediated by T-cells that previously had opportunities for interacting with peripherin autoreactive B-lymphocytes.

NOD-PerIg whole splenocytes transfers neuritis to NOD.scid but not NOD.scid-PerIg recipients

Next, we asked if transfer of whole splenocytes from NOD-PerIg donors could also induce neuritis in NOD.scid or NOD.scid-PerIg recipients, or whether neuritis development was a consequence of transferring purified T-cell populations. Splenocytes from 4–6-week-old NOD-PerIg mice were transferred into NOD.scid or NOD.scid-PerIg recipients. T1D developed at a faster rate in NOD.scid-PerIg than NOD.scid recipients (Figure 2A). Upon T1D development, or at the end of the study, recipient spleens were harvested for evaluation of engraftment, pancreases were harvested for insulitis analysis, and hind limbs examined for neuritis. As expected, NOD.scid-PerIg recipients had a large number of splenic PerIg B-lymphocytes (mean 2.82×10⁷) (Figure 2B). These cells primarily represented the peripherin autoreactive B-lymphocytes endogenous to NOD.scid-PerIg recipients, with a minor number derived from the donor splenocyte inoculum. Much lower numbers (mean 9.65×10⁵) of splenic B-lymphocytes were found in NOD.scid recipients (Figure 2B), all of which are derived from the donor mice. There was no difference in CD8⁺ T-cell engraftment in the spleens of either recipient type (Figure 2C). Conversely, there was a greater expansion of CD4⁺ T-cells in the spleens of NOD.scid-PerIg recipients (Figure 2D). While none of the NOD.scid-PerIg recipients had any hind leg neuritis, four NOD.scid recipients had neuritis in both hind limbs, one had a single affected limb, and two were free of visible lesions (Figure 2E). Thus, NOD-PerIg immunological effectors can only induce neuritis when transferred into completely lymphocyte bereft NOD.scid recipients, but not when transferred into NOD.scid-PerIg hosts in which peripherin autoreactive B-lymphocytes are present in large numbers. The ability of such peripherin autoreactive B-lymphocytes to quickly activate adoptively transferred diabetogenic T-cells may not allow NOD.scid-PerIg hosts to survive long enough post-engraftment to develop neuritis.

Severe damage of sciatic nerves by infiltrating T-cells

We next assessed if we could identify a means of predicting imminent neuritis development using measurable criteria. Towards this end, we generated another cohort of NOD.scid recipients engrafted with NOD-PerIg splenocytes and provided the mice to Jackson Laboratory’s Center for Biometric Analysis for blinded behavioral testing. Controls were NOD.scid mice injected with PBS. At week 15 post-transfer, NOD-PerIg recipients started to show symptoms of neuritis. Wheel running, rotarod, Von Frey, grip strength and hotplate analyses were carried out up to 16 weeks post-transfer. NOD-PerIg recipients displayed decreased balance and coordination in the rotarod assay and a modestly higher Von Frey response only at the 15/16-week post-transfer time point when neuritis was becoming overt (Supplementary Figure 3). Only at this late post-engraftment time point were NOD-PerIg recipients significantly less active during the night as measured by voluntary running wheel activity (Supplementary Figure 3E). Based on these data, it appears that prior to clinical symptoms, none of the tested mouse behaviors could be used to predict imminent neuritis onset. This is likely due to the clinical disease state developing suddenly, precluding an ability to observe any prior behavioral deficit.

Due to the overall negative results of the above behavioral analyses, we used a different approach to analyze the kinetics of nerve damage in this model. We set up a new cohort of NOD.scid mice engrafted with splenocytes from 4–6-week-old NOD-PerIg donors. Untreated NOD.scid mice provided controls. Mice were monitored weekly for T1D or neuritis development (Figure 3A). Prior to 15 weeks post-transfer, mice with T1D but no clinical signs of neuritis were removed from incidence and sciatic nerves analyzed for cellular infiltration (Figure 3D–K). After 15 weeks post-transfer, upon development of T1D or neuritis, or at the end of incidence, mice were analyzed for nerve conduction velocity, sciatic nerve histology, and sciatic nerve cellular infiltration via flow cytometry (Figure 3B–K). We should note, three mice with neuritis, one without clinical symptoms, and all surviving controls from the behavioral analysis cohort are also included in the below analyses (Figure 3B–K).

While control NOD.scid mice had motor nerve conduction velocities of 26.72 ± 1.482 m/s, the recipients of NOD-PerIg splenocytes with visual signs of hind leg impairment had nearly undetectable responses, with only four of eleven producing readable CMAP amplitudes (Figure 3B). Due to a lack of detectable CMAPs in 7/11 mice with clinical neuritis, nerve conduction velocities can only be calculated on four mice with clinical symptoms. Of the four clinical neuritis mice with detectable CMAP amplitudes, calculated nerve conduction velocities were similar to controls (22.86 ± 1.316m/s). In contrast, all mice without visible signs of neuritis had CMAPs (Figure 3B) and calculated motor nerve conduction velocities (22.36 ± 1.571 m/s) similar to controls. We should note, the large spread in CMAPs seen in control and no-clinical neuritis mice is due to technical variation resulting from slight differences in electrode placement. However, it is important to reiterate, all control and no-clinical neuritis mice had detectable CMAP amplitudes, whereas only 4/11 of those with clinical neuritis had such detectable responses. Sciatic nerve histology revealed that mice with clinical symptoms had hyper-cellularity and fewer axons. In the axons that did remain, there were signs of complete demyelination as well as the presence of myelin ovoids indicative of ongoing demyelination (Figure 3C, D). There were also occasional “onion bulb” structures, indicative of chronic demyelination and repair cycles (Figure 3D). The combined loss of axons and demyelination of remaining axons likely explains the inability to generate detectable responses to nerve stimulation under the test conditions utilized (0.5 Hz, 1–3 mA) in a majority of recipients with overt neuritis (Figure 3B). Recipients of NOD-PerIg splenocytes with no clinical indication of neuritis had nerve histology that was intermediate between controls and those with overt disease. While some mice without clinical symptoms had histology that was indistinguishable from controls, others had signs of demyelination and increased inter-axon invasive cellularity compared to controls. For the most part, mice without clinical symptoms had nerves that were closer in appearance to controls than nerves from mice with overt neuritis (Figure 3C). In fact, three blinded observers could not consistently assign a status of “control” or “no clinical symptoms” based solely on the appearance of nerve histology. Thus, these histological analyses also support an acute onset of disease pathology.

We prepared single cell suspensions of sciatic nerves from NOD.scid controls or those engrafted with NOD-PerIg splenocytes to determine the makeup of infiltrating leukocyte populations. Mice with clinical symptoms of hind leg neuritis displayed a higher frequency of CD45.1⁺ leukocytes within sciatic nerves compared to NOD.scid controls (Figure 3E, Supplementary Figure 1E). Mice without clinical signs of neuritis had a much wider spread of CD45.1⁺ leukocytes within sciatic nerves. This is likely explained by the fact that some of these mice were likely near the cusp of developing clinical symptoms, while others were much earlier in the disease process. Amongst CD45.1⁺ cells, NOD.scid controls had little non-specific background staining for CD90⁺ TCRβ⁺ cells, whereas ~60% of leukocytes were T-cells in recipients of NOD-PerIg splenocytes exhibiting overt neuritis (Figure 3F, G). Mice without clinical signs of neuritis had decreased levels of T-cells compared to those with overt disease, however, they had increased levels compared to NOD.scid controls (Figure 3F, G). For all recipients of NOD.scid-PerIg splenocytes, infiltrating T-cells were made up of both CD4⁺ and CD8⁺ T-cells in a roughly 60/40 ratio (Figure 3H, I). There was very little B-lymphocyte infiltration (<4% of CD45.1⁺ cells) in recipients of NOD.scid-PerIg splenocytes (Figure 3J, K). We also examined changes amongst CD45.1⁺ non-lymphocytes and found that while there was an increase in CD11c⁺ cells and a decrease in Ly6c⁺ cells amongst CD11b⁻ CD11c⁻ cells (Supplementary Figure 1F), there were no other major differences in CD11b⁺ or CD11c⁺ subpopulations based off Ly6c or Ly6g gating (Supplementary Figure 1G–I). Collectively these data indicate the hind limb neuritis is, as expected, an overwhelmingly T-cell mediated disease causing widespread damage to peripheral nerve architecture.

CD4⁺ T cells are necessary and sufficient for neuritis development

The CD4⁺ subset predominates T-cells infiltrating sciatic nerves. However, we wondered whether neuritis transfer required both CD4⁺ and CD8⁺ T-cells from NOD-PerIg mice, or whether either population was sufficient to induce neuritis. To address these possibilities, we transferred NOD-PerIg splenocytes into NOD.scid recipients. Recipients were then treated with depleting CD8 or CD4 antibodies. Controls were two separate cohorts of PBS-vehicle treated mice that were injected on the same schedule as those receiving anti-CD8 or anti-CD4 treatments. PBS mice are combined in the resulting analyses. It should be noted the CD4 and CD8 antibodies also served as reciprocal controls for each other. While ~60% of PBS treated control recipients developed T1D by 25 weeks post-transfer, none treated with either anti-CD8 or anti-CD4 did so (Figure 4A). We analyzed sciatic nerves of mice upon either diabetes or clinical neuritis development or at the end of incidence for unafflicted mice. Amongst non-diabetic survivors, recipients treated with anti-CD8 or anti-CD4 had reduced insulitis scores compared to controls (Figure 4B). Additionally, insulitis scores were lower in anti-CD4 than anti-CD8 treated mice (Figure 4B). No differences were observed between PBS and anti-CD8 treated mice in terms of severity of neuritis (Figure 4C, E, Table II). No anti-CD4 treated mice developed any clinical signs of neuritis and only 4 mice had a few sporadic (i.e. non-clustered) lymphocytes in the sciatic nerves. This level of infiltration did not rise to the minimal neuritis designation observed in control or anti-CD8 treated mice (Figure 4C). Interestingly, since clinical symptoms progressed slightly further than in previous experiments, some individuals amongst anti-CD8 injected mice and PBS-injected controls developed myositis (Figure 4D, F, Table II). Anti-CD4 treated mice were completely free of myositis (Figure 4D). Severity levels of neuritis and myositis in the NOD.scid recipients were not completely overlapping. However, no myositis was observed without at least a minimal level of co-occurring neuritis (Table II). Anti-CD4 or CD8 treatment specifically removed their intended target populations (Figure 4G, H). Anti-CD4 treated mice did have an expansion of splenic CD8⁺ T-cells compared to control recipients (Figure 4H). No differences were observed in splenic B-lymphocyte engraftment (Figure 4I).

Figure 4 – CD8⁺ and CD4⁺ T-cell depletion prevents T1D but only CD4⁺ depletion prevents neuritis development.

Whole splenocytes from 6-week-old NOD-PerIg donors (normalized to contain 1×10⁶ T-cells) were injected i.v. into NOD.scid recipients. One group of recipients were then injected i.p. every three weeks with PBS or 250μg anti-CD8. Another group was injected i.p. once a week for four weeks, followed by once every two weeks with either PBS or 250μg anti-CD4. (A) T1D incidence in NOD.scid recipients of NOD-PerIg splenocytes subsequently treated with PBS (both cohorts combined), anti-CD8, or anti-CD4 treated groups. (B) Mean insulitis scores for non-diabetic survivors (n=10 PBS, n=13 anti-CD8, n=13 anti-CD4). (C) Quantification of neuritis severity. Each dot represents a mouse with a pathology finding of Severe, Moderate, Minimal, or No Visible Lesion (NVL) in hind leg pathology samples (n=24 PBS, n=13 anti-CD8, n=13 anti-CD4). To display Mean±SEM, data was entered in Prism as Severe = 3, Moderate = 2, Minimal = 1, NVL = 0, although the y-axis displays the pathologist’s description. For the anti-CD4 group, 4 mice had some sporadically spaced, non-clustered lymphocytes, technically indicating infiltration (since these are NOD.scid recipients). This infiltration, however was closer in appearance to NVL than Minimal, but could not fully be classified as being free of lymphocyte infiltration. Therefore, for purposes of plotting, they were given a score of 0.25 to indicate that they were not free of infiltration, but were closer to NVL than the representative Minimal seen in E. (D) Quantification of myositis severity. Each dot represents a mouse with a finding of myositis, minimal myositis or no myositis (n=24 PBS, n=13 anti-CD8, n=13 anti-CD4). To display Mean±SEM, data was entered in Prism as Myositis = 2, Minimal Myositis = 1, No Myositis = 0, although the y-axis displays the pathologist’s description. (E) Representative histology showing neuritis at NVL, Minimal, and Severe levels (Scale bar 200 μm). Black arrows designate representative area of infiltration in affected tissue. (F) Histology showing myositis at NVL, Minimal, and Severe levels (Scale bar 200 μm). For Myositis, the most extreme case is shown. Black arrows designate a representative area of infiltration within affected tissue. (G) Quantification of CD4⁺ T-cells in recipient spleens. (H) Quantification of CD8⁺ T-cells in recipient spleens. (I) Quantification of CD19⁺ B-lymphocytes in recipient spleens.

Table II –

Histological Findings for mice injected every three weeks with either PBS or Anti-CD8

Mouse ID	Treatment	Neuritis	Myositis
AF18748	PBS	NVLa	No
AF18749	PBS	NVL	No
AF18750	PBS	Minimal	No
AF18723	PBS	Severe	Yes
AF18724	PBS	Severe	Yes
AF18725	PBS	Severe	No
AF18739	PBS	NVL	No
AF18740	PBS	Severe	No
AF18741	PBS	Severe	Yes
AF18742	PBS	Severe	Yes
AF18729	Anti-CD8	Severe	Yes
AF18732	Anti-CD8	Minimal	Mild
AF18606	Anti-CD8	Severe	Yes
AF18607	Anti-CD8	Severe	Yes
AF18608	Anti-CD8	Severe	Yes
AF18764	Anti-CD8	Minimal	No
AF18765	Anti-CD8	Minimal	No
AF18766	Anti-CD8	Minimal	No
AF18767	Anti-CD8	Moderate	Mild
AF18773	Anti-CD8	Mild	No
AF18774	Anti-CD8	NVL	No
AF18775	Anti-CD8	Severe	No
AF18776	Anti-CD8	Minimal	No
AF18777	Anti-CD8	Minimal	No

^a– NVL=No Visible Lesion

Phenotype of T-cells causing neuritis

To begin addressing the potential mechanisms by which CD4⁺ T-cells initiate nerve destruction, we setup three additional cohorts of NOD.scid recipients (Figure 5A). In addition to analyzing sciatic nerve infiltrating T-cells, we also examined islet-infiltrating T-cells in non-diabetic neuritic or unafflicted survivors. Both islet and sciatic nerve infiltrating CD4⁺ T-cells had a small expansion of IFN-γ producing cells compared to those in spleens (Figure 5B, C). In contrast, both sites of inflammation had reductions in TNF-α producing CD4⁺ T-cells (Figure 5B, C) compared to the spleen, with the greatest reduction being found in those within sciatic nerves. For all three tissues, the proportion of TNF-α producing CD4⁺ T-cells was greater than IFN-γ producing cells, though the disparity was greater in the spleen and islets compared to the sciatic nerve (p<0.0001 for spleen and islet, p=0.0458 for sciatic nerve). Very little TH2 (IL-4, 5, and 13) or TH17 (IL-17) CD4⁺ T-cells were observed. (Figure 5B, C). Taken together, these data indicate the potential contribution of IFN-γ and TNF-α in sciatic nerve damage.

The above analyses did not account for co-expression of IFN-γ and TNF-α. Thus, we examined the extent these populations of IFN-γ and TNF-α expressing cells may overlap. No differences were observed in IFN-γ and TNF-α co-expressing cells, while there was a slight increase in IFN-γ⁺ TNF-α⁻ CD4⁺ cells in both islets and sciatic nerves compared to the spleen (Figure 5D, E). There was a decrease in the proportion of cells IFN-γ⁻ TNF-α⁺ CD4⁺ T-cells in islets, and an even greater reduction within sciatic nerves (Figure 5D, E). Most significantly, both islets and sciatic nerve CD4⁺ T-cells had an expansion of IFN-γ⁻ TNF-α⁻ cells (Figure 5D, E). Taken together the modest expansion of IFN-γ producing cells and the drop in those producing TNF-α producing cells may indicate a greater role for the former cytokine in sciatic nerve damage. However, the greatest change appears to be the expansion of IFN-γ⁻ TNF-α⁻ cells.

The expansion of IFN-γ⁻ TNF-α⁻ cells within sites of inflammation was surprising. To further dissect the nature of CD4⁺ T-cells in the spleen and sites of inflammation, we next examined markers for cell surface activation (Figure 5F) and tolerance (Figure 5G–J). As expected, compared to those within spleens, CD4⁺ T-cells infiltrating islets or sciatic nerves were highly activated, having increased expression of CD69, ICOS, and IL-7Rα (Figure 5F). Furthermore, based on a CD4⁺CD25⁺ phenotype, proportions of regulatory T-cells (Tregs) with a potential ability to modulate the activation state of effector populations did not differ in spleens, islets, and sciatic nerves of NOD.scid recipients engrafted with NOD-PerIg splenocytes (Figure 5G). Taken together, these data indicate that despite the expansion of an IFN-γ⁻ TNF-α⁻ population in both sciatic nerves and islets, CD4⁺ T-cells at these sites are more activated than their splenic counterparts.

Next, we examined markers of function and tolerance. We were unable to detect much expression of FasL, CD40L, OX40, or LAG-3 on CD4⁺ T-cells (data not shown). However, islet-infiltrating CD4⁺ T-cells had increased expression of CD95 compared to those within spleens and sciatic nerves (Figure 5H). PD-1 expression was increased in sciatic-nerve infiltrating CD4⁺ T-cells, and to a greater extent by those within islets than that observed within spleens (Figure 5H). Finally, we found that sciatic nerve-infiltrating CD4⁺ T-cells had an expansion of TIM-3^Hi cells compared to those in spleens and islets (Figure 5I, J). Taken together, these data indicate that while exhibiting some differences, islet and sciatic nerve infiltrating CD4⁺ cells do have high expression of negative co-stimulatory molecules indicating they have progressed through the end stages of activation.

NOD-PerIg mice can develop neuritis if diabetes onset is sufficiently delayed

Finally, we asked whether the neuritis phenomenon was limited to transfer conditions, or if T1D could be attenuated, this pathology could eventually develop in NOD-PerIg mice. The accelerated T1D development in NOD-PerIg mice is due to peripherin reactive B-lymphocytes (25). Despite known issues of lessened efficacy of anti-CD20 in NOD mice (34, 35), we reasoned that even partial depletion of NOD-PerIg B-lymphocytes from an early age might delay T1D development. Starting at 4 weeks of age, we treated NOD-PerIg mice with anti-CD20. Anti-CD20 treatment significantly delayed the time of disease onset, but not the overall high penetrance of T1D in NOD-PerIg mice (Figure 6A). One of two mice which developed T1D at 18 weeks of age showed signs of neuritis (Figure 6B). Two mice, taken down at 27 weeks of age showed no signs of neuritis. A final mouse, taken down at 29 weeks of age with no T1D, showed signs of both neuritis and myositis (Figure 6C, D). Taken together, these data indicate NOD-PerIg mice are capable of developing neuritis, but their accelerated T1D status normally masks their susceptibility to this pathology. Additionally, these data indicate neuritis inducing NOD T-cells only require early transient interactions with peripherin reactive B-lymphocytes.

Figure 6 – B-lymphocyte depletion reveals neuritis can develop in NOD-PerIg mice if T1D is delayed.

NOD-PerIg mice were injected i.p with 250μg anti-CD20 every 2 weeks starting at 4 weeks of age. (A) T1D incidence comparing unmanipulated or anti-CD20 treated mice. (B) Representative histology of a diabetic mouse with neuritis. Black arrow designates representative area of infiltration in affected nerve tissue. (C, D) Mouse that survived to 29 weeks of age without T1D showing representative histology of (C) neuritis and (D) myositis. Black arrow designates representative area of infiltration in affected tissue.

Discussion

To our knowledge, this is the first model of peripheral neuritis in NOD mice that is based off expansion of T-cells mediated by T1D contributory peripherin-autoreactive B-lymphocytes. Previous reports of peripheral neuritis in NOD mice have been associated with modulations to immune co-stimulation (4, 6–8), as opposed to our model where disease is initiated by a naturally occurring autoreactive B-lymphocyte population. Although T1D can be transferred by T-cells from NOD-PerIg mice, depending on the kinetics of that syndrome, anti-peripheral nerve autoimmune responses can also occur. Our current findings also provide further evidence of the overlap in the autoimmune repertoire that targets islets versus nervous system components, as has been reported in T1D and multiple sclerosis patients (36).

We have several lines of evidence that peripherin reactive B-lymphocytes interact with cognate T-cells in the NOD.PerIg mouse models. First, in NOD.PerIg mice, there is an increase in both germinal center B-lymphocytes and T-follicular helper cells compared to standard NOD controls (25). Second, naïve T-cells (NOD-IgHEL.Igμ^null) transferred diabetes to NOD.scid-PerIg mice at a faster rate than to standard NOD.scid mice, and disease exacerbation was associated with an expansion of CD4⁺ T-cells (25). Third, we know NOD-PerIg B-lymphocytes actively participate within the islet lesion, as evidence by their extensive proliferation at this site (25), which would require both cognate antigen and T-cell help (25). And finally, preliminary studies have found that purified NOD-PerIg B-lymphocytes engraft better in NOD.scid recipients when co-transferred with cognate NOD-PerIg, but not NOD T-cells (Supplementary Figure 4).

In all previously reported cases of neuritis, nerves have undergone demyelination (4, 6–8), with myelin itself being identified as an antigenic target in several of these models (5, 6). Thus, these represent primary demyelinating models. In the peripheral nervous system, peripherin is a subunit of neurofilaments (37) within axons. The demyelination observed in this study is likely secondary to axon damage initiated by T-cells that have been expanded by anti-peripherin but not polyclonal B-lymphocytes. If peripherin remains the target antigen in the peripheral nerves, we propose two sites where peripherin-reactive T-cells would likely initiate axonal destruction: unmyelinated neurons or nodes of Ranvier. However, the presence of unmyelinated axons might indicate primary reactivity against Schwann cells and myelin and not peripherin itself. Since inflammatory cytokines can cause the formation of peripherin aggregates, which in turn initiates apoptosis of motor neurons (38), there also exists the possibility of early unmyelinated nerve damage caused by peripherin reactive T-cells triggering a cascade of surrounding neuronal death that could expand the immune response against other neuronal antigens. It is also possible that an immune response directed against myelinated axons at the exposed nodes of Ranvier leads to damage of the flanking Schwann cells triggering the demyelination observed in this study. An immune response targeting axonal antigens at the nodes of Ranvier would not be unprecedented, as recent evidence suggests that nodal regions are the target of B-lymphocytes in Guillain-Barré syndrome and chronic inflammatory demyelinating polyneuropathy (39).

Few NOD-PerIg B-lymphocytes are present within neuritis lesions despite the active role they play in insulitis. This, coupled with the fact that purified T-cells from NOD-PerIg mice transfer neuritis without the need to co-transfer PerIg B-cells indicate these B-cells play a minor role in the lesion after initiating the expansion of neuritis causing T-cells. The reason for the disparity between islet and nerve infiltration of NOD-PerIg B-lymphocytes is currently unknown. Whether an initial anti-peripherin immune response in the islets (driven by the interaction between the peripherin reactive PerIg B-lymphocytes and cognate T-cells) eventually expands anti-myelin (or other neuronal antigen) autoreactive cells is currently unknown. To address whether peripherin does remain the specific antigen being targeted in the peripheral nerves, work is currently ongoing to produce a direct-in-NOD CRISPR/Cas9 Prph1 knockout mouse. Since peripherin plays an important role in the development of unmyelinated neurons (40), this may provide its own set of complications.

As previously stated, our neuritis model relies on the expanded presence of a naturally occurring diabetogenic peripherin autoreactive B-lymphocyte population (25), and not changes to costimulatory signaling (4, 6–8). While only two anti-CD20 treated NOD-PerIg mice developed neuritis (Figure 6A), this is likely because while exhibiting a somewhat slowed disease onset, they were still characterized by aggressive T1D development. We have not observed any clinical signs of neuritis development in standard NOD mice undergoing types of B-lymphocyte deletion therapies (34, 35) out to 40 weeks of age. Furthermore, we are currently unaware of any reports on NOD mice given long term insulin injections to extend survival developing sudden onset neuritis. However, there have been cases of age-related neuritis/meningitis development in unmanipulated NOD/ShiLtJ mice (1, 2) a close relation to our NOD substrain (NOD/ShiLtDvs) (27). We suggest that longer term pre-clinical T1D intervention studies in NOD mice may be required to determine whether when T1D is attenuated, other autoimmune diseases become more prominent. T1D and multiple sclerosis patients reportedly have overlapping T-cell repertoires with a potential to target both nervous systems components and islets (36). Thus, it is also possible that an intervention ultimately found to inhibit T1D development in humans at high risk for this disease might engender the appearance of other autoimmune pathologies in such individuals.

We have previously reported myositis development in NOD mice carrying a transgene encoding a CD2 promoter driven IFN-γ receptor β-chain (41). In that model, myositis appears to develop in the lumbar region of the spinal column prior to migrating to the limbs, without the appearance of neuritis. NOD-PerIg T-cells appear to target peripheral nerves first as we documented myositis development in both anti-CD8 and PBS treated recipients when mice were not immediately removed from incidence studies upon initial visible neuritis onset. As they were not allowed to deteriorate further after the appearance of overt neuritis, it is unknown if these mice would have subsequently developed a similar level of myositis characterizing the CD2 promoter driven IFN-γ receptor β-chain transgenic NOD stock (41). Furthermore, no myositis was observed without at least minimal neuritis (Table II). We did observe cases of severe neuritis without accompanying myositis (Table II). Whether the T-cells that infiltrate the nerves are different than those causing myositis, or whether the populations involved in the myositis here are similar to those observed in our previous study (41) is not currently known.

We were surprised by the expansion of IFN-γ⁻ TNF-α⁻ T-cells in islets and sciatic nerves compared to the spleen (Figure 5D, E). However, due to the highly activated nature of the cells in islets and sciatic nerves (Figure 5F), and the expansion of negative-feedback co-stimulatory molecules (Figure 5H–J), it is possible we are detecting the expansion of terminally differentiated cells that are no longer participating in disease pathogenesis. Interestingly, the pattern of these markers differs between islets and sciatic nerves (Figure 5H–J) likely indicating these cells differ between the two sites of inflammation. The expansion of TIM-3^HI cells in the sciatic nerves (Figure 5I, J) may indicate a narrower range of antigen reactive T-cells than within islets, hence the progression to a more TIM-3⁺-exhausted phenotype.

Finally, the finding that CD8-depleted splenocytes from NOD-PerIg donors allowed development of neuritis, but not T1D is interesting. The lack of T1D development resulting from CD8⁺ T-cell depletion was expected, as such cells from standard NOD donors are required to transfer this disease to NOD.scid recipients unless the donors are already hyperglycemic (42). However, the ability to transfer aggressive neuritis may indicate that sufficient anti-neuronal CD4⁺ T-cell responses developed at an early age in NOD-PerIg donors, whereas young NOD-PerIg mice have not yet generated a repertoire of diabetogenic CD4⁺ T-cells capable of independently inducing hyperglycemia in secondary recipients. We do not know how T-cells infiltrating sciatic nerves may be related to those recognizing peripherin that attack pancreatic ß-cells. However, our cell surface marker analysis reveals a slightly different phenotype of CD4⁺ T-cells in islets versus sciatic nerves, which may indicate a divergent pool of antigen-specific T-cells. To fully dissect this issue, future studies identifying unique and overlapping TCR sequences between islet and nerve infiltrating T-cells is warranted but is outside the scope of this initial report.

Key Points.

T-cells or whole spleen from NOD-PerIg mice transfer neuritis to NOD.scid recipients.

CD4⁺ T-cells are necessary and sufficient to transfer neuritis.

Islet and sciatic nerve CD4⁺ T-cells have differing patterns of CD95, PD-1 and Tim-3.

4. Abbreviations

(CMAP)

compound muscle action potential

(NVL)

no visible lesion

(T1D)

type 1 diabetes