Инженерная иммунологическая ниша направляет терапевтическое развитие в модели прогрессивного множественного склероза.

Multiple sclerosis (MS) is a demyelinating autoimmune disease characterized by immune cell infiltration into the central nervous system (CNS) that affects nearly three million people ( 1 ). Of MS patients, approximately 15% exhibit a primary progressive (PPMS) disease course associated with accumulating CNS damage and associated morbidity ( 2 ). The disease course of PPMS is more rapid than for other forms of MS, disability is typically irreversible, and in 7 to 15 y, 50% of patients require at least a cane for assistance walking ( 3 , 4 ). Because of the rapid progression and associated disability, effective and efficient treatment of PPMS patients is critical. The first treatment for PPMS, ocrelizumab, an anti-CD20 antibody that depletes B cells and slows disability progression, was approved by the FDA in 2017. Unfortunately, ocrelizumab does not yield complete remission, only modestly slows disability progression, and does not reverse existing disability ( 5 ). Moreover, ocrelizumab broadly suppresses immune function leading to opportunistic infections and attenuating responses to life-saving vaccines ( 6 , 7 ). Other treatments are not commonly used because of a lack of convincing evidence for their success in PPMS patients.

Better therapeutics require a deep understanding of immune dysfunction in disease and the dynamics of cell–cell communication. Unfortunately, the complexity of autoimmunity—including innate and adaptive immune cells exhibiting extensive crosstalk as well as dynamic cell trafficking—creates a challenging system to understand. Additionally, collection of molecular data from diseased tissue is challenging in model systems (typically requiring killing of animals) and certainly not routine in the clinical context (biopsy of the CNS is impractical and cannot be performed repeatedly). We hypothesize that the development of a surrogate inflamed tissue containing both immune and stromal compartments could help identify potential cell–cell communication in PPMS that could be precisely targeted. In prior studies, we have demonstrated use of subcutaneously implantable porous biomaterials, termed immunological niches (INs), that form vascularized tissues containing both immune and stromal cells ( 8 , 9 ). Cells within the INs contain unique information relative to blood, likely resulting from a combination of selective recruitment and phenotype changes, which can be used to diagnose and monitor autoimmune conditions such as the relapsing–remitting form of murine experimental autoimmune encephalomyelitis (EAE) ( 8 ). Other work has demonstrated utility of implantable biomaterial scaffolds as a tool to understand dynamics in cell phenotype and regulation of both immune and stromal cells, particularly in the context of metastatic cancer ( 10 – 17 ).

In this report, we investigate a biomaterial scaffold as a diseased tissue surrogate to monitor immune function, identify potential drug targets, and guide the development of a therapeutic. Using single-cell RNA sequencing (scRNAseq), complex and dynamic populations of both immune and stromal cells were observable within the niche. Analysis revealed that communication between these cell populations is altered by disease, with CC chemokine signaling increased in myeloid cells. Additionally, communication to regulatory T cells (Tregs) was reduced at the IN in disease. These observations motivated the application of a nanoparticle-based therapy previously shown to reduce CCL2 expression to mitigate the enhanced CC chemokine signaling, while simultaneously boosting a Treg response ( 18 ). Within these poly(lactide-co-glycolide) (PLG) nanoparticles (NPs), we encapsulated disease-relevant antigen that promotes Treg populations and mitigates autoreactive responses. Overall, we demonstrate that the IN contains molecular insights into disease progression and guides the development of NPs effective at treating chronic progressive EAE.

Methods

Reagents.

All chemical reagents were purchased from Sigma-Aldrich. Peptides used were custom synthesized by GenScript.

Animals.

All procedures were performed in accordance with the regulations approved by the Animal Care and Use Committee of the University of Michigan (approval # PRO00011621). For all experiments, female C57BL/6J mice (#000664) were purchased from Jackson labs at an age of 5 to 6 wk. Animals were housed in a temperature-controlled room with 12-h light and dark cycles and with food and water available ad libitum.

Scaffold Fabrication and Subcutaneous Implantation.

Microporous poly (caprolactone) (PCL) scaffolds were prepared and implanted one week after arrival (week 6 to 7) as previously described in detail ( 19 , 20 ). In brief, 13 mm (diameter) by 2 mm (height) cylindrical scaffolds with an interconnected network of 250 to 425 µm pores were fabricated by a polymer-salt sintering and salt-porogen leaching process. Scaffolds were disinfected in 70% ethanol and stored at −80 °C until surgical implantation. Mice received isoflurane anesthetic (2 to 4%) before subcutaneous implantation of 2 scaffolds per mouse on the dorsum, and incisions were closed using surgical clips (Reflex 7 mm, Roboz Surgical Instrument Co.). Immediately before and 24 h after surgery, all mice received subcutaneous injections of carprofen (5 mg/kg) and were monitored daily for 10 d postsurgery. Animals did not show signs of distress from scaffold implantation and no further analgesia was necessary. In scRNAseq experiments, one animal (OVA d7) was euthanized prior to end point according to the approved procedure by the institutional guidelines at University of Michigan and excluded from analysis due to loss of surgical clips and wound reopening.

Adoptive Transfer of T Cells.

EAE was induced by adoptive transfer of T cells reactive to myelin oligodendrocyte glycoprotein peptide 35 to 55 (hereafter MOG) similarly to previously published protocols, and control animals received an adoptive transfer of T cells reactive to ovalbumin peptide 323 to 339 (hereafter OVA) ( 8 ). For mice receiving a subcutaneously implanted PCL scaffold, scaffolds were implanted 14 d before the adoptive transfer of T cells. Donor mice were first immunized with an emulsion containing 1:1 complete Freund’s adjuvant [incomplete Freund’s adjuvant (Fisher) with 4 mg/mL heat-killed M. tuberculosis H37 Ra (Fisher)]: PBS with 4 mg/mL peptide (MOG or OVA). Donor mice were injected at three distinct subcutaneous sites with 50 µL of emulsion for a total volume of 150 µL. After 10 d, spleens and lymph nodes (inguinal, axillary, and brachial) were harvested, processed to a single-cell suspension, and cultured in RPMI with 10% FBS, 2 mM L-glutamine, 100 IU/mL penicillin, 0.1 mg/mL streptomycin, 1× nonessential amino acids solution, 1 mM sodium pyruvate, 10 mM HEPES, 50 µg/mL peptide antigen, and 25 ng/mL rmIL-12 p70 (R&D Systems). After three days nonadherent cells were harvested, counted, and injected i.p. into recipient mice. Mice were not randomized for disease induction or treatment, but disease severity was monitored daily by blinded investigators on a 0 to 5 scale as previously described ( 21 ): 0 = no disease, 1 = hindlimb weakness or limp tail, 2 = hindlimb weakness and limp tail, 3 = partial hindlimb paralysis, 4 = total hindlimb paralysis, and 5 = moribund.

Tissue Isolation and Preparation.

Mice were euthanized with CO 2 in accordance with the approved procedure by the institutional guidelines at University of Michigan and tissues were harvested and placed in HBSS on ice. For spleen harvest, spleens were processed into a single-cell suspension by physical agitation through a 70 µm filter, RBCs were lysed in an ACK solution (Gibco), and cells washed in PBS. Isolated splenocytes were then cultured as described below.

Bead-Based Multiplex Immunoassay.

PCL implants were harvested 9 d after adoptive transfer of either MOG or OVA-reactive T cells and flash frozen in isopentane prior to protein isolation and cytokine and chemokine quantification. Briefly, scaffolds were thawed and cut into 1 to 2 mm pieces prior to homogenization in PBS. Cells were then lysed in RIPA buffer (Thermo) supplemented with protease inhibitors (Thermo) for 30 min at room temperature. Samples were centrifuged at 14,000× g for 15 min to remove debris and a BCA assay (Thermo) was used to quantify total protein content in each scaffold according to the manufacturer’s directions. LegendPlex multiplex assays (BioLegend) were used to quantify proinflammatory chemokines (Cat # 740451) and cytokines (Cat # 740446) according to the manufacturer’s instructions.

Splenic Recall.

Splenocytes were cultured in RPMI 1640 supplemented with MOG or OVA peptide (5 µg/mL) at 800,000 cells per well in 96-well flat-bottom plates. After 72 h, the ELISA was conducted to measure cytokine secretion in cell-culture supernatants (University of Michigan Cancer Center Immunology Core).

NP Fabrication.

Peptide-encapsulating NPs were fabricated with the double emulsion method as previously described ( 22 ). Antigen (MOG or OVA peptide) was dissolved in PBS to a concentration of 50 mg/mL and added to PLG dissolved in dichloromethane [20% w/v (inherent viscosity: 0.17dL/g, Lactel)]. Approximately 5% of PLG was labeled with cyanine 5.5 for biodistribution studies. This solution was sonicated at 100% amplitude for 30 s before adding 10 mL of 1% w/v poly(ethylene-alt-maleic anhydride) (PEMA) in water and a subsequent sonication. This double emulsion was stirred in 0.5% PEMA overnight to form particles, particles were washed, and lyophilized in a sucrose/mannitol (3/4% w/v) cryoprotectant solution.

NP Characterization.

The size and zeta potential of NPs were measured at a concentration of 250 ng/mL by dynamic light scattering using a Malvern Zetasizer ZSP. Total antigen loading was quantified by degrading particles in 0.1 M NaOH and subsequent Micro BCA assay (Pierce), as previously reported ( 23 , 24 ). In brief, the Micro BCA assay was done following the manufacturer’s instructions with the modification that the corresponding peptides were used for the standard curve to improve accuracy of the assay to quantify peptide concentration compared to protein concentrations. Antigen release over six days was determined by suspending particles in PBS, incubating at 37 °C and centrifuging particles to change to fresh PBS at each time point. Supernatants were frozen until the final time point at which time a Micro BCA assay was used to determine antigen release. For scanning electron micrographs of the particles, particles were sputter coated with gold before imaging with a TESCAN MIRA3 by the Michigan Center for Materials Characterization.

Bone Marrow–Derived Antigen-Presenting Cells.

Bone marrow–derived macrophages (BMDMs) or bone marrow–derived dendritic cells (BMDCs) were prepared as described previously ( 25 ). Briefly, femurs were isolated from C57BL/6J mice, and the bone marrow was flushed using macrophage growth medium, RPMI 1640 (Sigma) supplemented with 10% fetal bovine serum (FBS), penicillin/streptomycin, 1 mM sodium pyruvate, 25 mM HEPES, 2 mM L-glutamine, 50 μM 2-mercaptoethanol. Cells were pelleted by centrifugation before resuspension in ACK buffer (Gibco) to induce red blood cell lysis. Cells were preincubated on petri plastic for 4 h and those remaining in suspension were then cultured for 6 d in growth media containing 20 ng/mL macrophage-colony stimulating factor (M-CSF) (R&D Systems). The procedure for generating BMDCs was the same, but 20 ng/mL of granulocyte-macrophage colony-stimulating factor (GM-CSF) (R&D Systems) was added instead of M-CSF. Cells were pelleted, counted, and replated at 100,000 cells per well in 24-well plates in growth media containing cytokine. The next morning, wells receiving NPs were incubated with 300 μg/mL NPs for 3 h. Previously published experiments using PLG nanoparticles demonstrated greater than 90% of cells having phagocytosed particles by 3 h ( 18 , 23 ). After 3 h, all wells were washed and fed with media. For conditions containing lipopolysaccharide (LPS), 100 ng/mL LPS (Sigma) was used. Cells were cultured for 2 d before the ELISA was conducted to measure cytokine secretion in cell-culture supernatants (University of Michigan Cancer Center Immunology Core).

RAW 264.7 Cells.

RAW 264.7 cells were obtained from ATCC and maintained in DMEM with 10% FBS, and penicillin/streptomycin. For cytokine assays, cells were pelleted, counted, and replated at 25,000 cells per well in 24-well plates in growth media. The next morning, wells receiving NPs were incubated with 300 μg/mL NPs for 3 h. After 3 h, all wells were washed and fed with media. For conditions containing LPS, 100 ng/mL LPS (Sigma) was used. Cells were cultured for 2 d before the ELISA was conducted to measure cytokine secretion in cell-culture supernatants (University of Michigan Cancer Center Immunology Core).

Single-Cell RNA Sequencing.

PCL implants were processed for scRNAseq ( 26 ) in a similar manner to previous work ( 9 ), and 2 to 3 mice were pooled for each condition at each time point to account for biological variability. Briefly, scaffolds were minced and incubated in RPMI containing 0.2 U/mL Liberase TL (Roche) at 37 °C for 20 min before passing through a 70 µm filter to form a single-cell suspension. RBCs were lysed in an ACK solution (Gibco), and cells washed in PBS before being resuspended to 1 million cells/mL in PBS containing 0.04% w/v BSA.

Single-cell cDNA libraries (approx. 10,000 cells / sample) were prepared by the Advanced Genomics Core at the University of Michigan for 5’ sequencing using the 10x Genomics Chromium technology before sequencing on a NovaSeq 6000 (Illumina) to a depth of 50,000 reads/cell. The Cell Ranger pipeline was used to preprocess data with reads aligned to mouse reference GRCm38 (mm10). The Seurat R package (v.4.3.0) ( 27 ) was used to remove cells with more than 5% mitochondrial genes or fewer than 200 and greater than 3,500 transcripts. NormalizeData and ScaleData were used to normalize the data for downstream analysis and an “anchor-based” integration of each sample was done by FindIntegrationAnchors. PCA was used to reduce dimensionality and identify the top 2,000 highly variable features, and an elbow plot helped determine a dimensionality of 20.

The FindNeighbors and FindClusters functions were used to enable graph-based clustering of the data, with dimensional reduction via UMAP. Markers used to annotate cell types were identified with the FindConservedMarkers function and FindMarkers was used to identify differentially expressed genes. The CellChat package (v.1.6.1) ( 28 ) was used to identify communication between cells, and alterations in communication with disease were identified using differential expression analysis workflow between conditions.

CNS scRNAseq was acquired from the NCBI’s Gene Expression Omnibus repository (GEO Accession GSE200901 ) ( 29 ) and processed as described above. In brief, 8 to 12-wk-old C57BL/6 mice were given an adoptive transfer of MOG-reactive T cells, and CD45+ cells FACS sorted from 4 spinal cords at days 6 and 10 after adoptive transfer were submitted for sequencing ( 29 , 30 ). CellChat comparison analysis between spinal cord and scaffold tissues was done by comparing day 6 from the spinal cord to signaling at day 7 in the IN as well as day 10 from the spinal cord to day 9 in the IN. Cell–cell communication in common between the tissues was identified by comparing communication upregulated in diseased IN (MOG upregulated compared to OVA IN) and communication at the spinal cord at the corresponding timepoint. Cell–cell communication differing between the tissues at the corresponding timepoints was identified in the IN as communication that is not in common or present in the control (OVA) IN and identified in the spinal cord as communication not in common or present in the MOG or OVA IN.

NP Administration.

To administer NPs for in vivo analysis, lyophilized particles were washed in sterile water and centrifuged to pellet particles. Then supernatant containing cryoprotectants was removed, particles were resuspended in PBS and stored on ice until administration. Cryoprotectants and the inherent negative charge of the PLG particles prevent aggregation. Preparation of particles is similar to those used in clinical trials ( NCT03486990 ) ( 31 ). Each mouse received a single tail vein injection of 2.5 mg NPs at the indicated time point.

Biodistribution.

C57BL/6J mice (Jackson Labs) were intravenously injected with 2.5 mg of Cyanine 5.5–labeled particles into the lateral tail vein. After 24 h, mice were euthanized following the approved procedure by the institutional guidelines at University of Michigan. Organs were harvested and imaged with an In Vivo Imaging System Spectrum (IVIS Spectrum, Perkin Elmer).

Flow Cytometry.

For flow cytometric characterization of anergic and regulatory T cells, mice received an adoptive transfer of MOG-reactive T cells, intravenous injection of NPs on day 2 and their spleens were harvested on day 12. Spleens were mechanically processed into single-cell suspensions and red blood cells lysed. Splenocytes were prepared for flow cytometry using previously described protocols ( 8 ). Cells were blocked with anti-CD16/32 (clone 93, eBioscience). Cells were stained with Live/Dead Fixable Violet (Life Technologies), BV605 anti-CD25 (clone PC61, BioLegend), FITC anti-CD4 (clone RM4-5, BioLegend), PerCP/Cy5.5 anti-FR4 (clone 12A5, BioLegend), PE/Cy7 anti-CD8 (clone 53-6.7, BioLegend), APC anti-CD73 (clone TY/11.8, BioLegend), and PE anti-FoxP3 (clone MF-14, BioLegend). Surface markers were stained first, followed by intracellular FoxP3 staining with True-Nuclear Transcription Factor Buffer Set (BioLegend) according to the manufacturer’s instructions.

Imaging Myeloperoxidase in the CNS.

Quantification of autoimmune activity in the CNS by fluorescent and luminescent imaging of live mice has been demonstrated previously ( 32 – 34 ). To image inflammatory activity in the CNS, mice were adoptively transferred with MOG-reactive T cells and treated i.v. with NPs on day 2. Mice were shaved and hair depilated before injection i.p. with 5 mg luminol in 100 uL PBS for in vivo bioluminescence imaging (BLI). Mice were imaged 10 min after injection with an IVIS spectrum (Perkin Elmer) at an exposure time of 3 min.

Statistical Analysis.

All statistics were calculated with GraphPad Prism software, or R. CellChat statistics were calculated using the default function parameters. In brief, differentially expressed genes were identified using a Wilcoxon rank-sum test with P < 0.05 considered significant and a threshold for differential communication between MOG and OVA conditions was set at an absolute log 2 fold change of at least 0.2 for upregulation and 0.1 for downregulation. Normality of data was assessed using a Shapiro–Wilk test. To analyze flow cytometry data, to compare ELISA results from NP+LPS to LPS-treated groups, and to compare BLI measurements, Student’s t tests were used for normally distributed data, and Mann–Whitney tests were used otherwise. For splenic recall experiments, two-way ANOVA with a Sidak corrected post hoc multiple comparisons test was used. For NP efficacy experiments, differences between disease courses were analyzed using the Kruskal–Wallis test (one‐way ANOVA nonparametric test) ( P < 0.05 considered significant). Data prior to treatment were excluded from analysis [e.g., PLG(MOG) treatment on day 9 (d9) was compared to PLG(OVA) beginning on day 10 (d10)]. A one-way ANOVA with Tukey’s multiple comparison test was used to compare clinical scores at the terminal time point. For all conditions, P < 0.05 was considered statistically significant.

Results

IN as a Disease Tissue Surrogate for Analyzing Immune Dynamics.

INs were created by implanting the microporous PCL scaffolds (13 mm diameter, 2 mm height, 250 to 425 µm pores) into the subcutaneous space of C57BL/6J mice ( Fig. 1 ). After two weeks, adoptive transfer of T cells was used for disease induction (reactive against MOG) or served as controls (reactive to OVA). INs were harvested early in disease onset at day 7 or day 9 posttransfer and processed for scRNAseq. We identified 20 distinct cell types within the IN, that indicate a complex niche including stromal cells, as well as many populations of myeloid and lymphoid cells ( Fig. 2 A ). Monocytes and macrophages of various phenotypes comprise approximately half of the cells within the IN, consistent with literature examining chronic stages of the foreign body response (FBR) to materials ( 19 , 35 – 37 ). Interestingly, we observe an increase in CD8 T cells and B cells within the IN of diseased animals that increases as disease progresses, suggesting the complex inflammatory dynamics in EAE are reflected within the IN ( Fig. 2 B ). Because we saw differential recruitment of immune cells in disease and chemokines are known to regulate recruitment, we examined chemokine expression but few changes in mRNA were observed. To further characterize the IN and confirm this finding, we used two bead-based multiplex immunoassays (LegendPlex) to quantify various cytokines and chemokines in the INs at day 9 ( SI Appendix , Table S1 ). CXCL9 and IL-6 were both significantly reduced in the diseased condition. IFNγ was below the limit of detection for the kit [consistent with other work ( 38 )].

Schematic illustrating experimental workflow for scRNAseq studies in chronic progressive EAE. Mice were implanted with INs (on day-14) and two weeks later received an adoptive transfer of T cells reactive against MOG peptide (EAE) or OVA peptide (Control). Implants were harvested 7 to 9 d after transferring T cells and analyzed via scRNAseq. Mouse cartoon from Servier and sequencer from DBCLS, both obtained from bioicons.com.

scRNAseq illustrates that INs contain various cell types and cell–cell communication that are dynamic with disease state. IN in mice is composed of a complex population of both immune and stromal cells ( A ), and these populations are altered with both time and disease in EAE models compared with OVA controls ( B ). The cells within these INs communicate with each other, but this cell–cell interaction is altered by disease as analyzed via CellChat ( C – F ). ICAM-1 signaling at d9 demonstrates reduced signal from both CD11b+ DC and Tregs in disease [MOG] ( D ) when compared to healthy [OVA] ( C ). Similarly, signaling via Itgb2 occurs in various cells ( E and F ), but this signaling is reduced from both CD11b + DC and Tregs in disease ( F ).

Because few changes occurred in quantification of cytokines and chemokines, we sought to conduct a deeper analysis of receptor–ligand interactions with the scRNAseq data. CellChat ( 28 ) was used to examine cell–cell communication within the IN, and we observed that within the complex environment of the IN, an array of communication pathways were present ( SI Appendix , Fig. S1 ). Pathways known to be important in EAE, such as signaling through ICAM1 and LFA-1 are altered in disease when compared with INs from otherwise healthy mice that received control T cells (OVA) ( 39 , 40 ). Within the IN, a large variety of cells interact with each other via ICAM-1 and its receptors LFA-1, CD11c, and CD43 ( Fig. 2 C and E and SI Appendix , Fig. S2 and S3 ). In EAE, communication with Tregs via ICAM-1 diminishes within the niche, highlighting reduced activity of this regulatory component during the development of autoimmunity ( Fig. 2 D and F ). Furthermore, communication with Treg via Treg-produced laminin declines in the IN in disease, but interactions with stromal cell produced laminin remain ( SI Appendix , Figs. S1 C and D and S4 ).

In investigating innate cell signaling, we observed significant alterations in chemokine signaling in disease (compared with OVA control). CC chemokines expressed by myeloid cells and interacting with receptors on both myeloid and lymphoid cells were robustly increased in EAE at both day 7 (d7) and day 9 (d9) ( Fig. 3 A and B ). CCL2 communication via its receptor CCR2 is increased from monocytes on d7 and inflammatory monocytes on d9 to a variety of target cell types. Additionally, CCL5 communication via CCR1 and CCR5 is increased from ferritin+ myeloid cells (d7) and CD4 T cells (d9). CCL4 communication via CCR5 is increased in dendritic cell populations (d9). Finally, communication via CCL6 is increased in several myeloid cells throughout disease. Overall, the INs indicated a variety of CC chemokine signaling networks were dysregulated in EAE when compared with OVA controls, and importantly, they did so without requiring the biopsy or harvest of a vital organ. A comparison with previously published scRNAseq data from spinal cords of immunized mice with EAE also demonstrates CC chemokine dysregulation ( SI Appendix , Fig. S7 ).

scRNAseq of the immunological niche (IN) shows increased CC chemokine signaling in diseased mice (when compared with OVA controls) at both d7 ( A ) and d9 ( B ). Producers of these chemokines are predominantly myeloid cells. At both time points several CC chemokine–receptor pairs are exhibiting increased signaling. All pathways shown are significantly increased in disease (compared with OVA controls) as determined via CellChat ( P < 0.05).

Development of NP Therapeutic Targeting CC Chemokines.

We next investigated polymeric NPs that were previously reported to alter inflammatory dynamics and inhibit expression of a key CC chemokine, CCL2 ( 18 ), for their ability to attenuate the immune dysfunction. Although prior work demonstrated that these NPs inhibited CCL2 signaling, these studies did not examine their effects on other CC chemokines, but we hypothesized that the effects of the NPs may extend beyond CCL2. Furthermore, this platform has been used for antigen delivery to induce antigen-specific tolerance in autoimmunity ( 22 , 41 , 42 ). PLG NPs encapsulating antigen (MOG or OVA) were prepared using double-emulsion solvent evaporation. NP properties between MOG and OVA-loaded were similar and consistent with previous work, with diameters of approximately 400 nm, a highly negative zeta potential, and antigen loading of approximately 7 µg/mg ( Fig. 4 A – C ). When suspended in PBS in vitro, NPs released peptide over 6 d. The efficacy of autoantigen-loaded NPs in reducing aberrant CC chemokine communication was investigated in vitro by delivery of MOG-loaded NPs to the culture media of macrophages (BMDM and RAW 264.7). We confirmed that antigen-loaded particles reduced CCL2 expression in RAW cells ( Fig. 4 E ), consistent with prior reports using unloaded particles ( 18 ). We also observed decreased expression of CCL4 and CCL5 when BMDM or RAW cells were treated with NPs encapsulating MOG ( Fig. 4 E and F ) and decreased expression of CCL5 when BMDCs were treated with NP ( SI Appendix , Fig. S5 ).

MOG-loaded PLG NPs inhibit expression of CC chemokines in vitro. MOG ( A ) and OVA ( B ) loaded NPs have diameters of approximately 400 nm (Scale bar, 1 µm) and Zeta potential <−50 mV ( C ). In vitro MOG released into PBS over six days ( D ). Raw 264.7 cells treated with 300 µg/mL MOG-loaded NPs before stimulation with LPS demonstrated decreased expression of CCL2, CCL4, and CCL5 compared to cells stimulated with LPS, but not treated with NPs ( E ). Similarly, bone marrow–derived macrophages (BMDM) exhibited decreased expression of CCL4 and CCL5 when treated with NPs ( F ). Data are represented as the mean ± SEM. Each point represents an individual biological replicate (n = 6). Student’s t test was used to compare LPS with LPS + NP for each protein (n = 6). * P < 0.05, *** P < 0.0005, and **** P < 0.0001.

The biodistribution of the MOG-loaded particles was investigated in healthy C57BL6/J mice. Prior work by our group and others has demonstrated that NPs in the size range produced here typically accumulate in the reticuloendothelial system ( 41 , 43 – 45 ). The MOG-loaded particles were labeled with Cy5.5 and injected i.v. via the tail vein. Necropsy was performed after 24 h, and distribution quantified via an IVIS. Our findings were similar to previous work with PLG NPs predominantly accumulating in the liver, spleen, and kidney with less accumulation in the lymph nodes, lungs, and blood ( Fig. 5 ).

Biodistribution indicates NPs predominantly accumulate in the liver, spleen, and kidney. Fluorescent NPs (or saline control) were injected intravenously and necropsy was performed 24 h later. Fluorescence imaging (shown as radiant efficiency) performed with an IVIS demonstrated clear accumulation in the spleen, kidney, and liver ( A ). When quantified, the spleen, kidney, and liver showed large fold increases in fluorescence (radiant efficiency) with particle treatment compared to control ( B ). Data are represented as the mean ± SEM. Each point represents an individual biological replicate (n = 5).

We next sought to test the efficacy of the MOG-loaded CC targeting PLG NPs in EAE. EAE was induced with an adoptive transfer of T cells reactive to MOG on day 0. To investigate the timing of therapeutic delivery, 2.5 mg of particles loaded with MOG were administered two days after induction of disease (presymptomatic), on d7 (early symptomatic), or on d9 (peak disease), and OVA-loaded particles were delivered on day 2 as a control. Comparing clinical scores of the mice, MOG-loaded NPs reduce disease burden and earlier administration was associated with improved outcomes ( Fig. 6 A and B and SI Appendix , Fig. S6 ). In fact, for the presymptomatic delivery of NPs (d2), no symptoms were observed. Importantly, administration of NPs after symptomatic onset (d7) reduced disease burden to less than half of mice receiving control treatments.

MOG particles effectively treat EAE, and earlier treatment yields improved outcomes. Mice received IV injections of PLG(MOG) on day 2, 7, or 9 or PLG(OVA) (day 2; negative control) after adoptive transfer of T cells to induce EAE. Mice were scored daily to evaluate EAE symptoms ( A ), and the cumulative disease score was also calculated ( B ). Overall, PLG(MOG) reduced EAE scores when administered on day 2 or day 7 and cumulative disease burden was reduced for particles administered on day 2. Recall studies were conducted after the terminal time point, with splenocytes from each group. Splenocytes were cultured with either OVA (negative control) or MOG (recall) and cytokines produced were measured via the ELISA ( C – F ). Statistical analysis for disease scoring was determined by the Kruskal–Wallis test (one‐way ANOVA nonparametric test) compared to control [PLG(OVA)]. n = 7 to 8. * P < 0.05, ** P < 0.005, *** P < 0.0005, and **** P < 0.0001.

The molecular basis for improved outcomes in NP-treated mice was investigated by performing a recall assay to either MOG or OVA peptide on harvested splenocytes. The ELISA of the cell culture supernatant indicated reductions in key cytokines involved in EAE for the NP-treated groups responding to MOG peptide, including IFNγ, IL-2, IL-10, and IL-17 ( Fig. 6 C – F ). Splenocytes responding to OVA peptide had low cytokine expression regardless of treatment condition. Surprisingly, for the treatment time point that performed best (presymptomatic) IFNγ was unchanged compared to the control treatment, but it was changed for mice treated on d7 or d9 ( Fig. 6 C ).

Monitoring immune cell activity in vivo enabled further confirmation of reduced disease activity. We induced EAE in a cohort of mice to which NPs were administered on d2. On day 12 after disease induction, mice received i.p. injections of luminol, and bioluminescence of the CNS was measured with IVIS imaging ( Fig. 7 ). Injected luminol interacts with neutrophil myeloperoxidase (MPO) and produces a bioluminescent signal that is indicative of acute inflammation ( 46 ). Quantification of IVIS imaging showed a decrease in luminescence in the MOG-treated mice suggesting reduced inflammatory activity in the CNS of these mice. Spleens harvested from mice were analyzed for Tregs (CD4 + CD25 + FoxP3 + ) and anergic T cells (FR4 hi CD73 hi CD25 - FoxP3 - ). Percentages of both Tregs and anergic T cells were increased in particle-treated mice.

NPs increase Treg and anergic T cells and reduce neuroinflammation. Mice were treated 2 d after adoptive transfer with MOG-reactive T cells with either PLG(OVA) or PLG(MOG) and spleens were harvested (d12) for flow cytometry which indicated an increase in CD25 + FoxP3 + Treg cells ( A ) and a concomitant increase in FR4 hi CD73 hi CD25 - FoxP3 − anergic T cells ( B ) in the PLG(MOG)-treated group. Additionally, on d12 mice were given an intraperitoneal injection of luminol and imaged via an IVIS for MPO activity ( C ). Quantification indicated increased inflammation in PLG(OVA) compared to PLG(MOG) treated mice ( D ). Data are presented as mean ± SEM. Each point represents an individual biological replicate (n = 7). Student’s t test was used for statistical analysis. * P < 0.05 and ** P < 0.005.

Discussion

PPMS exhibits rapidly progressing symptoms and associated disability, presenting a major clinical challenge with limited treatment options. The gold standard treatment, B cell–depleting antibodies, only partially (and modestly) slows disability progression, does not reverse existing disability, predisposes patients to opportunistic infection, and blunts vaccine responses. A deeper understanding of the interplay between multiple cell types within disease tissue throughout progression could identify targets for treatment. We have previously demonstrated that subcutaneously implanted porous poly(caprolactone) (PCL) scaffolds that we termed INs integrate with the surrounding tissue, vascularize, and exhibit a dynamic invasion of immune cells ( 8 , 9 , 19 ). The porous nature of these implants and associated surface area drives a FBR with large quantities of cells from which samples can readily be isolated for monitoring. Examining differential gene expression within these implants had diagnostic and prognostic utility sufficient for informing treatment decisions. In this report, we identified potential targets and developed antigen-specific treatments for a mouse model of progressive EAE.

In this work, single-cell RNA sequencing analysis identified the complex cellular dynamics within the IN that reflect disease progression. The IN consists of multiple cell populations, including immune and stromal cells ( Fig. 2 and SI Appendix , Fig. S1 ). Cell populations within the IN are dynamic, and the composition changes as disease develops, including increases in B cells and CD8 T cells. A growing body of literature has used immunoassay arrays to quantify chemokines in the FBR to implanted materials ( 47 , 48 ), inspiring us to measure cytokines and chemokines using a bead-based immunoassay at day 9. Alterations in total proteins with disease state at the IN were subtle and unlikely to explain variations in cell recruitment during disease. This motivated a further examination of receptor–ligand pairs in the scRNAseq data using CellChat.

As expected, given the complex cellular composition of the IN, the cell–cell communication within the IN is varied, interconnected, and changes with disease state. Interestingly, communication via ICAM-1 and its receptors LFA-1, CD11c, and CD43 are reduced in the IN during disease. In particular, regulatory T cell communication via these molecules shows a dramatic reduction. Previous work indicates the importance of this pathway in regulatory T cell–suppressive functions, including more severe EAE in LFA-1 null animals ( 39 , 40 , 49 ). Of the many pathways altered in disease, CC chemokine signaling shows robust changes within the IN ( Fig. 3 and SI Appendix , Figs. S5 and S6 ). CC chemokines including CCL2, CCL4, CCL5, and CCL6 and their cognate receptors demonstrate increased signaling in INs in mice with EAE when compared with OVA controls. The discrepancy between cytokine signaling analysis and bulk measurement of cytokine expression is expected as cytokine and chemokine signaling is highly dynamic and simply measuring the total milieu does not capture granular details involved in cell–cell communication.

The CC chemokine pathways have been known to be involved in the FBR for at least a decade ( 50 – 53 ). CC chemokines have diverse functions in both EAE and MS. CCL2 and its receptor CCR2 are highly involved in EAE and are expressed at high levels in lesional brain tissue in MS ( 54 – 56 ). Similarly, CCL5 is overexpressed in MS and EAE, and polymorphisms that lead to reduced CCL5 expression also lead to reduced severity of MS symptoms ( 57 – 60 ). Although less well studied, CCL4 is overexpressed in EAE and MS ( 61 ). The cognate receptors for these ligands CCR2 and CCR5 are also overexpressed in MS and EAE ( 62 , 63 ). To date, therapies targeting elements of these pathways have been successfully tested in preclinical studies but have not shown sufficient efficacy in MS ( 64 , 65 ). A comparison with previously published scRNAseq data confirmed the presence of CC chemokine signaling in the diseased CNS ( SI Appendix , Fig. S7 ). However, the total immune cell population in the healthy CNS is low, making a differential analysis of the diseased:healthy CNS impossible and further highlighting the utility of the IN in measuring dysregulated immune signaling specific to disease. Collectively, the IN highlighted both the dysregulated nature of CC chemokine signaling in EAE and lack of adequate regulatory T cell function, which led us to the hypothesis that antigen-specific therapies that simultaneously target multiple CC chemokines and promote Tregs may yield a highly effective treatment.

We sought to develop a NP-based therapeutic that targets multiple CC chemokines and simultaneously delivers antigens to promote tolerogenic responses in chronic progressive EAE. Antigen-loading in polymeric NPs has been documented to induce tolerogenic responses in autoimmunity ( 23 , 31 , 66 – 72 ). In fact, several of these approaches are demonstrating efficacy in human clinical trials for celiac disease and anti-drug immunity in treating gout ( 31 , 73 ). Some polymeric NP formulations reduce CCL2 expression in macrophages and dendritic cells, but the NP effects on other CC chemokines have not been investigated ( 18 ). We hypothesized that polymeric nanoparticles that mitigate dysregulated CCL2 expression would also impact other CC chemokines (CCL4 and CCL5). Because the IN demonstrated broadly dysregulated CC chemokine signaling, we hypothesized that the NPs could be loaded with antigen and target immune dysfunction via multiple mechanisms to reduce disease burden with antigen-specific therapies in EAE. We fabricate PLG NPs encapsulating MOG peptide that release antigen in a sustained manner. Antigen-loaded NPs reduce expression of CC chemokines in vitro in BMDM, RAW 264.7 cells, and BMDCs ( Fig. 4 and SI Appendix , Fig. S8 ). RAW 264.7 cells are a convenient cell line for initial testing but are not equivalent to primary cells necessitating the need to study both ( 74 ). Overall, the in vitro studies are helpful for evaluating NP efficacy at targeting CC chemokine dysregulation, but species considerations are important limitations including possible effects of xenogeneic serum on the cells ( 75 ) (controlled for by using FBS in all conditions tested) and an understanding that murine and human immunology are not equivalent including differential expression of receptors and coreceptors (e.g. Fc receptors and CD4) ( 76 , 77 ). Thus, the findings in this work studying approaches to monitoring and treating murine autoimmune models may not necessarily directly translate to humans; yet this strategy could potentially be performed in a clinical study.

A single NP dose infused intravenously effectively treats EAE, with earlier administration associated with reduced disease burden ( Fig. 6 and SI Appendix , Fig. S9 ). Similar results have been observed with antigen encapsulating NPs in other systems ( 23 , 68 , 78 , 79 ). For PPMS, presymptomatic administration (particles delivered after transfer of encephalitogenic T cells, but before symptoms) completely abrogates disease progression. Yet even after onset of clinical disease, a single dose of particles can reduce disease burden to nearly that of naïve mice, after mice recover from initial inflammation. This response reflects a realistic clinical use case in that the animals exhibit clear symptoms at the time of administration. In addition to reducing symptomatic disease, BLI demonstrates reduced inflammatory activity in the CNS of NP-treated mice ( Fig. 7 ). The improved efficacy with early treatment motivates the need to develop strategies for early detection ( 72 , 80 , 81 ).

To interrogate effects of NPs on T cell functions in vivo, we examined splenic CD4 T cell populations after particle administration. Administering MOG-loaded NPs results in significantly more Tregs and anergic T cells ( Fig. 7 A and B ). Additionally, splenic recall assays from mice treated at each time point demonstrate significant reductions in inflammatory cytokines from cells isolated from mice treated with NPs ( Fig. 6 D – F ). IFNγ expression in recall assays was significantly reduced with later treatment timepoints, but not in the earliest case (most effective treatment). This result is consistent with literature highlighting the dual role and stage dependence of IFNγ in EAE and MS ( 82 – 87 ). Together, these findings suggest that targeting of autoreactive T cells may be a possible mechanism for the observation that MOG-loaded NPs ameliorate disease in the chronic progressive EAE model. These results are consistent with prior demonstrations that antigen-loaded NPs induce anergic and Tregs in an antigen-specific manner to mitigate autoimmunity ( 70 , 81 , 88 , 89 ).

In conclusion, by harnessing a subcutaneously implantable biomaterial in EAE in conjunction with scRNAseq, we identify specific cell–cell communication that we then target with a polymeric NP. These NPs target CC chemokine signaling while also administering an antigen-specific immunotherapy, leading to overall reductions in disease burden. Importantly, NPs dramatically reduce disease burden with only a single dose and are functional therapeutically. The NP system contains only polymeric carrier and the antigen of interest, but no other active ingredient (i.e. drug). Ultimately, this work demonstrates the potential utility of these polymeric systems for treating autoimmunity and illustrates a proof of concept of the biomaterial-based IN as a drug discovery platform.

Supplementary Material

Appendix 01 (PDF)

Acknowledgments

We would like to thank the UM Precision Health Scholars Grant for funding this work. This work was supported in part by NIH Pathway to Independence Award R00EB028840 to A.H.M and 1R01AI155678, 1R01CA243916, and R01CA272940 to L.D.S. This work was also supported in part by a Grant from University of Michigan BioInterfaces. S.N.W. is supported by an NIH T32 Grant TDE007057. We would like to thank the University of Michigan Flow Cytometry and Advanced Genomics Cores for flow cytometry and genomics resources, respectively. We would like to thank MC 2 for SEM resources and the UM Immune Monitoring Shared Resource for running the ELISAs.

Author contributions

L.M.R., K.R.H., S.N.W., J.T.D., L.D.S., and A.H.M. designed research; L.M.R., K.R.H., S.N.W., J.T.D., S.M.O., A.G., J.T., H.T., R.U., and A.H.M. performed research; K.R.H. and A.H.M. contributed new reagents/analytic tools; L.M.R., K.R.H., S.N.W., J.T.D., K.V.G., D.N.I., and A.H.M. analyzed data; and L.M.R., S.N.W., D.N.I., L.D.S., and A.H.M. wrote the paper.

Competing interests

This manuscript uses NPs for which L.D. Shea has an interest in Cour Pharmaceutical, who have licensed the technology. A patent application has been filed on the engineered niche idea (US17/286,154).

Footnotes

This article is a PNAS Direct Submission.

Contributor Information

Lonnie D. Shea, Email: ldshea@umich.edu.

Aaron H. Morris, Email: aharmorr@umich.edu.

Data, Materials, and Software Availability

Single-cell RNA sequencing data of the biomaterial-based immunological niches for this study have been deposited in the Gene Expression Omnibus (GEO) database (accession no. GSE287968 ) ( 26 ). Previously published single-cell RNA sequencing data of spinal cords ( 30 ) was accessed from GEO (accession no. GSE200901 ) ( 29 ).

Supporting Information

References

Associated Data

Supplementary Materials

Appendix 01 (PDF)

Data Availability Statement

Single-cell RNA sequencing data of the biomaterial-based immunological niches for this study have been deposited in the Gene Expression Omnibus (GEO) database (accession no. GSE287968 ) ( 26 ). Previously published single-cell RNA sequencing data of spinal cords ( 30 ) was accessed from GEO (accession no. GSE200901 ) ( 29 ).