Human immune cells infiltrate the lesioned spinal cord and impair recovery after spinal cord injury in humanized mice

Immune compromised mice require ~4 months of engraftment with human umbilical cord blood CD34+ stem cells to develop a full and functional human immune system The human neuroinflammatory response elicited after spinal cord injury in humanized mice is limited at 2 months post-engraftment but matures by 4 months Intraspinal neuroinflammation consists of a florid human T cell and macrophage response, and human T cells co-localize with human macrophages A human intraspinal neuroinflammatory response exacerbates lesion pathology and impairs functional recovery Abstract Humanized mice are a useful tool to help better understand how the human immune system responds to central nervous system (CNS) injury. However, the optimal parameters for using humanized mice in preclinical CNS injury models have not been established. Here, we show that it takes 3-4 months after engraftment of neonatal immune compromised mice with human umbilical cord stem cells to generate a robust human immune system. Indeed, sub-optimal human immune cell responses occurred when humanized mice received spinal contusion injuries at 2 months vs. 4 months post-engraftment. Human T cells directly contact human macrophages within the spinal cord lesion of these mice and the development of a mature human immune system was associated with worse lesion pathology and neurological recovery. Together, data in this report establish an optimal experimental framework for using humanized mice to help translate promising preclinical therapies for CNS injury.

Immunocompromised mice engrafted with human immune systems (i.e. "humanized" mice) are powerful pre-clinical 40 models for studying human immune cell function. However, the translational value of these mice has not been fully 41 realized, particularly in preclinical models of neuroinflammation and central nervous system (CNS) injury. Previously, 42 we documented the feasibility of using humanized mice to study systemic and neuroinflammatory changes caused by blood (UCB) stem cells to engraft bone marrow and secondary lymphoid tissues (e.g., spleen, lymph nodes) of 59 IL2 (rhIL2), human T cells increased expression of hCD69 ( Fig. 2A,B), a cell activation marker, accompanied by 120 robust proliferation (Fig. 2C,D; Fig. 2-figure supplement 1C) and production of human IFNg and IL-10 ( Fig. 2E,F). 121 122 When stimulated with a hCD40 activating antibody (clone 5C3) and rhIL4, purified human B cells increased hCD69 123 expression ( Fig. 2A,B) but did not proliferate or produce cytokines ( Fig. 2C-F). However, stimulation of human T cells 124 with hCD3/28 and rhIL2 induced robust hCD19 + B cell proliferation (Fig. 2C), suggesting human T and B cell 125 interactions are necessary for mediating human B cell function. 126 127 Lipopolysaccharide (LPS), a canonical activator of toll-like receptor 4 (TLR4) found mostly on myeloid cells, also 128 increased proliferation and production of human cytokines by human splenocytes (Fig. 2C-F). Similarly, LPS injected 129 in vivo (3mg/kg, i.p.) elicited production of human TNFα (hTNFα) by 1-hour post-injection (Fig. 2G). Human TNFα was 130 not detected in serum from naive hNSG mice or non-humanized NSG mice injected with LPS. We also detected 131 human IgG and IgM antibodies in blood serum of hNSG mice but not non-humanized NSG mice (Fig. 2H,I). Together,132 these data prove that human immune cells from hNSG mice are functional; they respond ex vivo and in vivo to 133 physiologically-relevant stimuli, producing a range of immune effector molecules (e.g., cytokines, antibodies). post-engraftment in humanized mice. Thus, one would predict that the effects of SCI on immune system activation and 140 the subsequent effects of neuroinflammation and lesion histopathology will change as a function of time post-141 engraftment. To test this hypothesis, a single human UCB donor was used to generate two cohorts of hNSG mice born 142 2 months apart. Both groups received a SCI at the same time -one at 2 months post-engraftment and the other at 4 143 months post-engraftment. Both cohorts survived to 35 days post-injury (dpi). Consistent with data from Fig. 1, 144 Figure 1. Long-lived engraftment of human immune systems in NSG mice occurs without toxicity. Human immune systems were derived from hCD34 + umbilical cord blood stem cells. A) Engraftment efficiency in mice is affected by time post-engraftment and source of stem cells. UCB = umbilical cord blood, BM = bone marrow. Data average ± SD; *p<0.05 ***p<0.001 compared to 2 months post-engraftment (UCB CD34 + only), oneway ANOVA with Tukey's multiple comparison test. B) Proportion of human CD45 + peripheral blood leukocyte (PBL) subsets. C) Human CD45+ cells in the blood (BLD), bone marrow (BM), and spleen (SPL) of hNSG mice 4-6 months post-engraftment. D,E) Human immune subsets (hCD3, hCD19, and hCD33) in blood (D) and spleen (E) of hNSG mice 4-6 months post-engraftment. F,G) Human CD3 + , CD4 + , and CD8 + T cells identified in thymus (F) and spleen (G) of naïve hNSG mice 4 months post-engraftment. Human T cell subsets in thymus are highlighted (F). uninjured hNSG mice had more human PBLs at 4 months post-engraftment than at 2 months ( Fig. 3A; Fig. 3- figure  145 supplement 1A) -all circulating human leukocyte subsets including T lymphocytes (hCD3), B lymphocytes (hCD19) 146 and myeloid cells (hCD33) increased in number and proportion as a function of time post-engraftment ( Fig. 3B; Fig. 3-147 figure supplement 1B).

149
Five weeks after SCI we noted marked differences in circulating leukocyte responses between the two cohorts. In the 150 2-month cohort, the total numbers of circulating human leukocytes increased after SCI (Fig. 3E,F; Fig. 3- figure  151 supplement 1E,F). However, this was unlikely an injury-dependent effect and is better explained by continued 152 maturation of the human immune system. Indeed, by 35 dpi cells from mice in the 2-month cohort were now engrafted 153 >3 months with the relative proportion of hCD45+ PBLs in most mice increased to levels identical to engraftment at 3 154 months (compare Fig. 1A,B) and similar to pre-injury values found in the 4-month post-engraftment cohort (compare 155 were modest in the 4-month post-engraftment cohort (Fig. 4E) and were marked by a selective increase in the 157 proportion of T cells ( Fig. 3F; Fig. 3-figure supplement 1F).

159
Given the importance of the spleen in regulating neuro-immune interactions after SCI (Blomster et al., 2013; Noble et  160 al., 2018), we also quantified human splenocyte subsets in both cohorts 35 dpi (Fig. 4A,B). The relative proportions of  161 human T, B, and myeloid cells in spleen reflected their proportions in the blood 35 dpi, which is consistent with the role 162 of the spleen as a reservoir for circulating human leukocytes. Spleens of both groups displayed preferential clustering 163 of hCD3 + T cells in the splenic white pulp. Again, the relative density of T cell clusters was increased in the 4-month 164 post-engraftment cohort (Fig. 4C). 165 166 Figure 2. Human innate and adaptive immune cells from hNSG mice are functional and respond to cell-specific stimulation. A) Human splenocytes upregulate cell surface expression of activation marker CD69 48 hours after stimulation with human CD3/28 antibody and rhIL2. B) Proportion of hCD4 + and hCD8 + T cells expressing CD69 48 hours after stimulation by hCD3/28 and rhIL2. C) Decrease in CFSE staining demonstrating robust proliferation of human splenocytes stimulated with hCD3/28 and rhIL2. D) Proportion of proliferating splenocytes 96 hours after cell specific stimulation. E,F) Quantification of human interferon gamma (IFNg) and IL10 in culture supernatants after 96 hours of cell specific stimulation. G) Human TNFa quantification in blood serum 1 hour after in vivo injection with 3 mg/kg lipopolysaccharide (LPS). Human IgG (H) and IgM (I) from blood serum in hNSG mice. Note the absence of human cytokines and antibodies in blood serum of non-engrafted NSG mice treated with LPS, demonstrating species specificity of ELISAs. ND=not detected. Data average ± SEM; n=2 biological replicates in B,D; n=4 biological replicates in E,F; n=3 mice per group in G,H; n=3 NSG and n=6 hNSG mice in I,J. leukocytes was increased in SCI mice from the 4-181 month cohort (Fig. 5A,B). Additionally, hCD3 + T 182 cells were increased ~10-fold in mice injured at 4 183 months post-engraftment compared to 2 months 184 post-engraftment (Fig. 5C,D).
Multi-color immunofluorescent confocal microscopy 187 identified hCD3 + /hCD4 + helper and hCD3 + /hCD8 + 188 cytotoxic T cell subsets within spinal cord lesions 189 (Fig. 6A). We never observed colocalization of 190 hCD4 + and hCD8 + labeling on any single hCD3 T 191 cell. In line with blood and spleen data in Figs. 4-5, 192 more human T cell subsets were consistently found 193 in the injured spinal cord of hNSG mice from the 4-194 month cohort (Fig. 6B)   injury or repair has not been possible in people with SCI. 237 238 Previously, we found that lesion pathology was exacerbated and functional recovery was impaired in hNSG mice as 239 compared to non-engrafted immunocompromised NSG control mice (Carpenter et al., 2015). However, our previous 240 study occurred before we understood the importance of post-engraftment timing on maturation of the human immune 241 system hNSG mice. Here, we compared lesion pathology and functional recovery in non-engrafted NSG mice and 242 age-matched hNSG mice with a mature (4-month post-engraftment) human immune system after spinal contusion 243 . Although the data in that report were the first to 252 illustrate the feasibility of using humanized mice to test hypotheses related to neuro-immune interactions after SCI, the 253 report was limited in scope and did not evaluate the implications of developmental timing on human immune cell 254 composition and function. 255 Data in this report show the relative efficiency of human chimerism and corresponding changes in human immune cell 256 composition change as a function of time post-engraftment. Human chimerism peaks between 3-4 months post-257 engraftment and remains stable for at least 12 months. Of the human immune cells that we characterized, T 258 lymphocytes are likely key to the development of a mature human immune system in hNSG mice and may ultimately 259 play a key role in influencing post-injury neuroinflammation, pathology and recovery of function. 260 261 Figure 6. Human helper and cytotoxic T cell subsets infiltrate the injured spinal cord, with T helper subsets directly contacting human macrophages. A) Immunofluorescent labeling of hCD3 + /hCD4 + helper and hCD3 + /hCD8 + cytotoxic T cells at the lesion epicenter 35 dpi in 2-and 4month post-engraftment hNSG mice. MIPAR automated quantification of total numbers (B) and relative proportion (C) of human T cell subsets. D) Immunolabeling of hCD4 identifies both small and large cells with morphology of T cells and macrophages, respectively. Small hCD4 + T cells were often found directly adjacent to large, phagocytic human macrophages, indicating T cell-macrophage interaction within the injured spinal cord (arrowheads). E) Confocal imaging of hCD3 + /CD4 + T cells directly contacting a hCD3 -/hCD4 + macrophage. *p<0.05 ****p<0.0001 student's t-test. Data average ± SEM.
In humanized mice, human T cells help promote the maturation and functional development of human B cells (Lang et  262 al., 2013). In athymic nude rats, which lack T cells, B cells are similarly impaired but can be rescued by reconstituting 263 adult rats with T cells (Milićević et al., 2005). More human macrophages were also found in spinal cord lesions in the 4-month cohort (as compared with lesions 287 from hNSG mice in the 2-month cohort). A more robust myeloid cell response to SCI was observed even though the 288 percentage of human myeloid cells in blood or spleen of hNSG mice was not different between mice in the 2-and 4-289 month post-engraftment cohorts. Again, this is evidence that a more functionally mature human immune system exists 290

It is not clear how T cells influence the development and maturation of 264 other immune cells but interactions between T cells and progenitor cells in
in hNSG mice at 4 months post-engraftment. 291 292 We also found that many large, phagocytic human macrophages expressed membrane hCD4. Rat CD4 + macrophages 293 are  is routinely used to phenotype helper T-cells, is also expressed by human and rat monocytes (Crocker et al., 1987; 297 Mestas and Hughes, 2004). Ligating this co-receptor on monocytes/macrophages triggers intracellular signaling 298 cascades that augment macrophage maturation and activation (Szabo et al., 1990, Zhen et al., 2014 injury . Therefore, humanized mice could prove to be a useful tool for studying the functional 303 importance of activating CD4 (and possibly CD8) co-receptors on CNS macrophages. 304 305 Using confocal microscopy, we confirmed several examples of cell-cell contact between human T cells and human 306 macrophages in the injured spinal cord. Membrane hCD3 expression was abundant at the interface between human 307 macrophages and human T cells (see Fig. 6-figure supplement 2). Redistribution of CD3, a major component of the 308 T In the present report, data suggest that the intraspinal human immune response contributes to pathology and 316 subsequent functional impairment. Indeed, after SCI in 4-month post-engraftment hNSG mice with mature functional 317 human immune systems, a prominent intraspinal human T cell infiltrate was associated with exacerbated lesion 318 pathology and impaired spontaneous recovery. Still, we previously provided evidence that axons migrate along lesion 319 matrix proteins in close proximity to human immune cells in hNSG mice, indicating possible reparative functions of 320 human immune cells. Together, these data demonstrate that in hNSG mice, the human immune system actively 321 responds to SCI hNSG mice and that these cells exert biological effects on the lesion epicenter. To prove a causal role 322 for human T cell-mediated pathology or repair, human immune cell manipulation studies are needed. 323 324 Considerations for using humanized mice in CNS injury models 325 326 Long-term health and viability of humanized mice may be a concern for those interested in brain or spinal cord aging 327 studies or if chronic survival times are required. We were consistently able to keep hNSG mice alive for 9-12 months of 328 age without major health concerns. Sporadic premature death of hNSG mice does occur but cause of death is difficult 329 to determine. Anemia may be one factor contributing to premature mortality problems (e.g., ruffled fur, cataracts in one or both eyes and spontaneous hindlimb inflammation). Importantly, SCI did 336 not increase the number or severity of these health issues. We did note that hNSG mice appear healthier when they 337 are engrafted with human UCB stem cells close to 72 hours after birth. Although there are various types of humanized mouse models, including several that include "next generation" 354 strategies for manipulating the composition and function of the human immune system in mice (see Shultz  Moreover, hNSG mice are available through commercial vendors, although current costs may be prohibitive for larger 358 studies (~$1,000 USD/mouse). Fortunately, it is possible to reduce costs and generate hNSG mice "in house". In our 359 experience, it is possible to generate a litter of 8-10 hNSG mice for ~$1,500 cord and skin treated with a sequence of betadine-70% ethanol-betadine. A small midline incision was made to expose 413 the mid-thoracic vertebra, then a partial laminectomy was performed. A 60 KDyne (mild) T9 contusion injury was 414 performed using an Infinite Horizon Impactor (Precision Systems and Instrumentation, Lexington, KY). Immediately 415 after impact, the muscle overlaying the injury site was sutured, followed by closure of the wound with sutures or 416 staples. After surgery, mice were placed in cages on heating pads and monitored until they recovered consciousness. 417 Animals were supplemented with sterile saline (1-2mL, s.q.) and softened food to eat ad libitum during recovery. 418 Bladders were expressed twice daily, and urine underwent weekly pH testing to detect bladder infections. Gentocin 419 antibiotic was subcutaneously administered once a day at 5 mg/kg for 5 days post-injury (dpi). 420 421 Hindlimb locomotor function assessment The open-field Basso Mouse Scale (BMS) was used to assess hindlimb paralysis and functional recovery (Basso et  424 al., 2006). Pre-SCI testing occurred after acclimating mice to the open field on multiple days with dim lighting to  425  minimize anxiety. Post-SCI testing occurred on 1, 3, 5, 7, 14, 21, 28, and 35 dpi. Mice explored the open field and  426 were scored over a period of 4 minutes. Two individuals that were blinded to the experimental condition averaged the 427 left and right hindlimb score for each mouse to obtain a single score per mouse at each time point. 428 429 The horizontal ladder test was performed before SCI and at 35 dpi. The horizontal ladder contained 33 rungs spaced 430 ~1 cm apart and elevated ~1 cm off of a clear mirrored surface for viewing of paws from the side and below using a 431 digital video recorder to capture each trial. Mice were acclimated to the ladder test over 3 separate days with a 432 minimum of 3 runs per session before acquiring baseline values. Mice were handled by a single experimenter (RSC). 433 During testing days, each mouse was allowed a minimum of 3 passes along the horizontal ladder to reduce trial-trial 434 variability. Mice were allowed to rest for ~15-20 minutes between each pass/trial. Digital videos were assessed by a 435 single experimenter (RRJ) blinded to group designations, with playback at 0.35x speed using VLC media player (v.3.0; 436 VideoLAN Organization) on a computer screen of a minimum 13-inch diameter. Each trial (3/mouse) was viewed and 437 analyzed 3 times by the experimenter and averaged to acquire a single value for each trial. This was done to reduce 438 observer error and variability, and a subset of trials were assessed by a second reviewer (RSC) to ensure accuracy. 439 440 Tissue collection 441 442 For survival blood sampling, blood was collected from the submandibular vein using a lancet and EDTA- spleen were obtained by standard hemocytometer counting techniques, while total blood was analyzed with a 452 Hemavet 950fs multi-species hematology system capable of analyzing whole blood cells with 5-part differential, 453 platelets, and red blood cells. To obtain serum, blood was collected as above, inserted into gel separation tubes 454 coated with clotting activator, and centrifuged at 10 5 x g for 5 minutes. For experiments using tissue sections, mice 455 were perfused with 0.1M PBS followed by 4% paraformaldehyde. and centrifugation for 5 minutes at 4°C. An LSR II flow cytometer (BD Biosciences) analyzed immunolabeled cell 510 samples. Forward scatter and side scatter parameters gated viable cell populations for phenotypic analysis using 511 antibody panels. Offline data analyses were completed with FlowJo v.10 software (Tree Star, Inc., Ashland, OR). 512 513 In vivo lipopolysaccharide challenge 514 515 hNSG and non-engrafted NSG mice (4-6 months old) were injected 3 mg/kg (i.p.) with lipopolysaccharide (LPS from E. 516 coli; O55:B5, Sigma-Aldrich) or 0.9% saline solution. Mice were anesthetized, blood collected for serum, and tissues 517 placed in 4% PFA overnight at 4°C. Blood was collected into a BD Microtainer serum separator tube with clotting 518 activator, and after 30 minutes tubes were centrifuged at 10,000xg for 5 minutes. Serum was collected, aliquoted into 519 1.5mL Eppendorf tubes, and stored at -80°C. 520 521 Ex vivo purification and stimulation of human splenocytes 522 523 hNSG mice (4-6 months old) were anesthetized and spleens dissected as previously described. Splenocytes were 524 diluted in IMDM and counted on a manual hemocytometer with a 1:1 ratio of 0.4% Trypan Blue to quantify live cells. 525 Contaminating mouse splenocytes were depleted from cell preparations by incubating with MACS anti-mouse CD45 526 magnetic microbeads (Miltenyi Biotec, Auburn, CA) and then washing through magnetic LD Columns as per 527 manufacturer's protocol. Enrichment of human cells was verified via flow cytometry of the depleted and cultured 528 fractions: depleted fraction contained >90% mouse CD45 + splenocytes, while the cultured fraction contained >90% 529 human CD45 + splenocytes (see Fig. 2-figure supplement 1  hNSG mouse injured at 4-months post-engraftment. The co-localization of hCD3 and hCD4 on small cells confirms 825 their T cell identity. In this example, two human T cells are in direct contact with a human CD4 + macrophage. Note the 826 absence of hCD3 labeling on the human macrophage. A large, isolated T cell with immunoblast-like morphology is also 827 identified. 828