Mice

A total of 23 NSG-SGM3 (14♀, 9♂), 34 humanised NSG-SGM3 mice (13♀, 21♂) and 19 C57BL6/J (15♀, 4♂) age-matched mice were used in this study; all mice were obtained from the Animal Resource Centre (Canningvale, Western Australia) and/or local breeding colonies maintained by The University of Queensland’s Biological Resources facility. Mouse age at each procedural step (i.e., humanisation of the immune system and surgery) and endpoints are as specified below and/or in the experimental timelines (see Figs. 1A and 2A). Humanised (hu) NSG-SGM3 mice were generated as described previously [11, 14]. In brief, cord blood was obtained from the Queensland Cord Blood Bank. Density gradients were used to isolate CD34 + haematopoietic stem cells, followed by positive selection using a CD34 + isolation kit (Miltenyi Biotec). Next, two-to-five-day old NSG-SGM3 pups (NOD.Cg-PrkdcscidIL2rgtm1Wjl/Tg(CMV-IL3,CSF2,KITLG)1Eav/MloySzJ; stock no. 013062, The Jackson Laboratory) were conditioned using total body irradiation (100cgy), followed by intrahepatic injection of human CD34 + cells 4 h after irradiation (4.0 × 104 cells for Cohort 1; n = 9 (5♀, 4♂) (Fig. 1A), and 4.2 × 104 cells for Cohort 2; n = 21 (7♀, 14♂) (Fig. 2A)). All mice were allowed to acclimatise for at least 14 days prior to experimentation and housed in individually ventilated cages under specific pathogen-free conditions in certified animal holding facilities at the host institution. They were maintained on a 12-h light–dark cycle and had unlimited access to food and water. Two male mice, both from Cohort 1, were euthanised prior to study completion as they reached humane endpoints on general wellbeing score sheets (weight loss, ruffled fur and hunched posture); these mice were excluded from all subsequent analysis. We otherwise did not observe any indications for an increased rate of infectious complications and/or mortality in NSG-SGM3 and huNSG-SGM3 mice under our experimental paradigm.

Fig. 1

Humanised NSG-SGM3 mice have worse functional recovery from SCI compared to non-humanised controls. A Experimental timeline. NSG-SGM3 pups were irradiated and engrafted with human CD34 + (huCD34 +) stem cells. Mice received spinal cord injury (SCI) surgery at 10 weeks post-engraftment. Functional recovery was assessed weekly using the Basso Mouse Scale (BMS). B Percentage of human CD45 + (huCD45 +) out of all live circulating cells at 8 and 16 weeks-post engraftment (n = 7 per timepoint; 5♀, 2♂). C Comparison of the change in engraftment efficiency between individual mice. Asterisk highlights significant temporal increase in huCD45 + cell when removing one animal with engraftment failure (n = 7 per timepoint; 5♀, 2♂). D Representative immunofluorescent images of spleens from non-engrafted NSG-SGM3 and humanised NSG-SGM3 (huNSG-SGM3) mice, at 16 weeks post-engraftment, stained for huCD45 (red; A488), mouse CD45 (msCD45; green; A647) and Hoechst (cyan). Scale bar is 50 μm. E Pooled BMS scores for huNSG-SGM3 (blue; n = 7) and NSG-SGM3 (green; n = 23 (14♀, 9♂) for days 0–7 post-SCI; n = 17 (9♀, 8♂) for days 14–35 post-SCI) mice following SCI. F BMS scores at 42 days post-injury (dpi) for huNSG-SGM3 (n = 7; 5♀, 2♂), non-humanised NSG-SGM3 (n = 17; 9♀, 8♂), and C57BL/6 J mice (orange; n = 19; 15♀, 4♂). Data are mean ± SEM for all graphs, with statistical analysis being either a paired t-test (B, C), two-way ANOVA (E), or one-way ANOVA with Tukey’s post-hoc (F); *, p < 0.05; ****, p < 0.0001; ns, not significant

Fig. 2

Profiling of the immune response to SCI in humanised mice. A Experimental timeline. huNSG-SGM3 mice were subjected to either laminectomy (sham), spinal cord injury (SCI), or no surgery (naïve controls) at 15 weeks post-engraftment and then sacrificed 7 days later. B Left: engraftment efficiency at 8- (n = 21; 7♀, 14♂) and 16-weeks post-engraftment (study endpoint; n = 7; 3♀, 4♂). Data points show individual animals, along with group means and standard error of the mean (SEM); unpaired t-test; ns, not significant. Right: Temporal change in engraftment efficiency for individual mice. Lines highlight the direction of change in engraftment efficiency over time. C Representative immunofluorescent images showing huCD45 + cells (red; A647) within the different areas of the lesion (i-iii in spinal cord cartoon) in huNSG-SGM3 mice with SCI (7dpi; scale bar is 20 μm); staining for glial fibrillary acidic protein (GFAP) is shown in green (A488) and cell nuclei are cyan. Arrows indicate huCD45 + cells. D Experimental workflow for single cell RNA sequencing (scRNAseq) experiments. huCD45+ cells were isolated by fluorescence activated cell sorting (FACS) from the blood and/or spinal cord for scRNAseq. E Heatmap showing the top-10 gene markers for each identified cell type; colour gradient shows the scaled expression for each cluster marker (centered data divided by the standard deviation). F UMAP of all immune cells collected and sequenced from blood and spinal cord samples. Cells were classified into 11 cell types/subsets based on marker expression. G MetaCell plot of all immune cells, coloured by cell type and subset

Surgical procedures

All surgeries were performed using aseptic technique in a UV-sterilised room. All solutions for injection (e.g., anaesthetic, saline) were either filter sterilised or drawn up in a sterile Biosafety Cabinet prior to injection. Mice for behavioural studies (Cohort 1) were 2–3 months of age at the time of surgery (Fig. 1A), and 3–4 months of age for Cohort 2 (Fig. 2A); mice were administered inhalable oxygen (50%) for a minimum of 5 min prior to anaesthetic injection.

Mice were anaesthetised with a combination of xylazine (20 mg/kg; intra-peritoneal (IP); Troy Laboratories) and tiletamine/zolazepam (50 mg/kg; IP; Virbac), followed by a thoracic contusive SCI as previously described [15]. In brief, paravertebral muscles over the target region were partitioned and anatomical landmarks used to identify T9 [16]. Following this, a dorsal laminectomy of T9 was performed and the spinal column clamped and stabilised. A severe (70kdyne) contusive SCI was then inflicted onto the exposed spinal cord segment using the Infinite Horizon Impactor (Precision Systems and Instrumentation) [17]. The actual applied force and associated tissue displacement were recorded post-impact for each animal, and not different between experimental groups and/or conditions (data not shown). Wounds were closed with 6–90 polygalactin dissolvable sutures (Ethicon) and Michel wound clips (Kent Scientific). Where relevant, sham-operated mice (i.e., ‘laminectomy only’) were also included to control for the effects of surgery itself. Naïve mice were not injured. All mice were kept in a Harvard Apparatus Small Animal Recovery Chamber to maintain body temperature during recovery from anaesthesia. For post-operative care, mice received a single dose of pain killer (buprenorphine in Hartmann’s sodium lactate; 1mkg/kg; sub-cutaneous (SC); Sigma Aldrich), immediately upon waking post-operation, and antibiotic treatment (gentamycin; 10 mg/kg/day; SC; Sigma Aldrich) once daily for 5dpi. Bladders of SCI mice were manually checked and voided twice daily for the duration of each experiment.

Flow cytometry

Flow cytometry was used to characterise select immune cell populations as part of engraftment checks. For this, mice were anaesthetised with methoxyflurane (inhalation), and blood collected at 8 weeks post-engraftment via retro-orbital bleed into lithium-heparin tubes (BD Biosciences); study endpoint blood samples (16 weeks post-engraftment) were obtained via cardiac puncture. Blood samples were processed for flow cytometry as described below under ‘Blood processing’. Additional routine haematological analysis was performed (where specified) on the Mindray BC-5000 Vet Haematology Analyser, using blood samples collected at room temperature (RT) in ethylenediminetetraacetic acid (EDTA) tubes (BD Biosciences) and kept on ice until analysis.

For all other flow cytometry and/or cell sorting experiments, collection procedures were performed under general anaesthesia with sodium pentobarbitone (100 mg/kg; IP; Virbac) prior to processing. The thoracic cavity and pericardial sac were then opened to allow for transcardial perfusion. Where appropriate, blood was collected via cardiac puncture (detailed below). A syringe needle was then inserted into the left ventricle of the heart through the apex, after which the right atrium was punctured. Mice were perfused with 10–15 ml 1 × Hank’s Buffered Salt Solution without Ca2+ or Mg2+ (HBSS; Life Technologies) containing 0.2% heparin (Pfizer) prior to organ dissection.

Blood processing

Blood was collected via retro-orbital bleed or cardiac puncture as described above. For collection via cardiac puncture, the heart was exposed under general anaesthesia by opening thoracic cavity, after which a heparinised syringe was inserted into the left ventricle chamber of the heart to draw up blood. Bloods were immediately transferred into lithium-heparin tubes (BD Bioscience). Next, 50 µl of blood was mixed with 150 µl of anti-coagulant buffer (4 mM EDTA; Sigma Aldrich) in Dulbecco’s phosphate-buffered saline (DPBS; 0.02% KCl, 0.02% KH2PO4, 0.8% NaCl, 0.1% Na2HPO4; 1:4 ratio), followed by incubation with 1 ml red blood cell (RBC) lysis buffer (1:5 ratio) for 7 min at RT. Samples were then centrifuged for 10 min at 300 × g (4 °C) and resuspended in 100 µl DPBS for staining.

General antibody staining procedure for flow cytometry

Appropriate single cell suspensions in DPBS were incubated with viability dye (Zombie Green/NIR/Violet (1:100); BioLegend) for 20 min on ice. Alternatively, where required, samples were stained with propidium iodide (added immediately prior to analysis) rather than Zombie dye as the live/dead exclusion marker (1/3200; Thermo Fisher). Cells were then centrifuged at 4 °C for 10 min at 300 × g (blood) or 500 × g (spinal cord). Cell pellets were resuspended in FACS buffer, i.e., 0.5% Bovine Serum Albumin (BSA; Sigma Aldrich) and 2 mM EDTA in DPBS, followed by a 10-min incubation with anti-CD16/32 (1:200; BD Bioscience) for Fc receptor blocking. Next, samples were incubated with an appropriate cocktail of fluorophore-conjugated antibodies (see Table 1) for 20 min (4 °C) and topped up with 1 ml FACS buffer. Samples were then centrifuged again for 10 min at 4 °C (300–500 × g) and cell pellets gently resuspended in FACS buffer. Propidium iodide-fluorescing counting beads (5 µl; Beckman Coulter) were then added to each tube as an internal standard to allow for accurate calculation of absolute cell numbers. Samples were analysed on the Fortessa flow cytometer (BD Biosciences) with BD FACS Diva software, and data was analysed with FlowJo software (v10.8, Tree Star, Inc.). Single-stain compensation beads (ABC Total Antibody Compensation Bead Kit; Thermofisher) were included to allow for removal of spectral overlap between fluorophores during the analysis phase.

Table 1 Antibodies used for immunofluorescence and flow cytometryBehavioural analysis

The Basso Mouse Scale (BMS) was used to evaluate locomotor performance of SCI mice at regular intervals post-surgery. Mice were observed in an open field (a circular, flat area) for 4 min by at least 2 single-blinded investigators. Published guidelines [18] were used to rate individual mice on a scale of 0–9 (9 indicating normal locomotion) based on their hindlimb movement, stepping ability, fore-hindlimb coordination and trunk/tail stability.

Tissue processing for histology

For histological assessment, mice were perfused with 10-15 ml saline containing 2% NaNO2 and 10 IU/ml heparin (Pfizer), to prevent coagulation, followed by 20-30 ml of phosphate-buffered Zamboni’s fixative (2% picric acid, 2% paraformaldehyde, pH 7.2–7.4). Next, spleens, livers and vertebral columns were removed and post-fixed overnight at 4 °C. Spinal cords were then dissected from the vertebral column on the following day and again post-fixed overnight. They were then cryoprotected along with other organ samples by immersion in 10% and then 30% sucrose solution at 4 °C. Following this, all tissue samples were embedded in Optimal Cooling Temperature compound (ProSciTech) and snap-frozen in methylbutane on dry ice. Transverse spinal cord, liver and spleen sections (20 µm) were cut using a Leica Cryostat CM3050-5, collected onto Superfrost Plus slides (1:5 series; Lomb Scientific), air-dried and then stored at −80 °C until further processing.

Immunofluorescent staining

For visualisation of select immune cell populations within the injured spinal cord (see Table 1), slides were defrosted and air-dried for 60 min at RT, after which they were washed in 1 × phosphate-buffered saline (PBS). Slides were then incubated for 1 h at RT in blocking buffer (2% BSA, 0.2% Triton X-100 in PBS) to reduce non-specific binding. Following this, they were washed again in PBS and incubated overnight at 4 °C with an appropriate cocktail of primary antibodies (diluted in blocking buffer). The following day, slides were washed again in PBS and incubated for 1 h at RT with secondary antibodies as well as Hoechst 33342 nuclear dye (1:1000; Sigma Aldrich). After a final round of washing, slides were cover slipped with DAKO fluorescent mounting medium (Sigma-Aldrich).

Stained spleen (Fig. 1D) and spinal cord sections (Fig. 2C) were imaged using a high speed automated Nikon sterology upright widefield fluorescence microscope with a 40 × air objective (Plan APO, numerical aperture 0.95, working distance 0.17–0.25 mm) and Hamamatsu Orca Flash 4.0 sCMOS camera (2048 × 2048 pixels, 6.5 × 6.5 μm pixel size, 82% QE). For Additional file 1: Fig. S4A-C, images were taken on a Diskovery Spinning Disk Inverted Confocal Microscope using a 70 μm spinning disk, 20 × air objective (CFI Plan APO VC, numerical aperture 0.75, working distance 1.00 mm) and Zyla 4.2 sCMOS camera (2048 × 2048 pixel resolution, 6.5 × 6.5 μm pixel size, 82% QE). All images were acquired at RT. NIS Elements Advanced Research software was used for image acquisition and brightness and contrast display settings were adjusted post-acquisition using ImageJ (v1.52p; display settings were made the same between images within a single figure panel).

H&E staining and GVHD assessment

For haematoxylin and eosin (H&E) stains, liver sections were stained with Mayer’s haematoxylin for 4 min, washed in running water for 2 min, followed by 70% EtOH for 2 min, and then stained with 80% ethanol-eosin for 7 s. Next, samples were again washed in 90% EtOH, followed by 100% EtOH (repeated three times), before being cleared in xylene, mounted with DepeX and then coverslipped. Sections were imaged using a Zeiss AxioScan Z1 slide scanner at 20 × magnification.

Graft-versus-host disease (GVHD) pathology in H&E stained liver sections was manually quantified by an investigator who was blinded to the experimental condition, using established grading criteria for the degree of mononuclear infiltrate [19, 20]. A score of 0 reflected no visible pathology, escalating from here with a score of 1 being defined as punctate infiltration of mononuclear cells, 2 as sporadic perivascular infiltration of mononuclear cells, 3 as sporadic perivascular infiltration of mononuclear cells with some additional infiltration of the parenchyma, or 4 as widespread perivascular infiltration of mononuclear cells with spread into the parenchyma.

Spinal cord processing for cell sorting

After perfusion with HBSS, the T8-T10 region of the spinal cord was dissected and weighed in an Eppendorf tube containing 500 μl HBSS (Life Technologies). Spinal cords were then mechanically dissociated using fine scissors, pelleted (10 min at 500 × g; 4 °C), and enzymatically digested for 20 min at 37 °C in 2.5 ml L-glutamine-containing Leibovitz’s L-15 Medium (Life Technologies) containing 0.1% papain (Worthington) and 0.1% DNase I (Roche). Next, spinal cord samples were triturated by gently pipetting the solution up and down (10–15 times), before being passed through a 40 μm cell strainer with 5 ml neutralising solution (10% Fetal Bovine Serum (FBS; Sigma Aldrich) in Dulbecco’s Modified Eagle Medium (Life Technologies)). Cell suspensions were then centrifuged for 10 min at 500 × g and resuspended in 100 µl DPBS for live/dead and antibody staining as detailed under ‘General antibody staining procedure for flow cytometry’.

Following this, spinal cord samples were additionally incubated with 20 µl anti-Cy7 MACs beads (Miltenyi Biotec) in 80 µl FACS buffer for 15 min at 4 °C, washed in 1 ml FACS buffer at 500 × g, and then run through MS Columns (Miltenyi Biotec) as per the manufacturer’s instructions to allow for positive selection/enrichment of human CD45 + (huCD45 +) cells prior to cell sorting.

Sample preparation and sorting for single-cell RNA sequencing

Male naïve and sham-operated huNSG-SGM3 controls (n = 1 technical replicate per condition) as well as huNSG-SGM3 SCI mice (n = 2 technical replicates), all from Cohort 2, were deeply anaesthetised with 4% isoflurane at 7dpi. Blood was collected via cardiac puncture from all mice, in addition to the T8-T10 spinal cord segments from SCI mice. All samples were obtained and swiftly processed on ice and/or at 4 °C as detailed earlier; a minimum of 350 µl blood was taken from each mouse and added volumes of 4 mM EDTA and RBC lysis buffer adjusted on an individual basis to account for any variations in sampled blood volumes.

Next, Fc receptors were blocked with mouse IgG (1 μg/ml in FACS buffer, Jackson) for 10 min, followed by incubation of the samples with anti-human CD45 APC-Cy7 (1:50, clone 2D1; BioLegend) on ice for 20 min. Spinal cord samples were additionally incubated with 20 µl anti-Cy7 MACs beads (Miltenyi Biotec) in 80 µl FACS buffer for 15 min at 4 °C. Spinal cord samples were then washed, passed through an MS MACS column (Miltenyi Biotec) and huCD45 + cells collected as per the manufacturer’s instructions. All samples were resuspended in 400 µl FACs buffer and stained, as appropriate, with propidium iodide for live/dead exclusion (1:400; Thermofisher) prior to fluorescence activated cell sorting (FACS); single-stained samples were included to allow for the removal of any spectral overlap between fluorophores. huCD45 + cells were sorted using a BD FACS Aria Cell Sorter, with all samples gated to exclude debris, doublets and triplets, and dead cells. huCD45 + events were sorted into sterile RNAse-free Eppendorf tubes containing 500 \(\upmu\)l sterile, ice-cold PBS + 10% FBS (Additional file 1: Fig. S3; orange gate).

RNA sequencing library preparation and Chromium 10 × single cell 3’ v3 sequencing

Library preparation and sequencing was performed at the Institute for Molecular Bioscience Sequencing Facility (The University of Queensland). Here, a cell count was performed first to determine cell viability (93–100%) and yield post-sorting. Single-cell suspensions were partitioned and barcoded using the 10X Genomics Chromium Controller (10X Genomics) and the Single Cell 3' Library and Gel Bead Kit (V3; 10X Genomics; PN-1000075 or PN-1000092). The cells were loaded onto the Chromium Single Cell Chip B (10X Genomics; PN-1000073 or PN-1000074) to target a total of 20,000 cells per sample. GEM generation and barcoding, cDNA amplification, and library construction were all performed according to the 10X Genomics Chromium User Guide. A total of 11 cDNA amplification cycles were performed, and one quarter of the cDNA was used as input for library construction. 12 SI-PCR cycles were used for final indexing PCR. Reactions were performed in a C1000 Touch thermal cycler with a Deep Well Reaction Module (Bio-Rad).

Single cell transcriptome libraries were quantified on the Agilent BioAnalyzer 2100 using the High Sensitivity DNA Kit (Agilent, 5067-4626). Gene expression libraries were pooled in equimolar ratios. The final pool was quantified by qPCR using the KAPA Library Quantification Kit—Illumina/Universal (KAPA Biosystems, KK4824) in combination with the Life Technologies Viia 7 real time PCR instrument. Denatured libraries were loaded onto an Illumina NextSeq-500 and sequenced using a 150-cycle High-Output Kit as follows: 28 bp (Read1), 8 bp (i7 index), 111 bp (Read2). Read1 supplies the cell barcode and UMI, i7 the sample index, and Read2 the 3’ sequence of the transcript. All samples were processed in one sequencing run.

scRNAseq analysis

Raw sequencing reads were processed and then mapped to a custom reference genome combining the human (GRCh38) and mouse (mm10) genomes in CellRanger v3.1.0 (10X Genomics). Cells were computationally divided into human and mouse fractions using k-means clustering (k = 2), with inspection of the gene content for the two clusters confirming the successful splitting of cells by species (data not shown). Cells matching the mouse genome were discarded and the resulting filtered count matrices from human cells used for downstream analysis. Median absolute deviation (MAD) filtering was implemented using scater v1.14.6 [21] to remove cell outliers with low library size and/or gene counts (> 3 MADs below the median value), or high mitochondrial or ribosomal gene percentage (> 3 MADs above the median value). Doublets were predicted with scds v1.2.0 [22] and cells were removed if they were predicted as doublets by at least two of the three included prediction methods (bcds, cxds, or hybrid), and expressed more than 3,000 genes. Gene counts were normalised using scran v1.14.6 [23]. All blood and spinal cord datasets were integrated for downstream analysis. Data scaling, dimensionality reduction, clustering and sub-clustering, data integration, and marker prediction were all performed in Seurat v3.1.4.9904 or Seurat v4.0.0 [24, 25]. Seurat objects were converted to SingleCellExperiment v1.8.0 objects where required [26]. Different cluster resolutions were tested and assessed using Clustree v0.4.2 [27]; a final resolution of 0.4 was selected after considering cluster stability, clustering patterns and top marker genes. Clusters were named based on assessment of top marker genes. The MetaCell pipeline [28] was implemented using default parameters. Differentially expressed genes (DEGs) were detected using Seurat, and visualised using ComplexHeatmap v 2.6.2 [29]. Gene ontology (GO) analyses were performed using the hypergeometric test in ClusterProfiler [30, 31] against the org.Hs.eg.db database v3.12.0 [32], and using a gene background as the union set of only genes expressed within each dataset. Ribosomal and mitochondrial genes were removed before DEG and GO analysis. Cell–to–cell interaction (CCI) analyses were run using CellChat v1 [33], with the built-in human interaction database; subsequent GO analyses were performed above using a gene background of expressed ligands or receptors only, as found in the CellChat database. All analyses were run on R v4.0.3 using a Benjamini-Hochberg-adjusted p-value of less than 0.05 as the significance threshold.

Finally, for comparisons of our in-house scRNAseq dataset from humanised mice with a publicly available bulk RNASeq dataset from human SCI subjects [1], we first log-transformed the counts for both datasets and averaged each gene. A linear model was then fitted over these pairs of average expression values for each gene, and a linear regression performed to identify correlated expression patterns between datasets. This was also repeated for each comparable condition between our data and the Kyritsis data, i,e, naïve mice against healthy human control subjects (HC), sham-operated mice against the human trauma control (TC) patients, and between spinal cord-injured (SCI) mice and human patients.

Statistical analysis

Statistical analysis for behavioural assessment, histology and flow cytometry experiments was conducted using either a student’s T-test, one- or two-way ANOVA with post-hoc corrections for multiple comparisons as appropriate. Group sizes for engraftment checks and behavioural studies were based on relevant previous experiments and the literature [9]. Our scRNAseq dataset was otherwise sufficiently large to detect rare cell (sub)types (as few as 5 cells) with over 99% sensitivity and a false discovery rate of less than 5% [34]. All data are presented as mean ± standard error of the mean (SEM), with statistical significant results defined as p < 0.05. For all graphs, *p < 0.05, **p < 0.01, ***p < 0.005 and ****p < 0.0001.