Patrolling human SLE haematopoietic progenitors demonstrate enhanced extramedullary colonisation; implications for peripheral tissue injury

Systemic lupus erythematosus (SLE) is an autoimmune disease where bone-marrow-derived haematopoietic cells have a key role in its pathogenesis with accumulating evidence suggesting an aberrant function of haematopoietic stem/progenitor cells (HSPCs). We examined whether patrolling HSPCs differ from bone-marrow HSPCs both in SLE and healthy individuals, and how they participate in peripheral tissue injury. By employing next-generation RNA sequencing, the transcriptomes of CD34+ HSPCs deriving from the bone marrow and those patrolling the bloodstream of both healthy and individuals with SLE were compared. Patrolling SLE and Healthy human HSPC kinetics were examined through their inoculation into humanised mice. Patrolling and bone-marrow HSPCs have distinct molecular signatures, while patrolling SLE HSPCs showed an enhanced extramedullary gene expression profile. Non-mobilised, SLE-derived circulating HSPCs demonstrated altered homing capacities. Xenotransplantation of circulating HSPCs in humanised mice showed that human peripheral blood HSPCs possess the ability for extramedullary organ colonisation to the kidneys. Circulating and bone marrow-derived HSPCs are distinct in steady and diseased states. Patrolling SLE CD34+ HSPCs are able to home at extramedullary sites such as the spleen and kidneys, potentially participating in peripheral tissue injury.


Results
SLE involves the dysregulation of the HSCs, resulting in a multi-organ disease phenotype 10 that could potentially involve circulating patrolling progenitors. To this end, we isolated total RNA from magnetically-isolated CD34 + cells extracted from the periphery (PBMCs) and the bone marrow aspirates [bone marrow mononuclear cells (BMMCs)] of age-matched healthy and SLE individuals and subjected them to NGS mRNA-seq. All DEG from each comparison were used as input on downstream analyses, using multiple pathway databases for maximum enrichment and predictive modelling 16 . Human PB CD34 + HSPCs differ from BM CD34 + HSPCs. We initially assessed whether the molecular signatures are distinct among the four haematopoietic progenitor populations (PB-SLE, PB-Healthy, BM-SLE and BM-Healthy). For this, we performed RNA-seq analysis in a human CD34 + cells from 26 human samples in total (3 healthy BM, 7 healthy PB, 6 SLE PB and 10 SLE BM). Principal component analysis (PCA) depicting all the DEG between those four groups, indicated distinct patterns in the transcriptome between BM-and PB-derived CD34 + progenitors, irrespective of disease status (four groups, Fig. 1A,B). This finding was also recapitulated when PB-and BM-derived progenitors were compared within the same disease setting (healthy or SLE, Fig. 1C, Supplementary Tables 1 and 2).
Next, we set to explore if transcriptome analysis could reveal significant gene expression differences in CD34 + HSPCs, due to disease status (i.e. Healthy vs SLE). PCA analysis showed that PB comparison (in contrast to BM comparison) did not show strong segregation, bearing a smaller number of DEG ( Supplementary Fig. 1A,B).
We looked further into this comparison by comparing the gene expression analysis of SLE DEG between PB and BM, by using the Healthy DEG PB versus BM output as a reference point (homeostatic cut-off). CLINVAR analysis provided a unique SLE-specific signature, providing a list of genes (and potentially their variants) altered in PB SLE CD34 + HSPCs ( Fig. 2A). Hierarchical clustering revealed that these genes were indeed differentially expressed in PB SLE versus PB Healthy progenitors. KEGG pathways analysis revealed that antigen presentation and processing that included Graft-versus-Host-Disease (GvHD), allograft rejection and oxidative phosphorylation, were overrepresented in SLE compared to healthy counterparts (Fig. 2B). Haematopoietic cell lineage, Th 17 cell differentiation, antigen processing and presentation as well as cell cycle, were also enriched in the PB setting, with the majority of genes being elevated (except cell cycle GO) in the SLE PB-derived cell populations, in comparison to healthy PB-derived progenitors. These data pinpoint a potential discrepancy between SLE PB-and BM-derived CD34 + cells that may cause significant differences in progenitor mobilisation and immune activation 17,18 . KEGG analysis of DE genes in SLE versus healthy CD34 + cells showed that genes involved in allograft rejection pathways were enriched in the SLE setting when compared to healthy CD34 + HSPCs (Fig. 2B).
Together, these analyses underscore the transcriptome differences in the circulating HSPCs in relation to BM CD34 + progenitors while underpinning their potential implication in SLE disease pathogenesis.
Human SLE PB HSPCs show an enhanced migration gene expression profile. The CLINVAR database for genes involved in human disease 19,20 indicated that SLE haematopoietic circulating progenitors differentially expressed key genes being involved in pro-inflammatory and immune-regulatory functions, including cell mobilisation [IFN-γ, IL-8R (CXCR1), CCL3L1-CCL3], as well as in SLE-specific immune regulation (CTLA4, STAT4, PCD1), when compared to Healthy PB circulating progenitors (Supplementary Fig. 2A). REACTOME Reactions and Pathways analyses 21 revealed significant changes in gene expression involved in extracellular matrix (ECM) organisation, immune system activation through CD3, and cytokine/chemokine regulation that involved IL-10 signalling and trafficking in inflammation sites 12,22 , with the majority of genes in the SLE-derived PB progenitors, being upregulated in comparison to Healthy-derived PB ( Supplementary  Fig. 2B). Finally, KEGG pathways analysis 23 was indicative of DEGs that involved the migratory and extramedullary increased potential (cytokine and cytokine-receptor interactions) as well as altered immune activation of  Specific transcription factor binding sites predicted in PB-and BM-derived CD34 + progenitors. Next, we sought to identify potential Transcription Factor (TF) binding sites based on the DE genes identified between PB and BM, in both healthy and SLE patient samples. The analysis revealed a non-exhaustive list of potential TF sites which pinpointed unique TFs that may regulate a cohort of gene regulation (Fig. 3). SLE PB-derived progenitor DEGs revealed a marked upregulation in SIRT7 and a downregulation in NRF1 and NRF1-dependent DEGs (131 genes), while LXR (NR1H3)-dependent gene expression showed mixed DEG patterns (36 genes), when compared to SLE BM-derived HSPCs (Fig. 3A,B and Supplementary Fig. 5A). BM-derived and PB-derived progenitors were used as an input on CLINVAR database, using as a reference cut-off the DEG found between healthy BM-derived and PB-derived progenitors. Graphical representation showing genes found to be either enriched in SLE (dark red) or Healthy (light green) progenitors, from genes reported in the CLINVAR database (p = 0.049). A heatmap of these DE genes indeed indicated higher expression levels in SLE PB-derived CD34 + , in comparison to SLE BM-derived CD34 + . (B) KEGG pie chart analysis of DEG between SLE-derived BM and PB CD34 + progenitors (red), and Healthy BM vs PB-derived (green). Pie chart connecting lines indicate kappa-score relations. Heatmaps indicating DEG that shown to be enriched in Healthy-derived PB (purple) when compared to Healthy-derived BM (light orange) progenitors, and SLEderived PB (red) when compared to SLE-derived BM (black) progenitors. All analyses were performed with an FDR < 0.05 (Bonferroni and Heidelberg) correction.  Fig. 3D and Supplementary Fig. 5E). SIRT7 expression of its nonsensemediated decay isoform showed higher expression in PB, in relation to BM (both SLE and Healthy), yet with a two-fold decrease (non-statistically significant) in SLE PB, in comparison to Healthy PB. In healthy PB-derived progenitors, predicted expression of altered TFs AP1 (JUN)-and CEBPB-dependent DE genes produced mixed differential gene expression patterns, (265 and 23 genes, respectively, Fig. 3C and Supplementary Fig. 5B).
We then investigated the prospect of TFs that may be related to the DE genes on PB-and BM-derived HSPCs between SLE and healthy patients. Prediction analysis showed a cascade of TF-dependent DEGs in the BM CD34 + progenitors, but none in the PB CD34 + progenitors ( Supplementary Fig. 5C,D). In most of the predicted TFs, DEGs were downregulated in the SLE BM-derived progenitors compared to controls ( Supplementary Fig. 6). was not statistically different in any of the four groups, albeit there was a trend of lower expression in the SLE PB. In contrast, SIRT7 expression was vastly and significantly elevated in PB CD34 + progenitors, in comparison to BM CD34 + progenitors. One-way ANOVA with Bonferroni's posthoc test, *p < 0.05, n = 3-6. SLE PB CD34 + progenitors showed an altered extramedullary differentiation in humanised mice. Based on the potentially altered migratory transcriptional profile of circulating HSPCs, we interrogated the behaviour of human PB CD34 + progenitors as xenotransplants. Human SLE or Healthy 2.5 × 10 5 CD34 + progenitors were injected into each of fourteen 2-3 month-old NBSGW mice (one human sample per mouse, 7 inoculated with SLE-derived and 7 with Healthy-derived PB CD34 + HSPCs, Figs. 4 and 5A). Mice were sacrificed at 9, 13, and 20-weeks post-injection to assess human cell colonisation potential. Mice did not develop alopecia, a characteristic of GvHD, up to 20-week post-injection 26 . We observed altered migration kinetics and www.nature.com/scientificreports/ lymphocyte/myeloid differentiation potential from SLE PB CD34 + human cells when compared to healthy PB CD34 + cells, in the murine BM (Fig. 4B).
In the spleen and kidneys differences in human CD19 + , CD14 + and CD3 + cell percentages, between SLEand Healthy-derived progenitors were evident (Figs. 4C and 5A). Of interest, a substantial number of human CD34 + was evident in the murine kidneys, indicating possible homing of secondary organs. We also examined whether HSPCs could have potentially adopted a BM niche-like profile [27][28][29] . We found that human SLE-derived PDGFRα + and CD105 + cells were present and in an increased percentage in the murine BM, kidneys and spleen, in comparison to healthy-derived human PB progenitors (Figs. 4B,C and 5). www.nature.com/scientificreports/ These xenotransplantation experiments indicate that human PB HSPCs mediated extramedullary colonisation with an increased ability of the SLE-derived HSPCs to home at extramedullary sites such as the kidneys, where they may participate in local pathology through the formation of primitive HSPC colonies (Fig. 5) 29 .
Increased frequency of MPPs in the peripheral blood of SLE patients, with increased extramedullary colonisation in humanised mice. In order to assess whether HSCs/MPPs numbers (found within the CD34 + cell population) differ in the periphery of SLE patients, PBMCs were subjected to immunostaining against the surface markers CD34, CD38, CD45RA, CD90 and CD49f., along with the cell viability dye 7AAD. Parallel gating indicated the presence of well-defined HSCs and MPPs in PBMC derived from both SLE patients and Healthy individuals (Fig. 6A). On gated CD34 + CD38 − populations, SLE MPPs were significantly more in BM and statistically significant more in PB (over a two-fold increase) when compared to healthy controls [Fig. 6B, and as we have shown recently concerning human healthy and SLE BM CD34 + transcriptomes 11 ]. No statistical difference was observed in CD34 + CD38 + cell populations (Supplementary Fig. 7A). In addition, SLE flares and a PGA ≥ 1.5, showed to be linked to a two-fold MPP increase in SLE PB (Fig. 7B and Table 1).
In light of this, SLE (derived from patients that exhibited SLE flares with a PGA ≥ 1.5) or Healthy MPPs were also injected into each of four 2-month old NBSGW mice, and assessed for colonisation potential in the BM, spleen and kidneys, 5 weeks later (one human sample per mouse, Fig. 6C and Supplementary Fig. 7B). FACS analysis indicated that MPPs colonise neither the bone marrow nor the spleen (Supplementary Fig. 8). In contrast, only the kidneys were occupied by human MPP-derived CD105 + and PDGFRα + cells, with SLE-derived cells showing an increased presence, when compared to Healthy-MPP-derived cells.

Discussion
Dysregulation of HSPC activity in the BM has been reported in several chronic inflammatory diseases, including inflammatory bowel disease, atherosclerosis and in SLE. Here, we investigated the potential gene expression differences that involve altered migration between PB and BM CD34 + HSPCs from both Healthy and SLE patients.
We confirm that PB CD34 + HSPCs are distinct from BM CD34 + HSPCs as previously shown in a study relying on surface marker intensity alone 30 . We show that the transcriptional landscape is vastly different between BM and PB HSPCs, likely underlying altered outcomes concerning extramedullary colonisation and engraftment success 31 . The authors of the latter study also pinpointed, as we did, to the importance of E2F TF family gene expression in cell cycle progression. This could be why we see an increase in the number of MPPs in the SLE setting, as confirmed in an SLE murine model in our laboratory 11 (and recent unpublished observations). In fact, what we observed was that PB HSPCs have decreased oxidative phosphorylation and cell cycle signatures as compared to BM HSPCS, indicating a more activated haematopoiesis along with a reduced self-renewal potential 32 .
In the SLE setting, PB HSPCs have a gene expression signature that could exacerbate inflammatory responses at local tissues such as the kidneys and other target organs in SLE such as joint and skin. This could be because a number of relevant transcriptional pathways is altered in PB HSPCs, in comparison to BM HSPCs. Of interest, the TLR-7/TNF-alpha/IFN-gamma axis that has been shown recently to be involved in extramedullary damage in murine SLE was shown to be highly upregulated in our human PB HSPCs, and especially in SLE PB HSPCs, when compared to BM HSPCs (data not shown) 33 . In humanised mice, xenotransplanted SLE-PB HSPC-derived cells were more prone to colonise extramedullary organs such as kidneys but not the BM nor spleen [possibly due to decreased oxidative phosphorylation and self-renewal capabilities 32 ], when compared to Healthy-PB HSPCs. SLE-PB CD34 + and MPP progenitors may participate in local inflammatory reactions in the periphery due to their increased homing potential and altered T h activation 34 while could be promoting an aberrant haematopoietic "niche" 35,36 . Indeed, Noroozinia A. et al. recently demonstrated the presence of human interstitial CD34 + cells in patients diagnosed with the active phase of lupus nephritis 37 . Additionally, chimeric human CD3 + CD34 + cell populations have been identified in human kidneys with active SLE, similar to the findings in our xenotransplantation experiments 38 . A future proof-of-concept experiment would be to fluorescently tag human SLE PB-derived HSPCs prior to inoculation and assess their presence in organs of interest through immunohistochemistry.
Of particular interest is NRF1 TF, a master mitochondrial regulator 39 , which showed a marked decrease in SLE PB, in comparison to SLE BM (but not healthy PB, data not shown). NRF1 is implicated in pathways related to mitochondrial biogenesis and systemic chronic inflammation, including SLE 40,41 . A recent publication indicated that SIRT7, a nutrient-sensing protein and a histone deacetylase, when bound to NRF1, is able to alter essential HSC functions related to ageing, possibly via mitochondrial function 15 and/or repopulation potential via invoking genome stability (when located to the nucleus) 13,42 . We did observe a multi-fold increase in SIRT7 gene expression, in PB-derived cells, in comparison to BM-derived cells, and a global decrease in oxidative phosphorylation gene expression (Figs. 4B and 5D, respectively). These events could be linked to a reduced HSC quiescence with altered regenerative and bone-marrow reconstitution capabilities 13,14,24,25 and, as far as SLE is concerned, direct multi-organ autoimmune inflammation 43 through extramedullary colonisation. Our results corroborate those of a recent study where E2F1 was identified as one of the susceptibility loci in SLE, directly involving NRF1 in Asian and European human populations (data not shown) 40 . Of note, SIRT7 and NRF1 genes were inversely, but not differentially, expressed in Healthy PB versus BM CD34 + HSPCs.
A recent study pinpointed that SIRT1 is involved in recurring infections in SLE patients 44,45 ; SIRT7 has the ability to bind directly to SIRT1, possibly being involved in lupus nephritis 46 . Recently, Alexander Kaiser et al. reported that SIRT7 levels influence remission responses after HSPC transplantation in myeloid leukaemia 47 . As such, the aforementioned SIRT could be one of the potential epigenetic targets for alleviating peripheral tissue injury in SLE 48 . Indeed, ongoing experiments will focus on investigating if this protein complex confers substantial changes to HSC's reconstitution capability and other properties such as multi-organ extramedullary www.nature.com/scientificreports/ colonisation conferring organ-specific SLE tissue injury (model proposed for this study, Fig. 7). To our knowledge, this is the first study that explores the transcriptome of human non-mobilised circulating HSPCs in the context of a systemic autoimmune disease. We recognise that an important limitation of the current study is that the expression profile of bulk CD34 + HSPCs cannot disentangle minor shifts in the transcriptome of progenitor subpopulations. We have shown that from all the major HSPC subpopulations examined primarily with CD34 and CD38 surface markers, MPPs, but not HSCs, were significantly enriched in SLE-derived PBMCs, when compared to Healthy-derived PBMC. Investigating more human HSPC subpopulations using a larger combination of cell surface markers should potentially reveal novel subgroups that persist in the SLE setting. Yet, this study does pave the way for revealing clinically important transcriptomic differences and migration kinetics between niche and patrolling HSPCs in humans; these findings need to be investigated further using single-cell mRNA-seq and/or mass cytometry technologies.
In summary, we show here that the human PB CD34 + HSPC transcriptome differs substantially from that of human BM CD34 + HSPC, leading to potentially altered homing capabilities. These findings underscore the need for studying further non-mobilised PB CD34 + cells for haematopoietic stem cell therapy (HSCT) regimes, since they may be more beneficial in some autoimmune disease treatments, including SLE, as reported recently 34 . These data also shed light to a pathway that involves TF NRF1 and histone deacetylase SIRT7 in SLE. Finally, we also demonstrate for the first time that in an autoimmune/inflammatory disease, human HSPCs are not only activated and circulating, but they are also able to migrate and survive in places other than the BM such as the spleen, a site of peripheral immune responses in SLE, as well as at sites of tissue damage, such as kidneys.

Methods
Animals. The NBSGW humanised mouse strain was purchased from Jackson Laboratories (JAX Stock No. 026622) and maintained as homozygotes (NOD.Cg-Kit W-41J Tyr + Prkdc scid Il2rg tm1Wjl /ThomJ). All animals used were 2-3 months of age upon the time of xenotransplantation experiments and fed chow diet in germ-free housing conditions. All mouse animal work and experimental protocols have been approved by the BRFAA ethics committee and the Attica Veterinary Department (758634/22-11/2019). The study is reported in accordance with ARRIVE guidelines.
Isolation of human CD34 + progenitors from BM and PB. Human BM aspirates and PB (10 ml each) were collected in EDTA-coated tubes from healthy and SLE patients and subjected to density gradient centrifugation using Histopaque-1077 (Sigma-Aldrich). Briefly, blood was diluted 1:2 with PBS (for PB) and 1:3 PBS (for BM) and carefully layered over Histopaque medium. Tubes were centrifuged at 400 g for 30 min (no break) at room temperature. White blood cell layer was carefully collected, and cells were washed with PBS, and treated with red-blood cell lysis buffer (Biolegend, cat no. 420301) prior to magnetic bead separation in order to obtain a minimum of 90% CD34 + -enriched cell population (human Diamond CD34 isolation kit, Miltenyi Biotec, Bergisch Gladbach, Germany 130-094-531).
For the xenotransplantation experiments in the humanised mice, a single dosage of 2.5 × 10 5 human PB CD34 + cells were injected retro-orbitally with 100 μl cell suspension into 2-3 months old NBSGW mice. Human conjugated antibodies used for assessing kinetics derived from Biolegend (1:100 dilution); huCD45-Brv421  NGS data analysis pipeline. Raw reads in FASTQ format were collected and quality control was performed using FASTQC 49 . Low quality bases (q < 30) and adapters were trimmed from the 3' end of the reads using Cutadapt 1.16 50 . Alignment was performed using the gapped-read STAR 2.6 mapper 51 against the human genome (hg38). Downstream analysis including quantification using the gencode.v29 annotation gtf file 51 , PCA plots, differential expression analysis using edgeR and heatmaps was performed using SeqMonk 1.44.0 software 52 . Genes with a false discovery rate < 0.05 (FDR-Benjamini and Hochberg correction 53 ) were considered significantly differentially expressed (DEG). Gene Ontology (GO) and pathway enrichment analysis as well as enrichment of transcription factor (TF) binding sites using DEG were performed using Toppgene 54 , Cytoscape and ClueGo 55 tools. GO terms, pathways and TF-binding sites with a FDR < 0.05 were considered statistically significantly enriched.
Semi-Quantitative PCR. RNA  Statistics. Statistical analysis on flow cytometric data was performed using ANOVA T-test using Mann-Whitney or Bonferroni post-hoc test, where appropriate (p < 0.05). RNA-seq data statistical analysis was initially performed using EdgeR with a p value (< 0.05) and multiple correction testing, for obtaining DEG lists. Downstream analysis involved False Discovery Analysis (FDR) based on Benjamini and Hochberg 53 . For the xenotranplantation inoculation outcome, the GraphPad Prism v8 software was employed.
Patients and study approval. Informed consent was obtained from all patients and human controls prior to sample collection (IRB protocol number 10/22-6-2017). Table 1 depicts the demographics and diseases characteristics of SLE patients at the time where PB samples were obtained. In all analyses presented here, all human subjects were female apart from three individuals (two SLE and one healthy), identified on relevant figures and figure legends. BM aspirates were obtained from SLE and healthy controls. BM SLE and healthy aspirates derived from agematched individuals (set for practical reasons to no more than 10 difference of age was a 10-year age). Patients met the 1997 American College of Rheumatology revised criteria for the classification of SLE 56 .