UM171 induces a homeostatic inflammatory-detoxification response supporting human HSC self-renewal

Elucidation of the molecular cues required to balance adult stem cell self-renewal and differentiation is critical for advancing cellular therapies. Herein, we report that the hematopoietic stem cell (HSC) self-renewal agonist UM171 triggers a balanced pro- and anti-inflammatory/detoxification network that relies on NFKB activation and protein C receptor-dependent ROS detoxification, respectively. We demonstrate that within this network, EPCR serves as a critical protective component as its deletion hypersensitizes primitive hematopoietic cells to pro-inflammatory signals and ROS accumulation resulting in compromised stem cell function. Conversely, abrogation of the pro-inflammatory activity of UM171 through treatment with dexamethasone, cAMP elevating agents or NFkB inhibitors abolishes EPCR upregulation and HSC expansion. Together, these results show that UM171 stimulates ex vivo HSC expansion by establishing a critical balance between key pro- and anti-inflammatory mediators of self-renewal.


UM171 triggers a specific inflammatory response in HSPC
In preliminary experiments using hematopoietic cell lines, we found that UM171 triggers simultaneous activation of pro-inflammatory (including IKK/NFkB) and anti-inflammatory (including detoxification responses) programs (Fig 1A-1C and S1 Fig). These programs are activated as early as 6 hours post-treatment, with the most significant changes reached at 48-72h.
To characterize these signals more precisely within the hematopoietic hierarchy, we deployed a single-cell sequencing strategy ( Fig 1D). We next performed integrated gene expression analysis of primitive and committed cell subsets subjected to both low and high UM171 concentrations. As proof of principle, we observed that upregulation of HSC specific genes, most notably PROCR or JAML [23] was highest in the primitive compartment and followed a dose dependency to UM171 (Fig 1E). In addition, we noticed a marked increase in genes encoding MHCassociated proteins, most notably HLA-A and B, as well as beta 2-microglobulin (B2M) in all conditions compared to DMSO in all conditions compared to DMSO (S2D Fig). Since HLA upregulation is a hallmark response to pro-inflammatory signals, this suggested that UM171 induces an inflammatory stimulus in CB cells across several cell-types. Supporting this hypothesis, gene set enrichment analysis revealed pathways related to MHC protein complex binding, inflammation and NFkB signaling as dominant features of UM171 treatment (Fig 1F, S2E Fig  and S1 Table).
Subpopulation analyses (primitive versus committed) showed common (e.g. HLA) and selective inflammatory responses. As shown in Fig 1G, inflammation genes such as BST2 and TNFSF10 appeared to be more induced in committed subpopulations upon UM171 exposure. In contrast, several members of the TNF/NFkB/cAMP signaling network were mostly induced by UM171 in the primitive subpopulation: These genes, known to be involved in fine-tuning the inflammatory response, included BIRC2-3, DUSP4, PDE4B, TIPARP (Fig 1H and S1  Table).
We previously reported the optimal UM171 dose-response for HSC expansion at around 35nM [19,23]. Higher levels eventually lead to reduced cell proliferation and to elevated inflammation which is accompanied by a preferential reduction in cell proliferation in the primitive subset (see MKI67, GO cell cycle and CFSE labeling in S3A and S3B Fig). These "inflammation high / proliferation low" primitive cells failed to expand ex vivo (S3C and S3D Fig), potentially linking inflammation to reconstitution.
Given the observed specific inflammatory signature upon UM171 stimulation, we next assessed the possibility that this signature is secondary to induction (global profiling: S4A Fig and S1 Table)  Moreover, exposure of CD34 + CB cells to low (10ng/ml) or high (50ng/ml) doses of TNFa or IFNg did neither cause upregulation of EPCR nor CD86 (S4C Fig), demonstrating the complexity of the UM171 inflammatory response which cannot be recapitulated by these 2 canonical pro-inflammatory agonists

Proinflammatory signaling is essential for UM171 induced HSPC expansion
Dexamethasone (Dex) treatment was able to completely suppress the UM171 mediated expansion of HSPC-enriched cell subsets (CD34 + EPCR + CD90 + CD45RA -: see shNT panels in Fig  2A). This effect was specific to the glucocorticoid receptor as it was markedly attenuated in the absence of this receptor (see shGR panels in Fig 2A). Of interest, reduced HSPC expansion in response to Dex was also observed in the absence of UM171, indicating a general involvement of inflammation in HSCs (Fig 2B and 2C). Most importantly, Dex treatment also reduced the long-term repopulating ability of human HSCs and significantly abolished UM171-driven expansion of these cells (Fig 2D).
We also tested the combination of UM171 with the NFB inhibitor EVP4593 on CD34 + cells. We observed a marked decrease in frequencies and numbers of CD34 + CD90 + EPCR + and CD34 + CD86 + cells (Fig 2E and 2F), both in the presence or absence of UM171, suggesting that the expansion of these subsets depend on NFB signaling activation. Interestingly, NFkB inhibition significantly reduced EPCR gene expression in HSC enriched population (S5A Fig).
Consistent with these results, Dex as well as other anti-inflammatory inhibitors (JNK, NFkB and NFAT antagonists) suppressed UM171 mediated EPCR and CD86 induction in the OCI-AML5 cell line model (S5B- S5D Fig).
Since cyclic AMP acts as an important anti-inflammatory messenger by inhibiting NFkB signaling through PKA activation [25,26], we also tested the impact of several cAMP elevating agents on UM171-mediated cord blood expansion. In line with above results, we found that cAMP elevation, whether using a phosphodiesterase inhibitor (IBMX), an adenylate cyclase activator (Forskolin) or cell permeable cAMP (db-AMP), dramatically reduced the proportion CD34 + EPCR + and CD34 + CD86 + in UM171-supplemented cultures (S6 Fig). Taken together, these results demonstrate that activation of pro-inflammatory programs, most notably of NFkB, represents a core requirement for UM171-driven ex vivo HSC expansion.

EPCR attenuates UM171-mediated pro-inflammatory signals
We previously showed that EPCR expression is rapidly induced by UM171 in primitive CD34 + CB cells and that this EPCR-positive subpopulation includes most short-and longterm HSCs. Moreover, we showed that EPCR is essential for the in vivo activity of human HSCs [23]. Since anti-inflammatory functions connected to NFkB activation have been Heat-map of differentially expressed transcripts after 6, 24, 48 and 72hrs of UM171 treatment in OCI-AML5 cell line (>2-fold change, q-value < 0.0001 at any timepoint compared to untreated). B: Examples of pro-inflammatory (IL1b and CD86) and anti-inflammatory genes (PROCR and SOD1) expression changes upon UM171 treatment in OCI-AML5 cells. C: GSEA plots showing enrichment for genes associated to inflammatory responses in OCI-AML5 cell line (250nM, 6, 24, 48 and 72hrs exposure).D: Topology of primitive (red) and committed (blue) cell subsets on top of t-SNE projection in either the combined dataset (left panel) or separated based on treatment (right panels). Barplots next to t-SNE graphs summarize cellular compositions of individual treatments into primitive/committed subsets. E: t-SNE heatmap of representative stem cell associated genes PROCR (also called EPCR or CD201) and JAML; imputed data (MAGIC). F: Heatmap summaries (average values) of selected inflammation associated genesets (corresponding GSEA in S2E Fig). G: t-SNE heatmap and corresponding violin plots (lower panel) of representative inflammatory genes modulated in response to UM171 in both primitive and committed cells (BST2 and TNFSF10); imputed data (MAGIC). H: Violin plots illustrating expression distribution of candidate genes associated to TNFa/NFkB pathway across primitive and committed subsets. Note the preferential UM171 mediated modulation of these candidate genes in primitive subset. p�0.05. reported for EPCR [27][28][29], we hypothesized that it might partially antagonize the inflammatory response in cells expanded under UM171 treatment. To test this hypothesis, we targeted the EPCR gene in the OCI-AML5 cell line using CRISPR (Fig 3A-3C). Strikingly, we observed a strong and proinflammatory/NFkB signature specifically in UM171 treated cells that was further exacerbated in three independent sgEPCR OCI-AML5 clones (Fig 3A-3C and S1 Table).
Of interest, EPCR knockout OCI-AML5 cells were hypersensitive to the combination of TNFa and UM171 (Fig 3D, left panel) which induced a hyper inflammatory response as detected by nitric oxide (NO) production ( Fig 3D, right panel). In the context of primitive CD34 + CD45RAcord blood cells, UM171 and TNFa treatment also resulted in NO hyperproduction and cell loss when EPCR was experimentally reduced ( Fig 3E). Together, these observations demonstrate that, in addition to causing a pro-inflammatory response, UM171 also triggers a critical negative feedback loop mediated by EPCR that protects HSPCs from inflammation borne cytotoxicity.

UM171 activates a ROS detoxification program through EPCR
We noted that UM171 rapidly induced an EPCR-dependent ROS detoxification response in OCI-AML5 cells (blue rectangle in Figs  Most importantly, UM171 also reduced ROS levels in primary CD34 + CB cells treated for 24h with various ROS inducing agents such as etomoxir, rotenone, oligomycine, CCCP ( Fig 4C). At day 7, when expansion of functionally defined HSCs reaches its maximum [19,23], steady-state levels of intracellular ROS were significantly blunted by UM171 in CD34 + CD45RAcells in the presence of oligomycin (Fig 4D and 4E

Discussion
In this study, we provide insights into how UM171 promotes the expansion of HSCs by coordinating pro-and anti-inflammatory responses. Interestingly, whereas pro-inflammatory cytokines like interferons or TNFa and anti-inflammatory drugs like dexamethasone are unable to establish permissive conditions for HSC expansion in culture, a low dose of UM171 (35nM) has the unique capacity to induce a rheostatic regulation of inflammatory and anti-inflammation/detoxification programs, hence enabling an effective context to promote HSC self-were evaluated by flow cytometry. B: CD34 + cord blood cells were cultured for 7 days in presence of vehicle or Dex only (100nM). Representative FACS profiles show a reduction of HSCs (CD34 + EPCR + ) and progenitors (CD34 + and CD34 + CD45RA + ) in presence of Dex. C: CD34 + cord blood cells were exposed to DMSO, UM171 (35nM), Dex (100nM) or UM171 and Dex for 7 days and co-stained with CD34, EPCR and CD90 antibodies. Total cell counts and counts of HSC enriched subsets are presented. Data show mean ±SEM of 2 independent experiments performed in triplicate. D: Day 7 cultures were transplanted in immunocompromised NSG mice (outcome of 2000 day 0 cells). Human CD45 and CD34 engraftment were assessed at 26 weeks post-transplantation. E: CD34 + cord blood cells were cultures for 5 days in presence of DMSO, UM171 (35nM), NFKB inhibitor (EVP4593, 100nM) and UM171 + EVP4593. FACS profile (left panel) and absolute counts (right panel) of CD34 + EPCR + , CD34 + EPCR + CD90 + and CD34 + CD86 + cells are shown. Data show mean ±SEM of 2 independent experiments performed in triplicate. https://doi.org/10.1371/journal.pone.0224900.g002 Pro-and anti-inflammatory signals determine human hematopoietic stem cell expansion renewal ( Fig 4H). This inflammatory response is observed within 6 hours of UM171 treatment and does not appear secondary to pro-inflammatory cytokines such as TNFa. Our results thus suggest that UM171 establishes a dosage-dependent and tightly regulated inflammatory tonus that equally relies on positive regulators such as NFkB and their integration in a negative feedback loop which prevents toxic accumulation of ROS/inflammation that is executed through upregulation of EPCR ( Fig 4I). We also show that several of these components are active in cells not exposed to UM171, further extending these observations to HSC self-renewal networks. Future work will determine the mediator(s) of this specific inflammatory response.
Supporting a role for pro-inflammatory signaling in HSC self-renewal, exposure of cord blood cells to immunosuppressors such as glucocorticoids led to a marked reduction of HSC repopulating ability and abolished UM171-driven HSC expansion in a glucocorticoid receptor dependent manner. These observations contradict recent findings showing a beneficial effect of glucocorticoids on HSC engraftment [30]. It is important to note that duration of glucocorticoid exposure was different in the 2 studies. While a 16 hour exposure led to an enhanced HSC engraftment [30], longer treatment as performed in the present study had detrimental effects on HSC function. In line with this, while chronic low doses of the pro-inflammatory TLR ligand LPS dramatically impaired HSC function [31], short-term treatment with higher doses of LPS led to increased HSC multi-lineage repopulation ability [32]. It appears therefore that the effects of inflammation on HSCs activity are dose and time-dependent. Consistent with this, while transient NFkB blockade (6 to 24 hours) enhances ex vivo HSC propagation [33], a minimal and critical threshold of NFkB activation is required to maintain HSC homeostasis [34].
Interestingly, HSPCs and committed cells seemed to adapt a distinct inflammatory response upon UM171 treatment. Of interest, many of the differentially modulated genes in primitive cells (such as DUSP4 and BIRC2/3) are known to act as molecular brake in inflammation. It is thus tempting to speculate that HSPCs have established a fine-tuning inflammatory program to mount a robust protective response and avoid self-destructive inflammation. It will be critical in future works to dissect more precisely these inflammatory programs.
In summary, and supporting the work of several ongoing investigations, our current work supports a central role for inflammation in adult stem cell self-renewal, with UM171 providing a favorable state for blood stem cells renewal in which appropriate pro-and anti-inflammatory/detoxifying conditions are achieved.

Human CD34 + cord blood cell collection
Umbilical cord blood units were collected from consenting mothers according to ethically approved protocol at Charles-Lemoyne Hospital, Montreal, QC, Canada. Human CD34 + cord blood (CB) cells were isolated using The EasySep™ positive selection kit ( StemCell   Fig 3. EPCR attenuates UM171-mediated pro-inflammatory signals. A: Heatmap representing differentially expressed genes after a 24 hours exposure to UM171 (250nM) in wild-type (sgAAVS1) and EPCR null (sgEPCR) context (>2-fold change, q-value < 0.0001 compared to DMSO/ sgAAVS1). Note that while UM171 mediated pro-inflammatory signaling (red labels) is exacerbated in sgEPCR transduced cells, UM171 mediateddetoxification responses (blue labels) is hampered in absence of EPCR. B: Heatmap of NFkB target genes (http://bioinfo.lifl.fr/NF-KB) after a 24 hours exposure to UM171 (250nM) in wild-type (sgAAVS1) and EPCR null (sgEPCR) context. C: Examples of pro-inflammatory genes expression changes upon UM171 treatment (dashed lines show differential response to UM171 in EPCR null vs wild-type context after 24-hour exposure). D: OCI-AML5 were transduced with shRen (CT) or shEPCR (Ametrin) lentiviral vectors and culture for 4 days in presence of DMSO, UM171, TNF (10ng/ml) or UM171+ TNFa. Cell viability (left panel) and nitric oxide (NO) generation (right panel) were assessed by flow cytometry. E: CD34 + cord blood cells were transduced with shNT or shEPCR (Ametrin) lentiviral vectors and culture for 7 days in presence of DMSO, UM171, TNFa (10ng/ml) or UM171 + TNFa. Percentage of transduced Ametrin + cells (right panel) and levels of nitric oxide (NO) (left panel) were assessed within the CD34 + CD45RA -(HSPC enriched) population by flow cytometry. Data are expressed as mean ± SEM of results from 2 cord blood donors performed in 3 replicates. https://doi.org/10.1371/journal.pone.0224900.g003 Pro-and anti-inflammatory signals determine human hematopoietic stem cell expansion Technologies Cat # 18056). Sorting for more primitive phenotypes was done in additional step using BD Aria II sorter.

Cell Lines
AML cell lines were purchased from the DSMZ German collection of Microorganisms and cell culture (Leibniz Institute) and the ATCC. OCI-AML3 and OCI-AML5 cells were cultured in MEM alpha media with 10% FBS and supplemented with 10 ng/mL of GM-CSF (PeproTech). NB4 and THP-1 cell lines were cultured in RPMI 1640 medium with 10% fetal bovine serum.

Single cell RNA sequencing, population mapping and data analysis
CD34+ cord blood (CB) cells were cultured for 48h, a timepoint before most cells undergo division. Experimental conditions included either DMSO or two different UM171 concentrations; 35nM, a dose previously established as optimal for HSC expansion [19] as well as 1000 nM. Single-cell RNAseq from each of these cultures was performed on a Chromium Single-Cell Controller (10X Genomics) using the Single Cell 3' Reagent Kit version 2 according to manufacturer's instructions. Target cell numbers were 6,000 per condition. Sequencing of the scRNAseq libraries were performed on a NovaSeq device using a S2 (PE 28x91) setup. A standard Cellranger v3.0.1 pipeline was used for read mapping (GRCh38 annotation) and demultiplexing. Subsequent analyses were done in Seurat (v2.3) [35] and included (i) exclusion of cells with less than 1,000 genes or unique molecular identifiers (UMI), (ii) exclusion of cells with more genes or UMIs than the respective means plus 2 standard deviations (likely representing multiplets), (iii) exclusion of cells with more than 6% mitochondrial gene expression (representing apoptotic cells). Expression counts were log-normalized (scale factor 10,000) and scaled including regression on number of UMI's, cell cycle scores and mitochondrial gene content. Multi-set Canonical Correlation Clustering and tSNE embedding was performed using variable genes (defined by mean.function = ExpMean, dispersion.function = LogVMR, x.low. cutoff = 0.02, x.high.cutoff = 5, y.cutoff = 2) excluding sex-specific genes. Average fold-change and p-values were obtained using the FindMarkers function in Seurat with logFC threshold and min.pct set to zero. Log2 fold-change ranked gene lists were used for subsequent GSEA analyses. Significantly regulated genes were defined by p-values smaller than 0.01 and 1.5-fold

UM171 activates a ROS detoxification program through EPCR. A: Assessment of total ROS generation (left panels) in OCI-AML5 cells (as measured by
DCFDA relative mean fluorescence intensity) and EPCR levels (right panels) at 30 minutes (upper panels) and 24h (lower panels) after addition of increasing UM171 concentration. B: Fold-change in expression of ROS detoxifying enzymes after exposure to UM171 in EPCR wild-type (sgAAVS1, black histograms) or EPCR null (sgEPCR, red histograms) context. C: CD34+ cord blood cells were exposed to DMSO or UM171 (35nM) for 24hrs and treated with different ROS-generating agents for 3 additional hours. ROS generation was then assessed by flow cytometry. D: CD34+ cord blood cells were exposed for 7 days to DMSO or UM171 (35nM) ± oligomycin. Mitochondrial ROS were then assessed in HSC enriched CD34+CD45RA-subset by flow cytometry using mitosox staining. Total and CD34+CD45RAabsolute cell counts are presented in E. F: CD34 + cord blood cells were cultured for 4 days in presence of DMSO or increasing dose of UM171, or pro-inflammatory cytokine TNFa (10ng/ml) or anti-inflammatory drug dexamethasone (100nM). ROS generation generated in each condition was evaluated using DCFDA staining. Representative FACS profile show ROS production in CD34+EPCR+ (HSC-enriched (blue)) vs CD34+CD45RA+ (progenitors (red)) subsets. G: Pro-and anti-inflammatory signals determine human hematopoietic stem cell expansion down or upregulation, respectively. For visualization purposes (S2B- S2D Fig and Fig 1E and  1G), data imputation was done using MAGIC (https://github.com/KrishnaswamyLab/ MAGIC) with t = 1 [36].
Initial population mapping was performed based on expression of the stem and progenitor associated genes AVP and HLF [23] (see S2B Fig, HSPC panel), of genes indicative of early lymphoid fate (SPINK2 and SELL; LMPP panel) and selective expression of myeloid, erythroid and megakaryocytic markers which were located to the periphery of the t-SNE projection (S2B Fig).
We next ranked cells using a stem-score (S2C Fig, top heatmap) that was calculated by averaging the normalized expression vectors of twelve commonly accepted stem cell specific genes [23,[37][38][39] (S2C Fig, middle heatmap). As expected, expression of differentiation associated genes anticorrelated with this ranking metric (S2C Fig, lower heatmap). Next, we categorized the top 5% stem-score ranking cells as 'primitive' and the lowest 20% as 'committed' subsets ( Fig 1D, bottom). Notably, t-SNE representation of these designated populations recapitulated a clear spatial distinction between HSPCs in the top central (Fig 1D, in red) and committed/ differentiated populations in the lower periphery of the projection space (Fig 1D, in blue). Albeit with small changes in frequency compared to the combined dataset, primitive and committed populations distributed well into all three experimental conditions (Fig 1D, middle  panel).

Transplantation assays
All experiments with animals were conducted under protocols approved by the University of Montreal Animal Care Committee. EPCR cell subsets purified from uncultured or expanded CD34 + CD45RA -CB cells were transplanted by tail vein injection into sub-lethally irradiated (250 cGy, <24 hr before transplantation) 8 to 16-week-old female NSG (NOD-Scid IL2Rgnull, Jackson Laboratory) mice. Human cells NSG-BM cells were collected by femoral aspiration or by flushing the two femurs, tibias and hips when animals were sacrificed at week 26.

Nitric oxide production
Intracellular NO was measured after staining for cell surface markers, by incubation of cells for 15 min at 37˚C with 10 microM DAF-FM diacetate followed by extensive washes, according to the manufacturer's instructions (Molecular Probes).

Cytokine assays
Cytokine levels in day 4 DMSO or UM171-exposed CD34+ cord blood cultures were measured using the LEGENDplex Human Inflammation Panel (13-plex) according to manufacturer's instructions. Data was analyzed using the LEGENDplexTM Data Analysis Software. In some conditions, cytokines secretion was also assessed after 4 hours PMA/ionomycin (20 ng/ ml, 500 ng/ml) stimulation.

Bulk RNA sequencing and data analysis
3-5 x 10 5 cells were FACS sorted from day 7 cultured cord blood derived CD34 + HSPCs and preserved at -80 C in TRIzol Reagent (Thermo Fisher Scientific Cat # 15596026). cDNA libraries were constructed according to TruSeq Protocols (Illumina) and sequencing was performed using an Illumina HiSeq 2000 instrument. Gene expression statistics were obtained using the kallisto/sleuth analysis pipeline (https://pachterlab.github.io/sleuth/about) and the GRCh38 version 84 annotation. Differential gene expression was determined using sleuth p-values and fold-change in transcripts per million (TPM) values as designated. GSEA analysis was done using the fgsea R package and the full Molecular Signatures Database (MSigDB, Broad Institute).

Accession codes
Gene Expression Omnibus: GSE57561, GSE138487 and GSE138680 Supporting information S1 Fig. UM171 induces upregulation of EPCR and CD86 in leukemic cell lines. A: Representative FACS profiles of CD86 and EPCR co-expression in monocytic derived cell lines exposed to DMSO or UM171 (500nM) for 24h. Various myeloid derived cell lines were screened for the upregulation of both EPCR and CD86 in response to UM171. Among them, acute myeloid leukemia cell lines OCI-AML3 and OCI-AML5 (FAB M4), promyelocytic leukemia NB4 (FAB M3) and monoblastic leukemia THP-1 (FAB M5) are shown. Note that OCI-AML5 cell line was used for all further studies as it shows the most consistent and highest response to UM171. Gene family annotations were downloaded from HUGO gene nomenclature committee (www.genenames.org). B: Amounts of pro-inflammatory cytokines IL1b, TNFa, IFNa2 and IFNg were measured by flow cytometry (LegendPlex) in day4 DMSO or UM171 exposed CD34+ culture media. Note that secretion of these pro-inflamatory cytokines were not induced by UM171 even after PMA/ionomycin stimulation. C: CD34 + cord blood cells were cultured for 4 days in presence of DMSO or UM171 (35 and 1000nM), or pro-inflammatory cytokine TNFa (10 and 50ng/ml) or IFNg (10 and 50ng/ml). CD34, EPCR and CD86 surface expression were assessed by flow cytometry. Representative FACS profile (upper panels) showing % of CD34+EPCR+ and CD34+CD86+ subsets and absolute counts (lower panels) of indicated populations in each condition. OCI-AML5 cells stably expressing Dox-inducible MitoTimer vector were exposed to increasing doses of UM171 for 24hrs and treated with Dox for 3hrs. Cells were the analysed by flow cytometry. Data show representative dot-plot profile of MitoTimer expressing cells (yaxis, green channel; x-axis, red channel). Note that while low dose of UM171 (125 to 500nM) reduces signals in red channel (consistent with lower ROS and improved mitochondrial quality), high dose of UM171 (above 1microM) increase ROS level (enhanced red signal). B: Representative FACS profile of ROS production at day 7 in CD34+ cells exposed or not to UM171 ± oligomycin. C: CD34+ cord blood cells were exposed for 7 days to DMSO or UM171 (35nM) in mild hyperthermia condition. Mitochondrial ROS were then assessed in HSC enriched CD34+CD45RA-subset by flow cytometry using mitosox staining. (TIF) S1 Table. AML5 and CD34+ Transcriptome and single cell RNAseq data. (XLSX)