key: cord-0271788-n6wl3l0l
authors: Calzetti, Federica; Finotti, Giulia; Tamassia, Nicola; Bianchetto-Aguilera, Francisco; Castellucci, Monica; Cavallini, Chiara; Mattè, Alessandro; Gasperini, Sara; Benedetti, Fabio; Bonifacio, Massimiliano; Tecchio, Cristina; Scapini, Patrizia; Cassatella, Marco A.
title: Identification and characterization of human CD34+ and CD34dim/- neutrophil-committed progenitors
date: 2021-04-30
journal: bioRxiv
DOI: 10.1101/2021.04.30.442138
sha: a36e937d5a7e5a316b2e32f142ebd98d9efabeca
doc_id: 271788
cord_uid: n6wl3l0l

We report the identification of human CD66b−CD64dimCD115− neutrophil-committed progenitors within SSClowCD45dimCD34+ and CD34dim/− bone marrow cells, that we named neutrophil myeloblast (NMs). CD34+ and CD34dim/−NMs resulted as either CD45RA+ or CD45RA−, with CD34+CD45RA−NMs found as selectively expanded in chronic-phase chronic myeloid leukemia patients. By scRNA-seq experiments, CD34+ and CD34dim/−NMs were found to consist of combinations of four cell clusters, characterized by different maturation stages and distributed along two differentiation routes. Cell clusters were identified by neutrophil-specific gene profiles, one of them associated to an interferon-stimulated gene (ISG) signature, hence supporting recently identified expansions of mature neutrophil subsets expressing ISGs in blood of diseased individuals. Altogether, our data shed light on the very early phases of neutrophil ontogeny.

Recent studies, performed by single-cell sorting, transcriptional analysis, single-cell transplants and clonal tracking analysis [1] [2] [3] [4] [5] [6] [7] [8] [9] , have challenged the classical hierarchical tree-like model of hematopoiesis 10, 11 . Accordingly, hematopoietic stem cells (HSCs) are currently considered as a heterogeneous cellular population in terms of lineage potential and transcriptional profile, instead of multi-potent, homogeneous clusters of cells 12 . In fact, it has been ascertained that hematopoiesis occurs as a continuous process along developmental trajectories following early erythroid/megakaryocyte/eosinophil/basophil and lympho/myeloid, ultimately generating mature, terminally differentiated cells. Moreover, transcriptomic profiling demonstrates that intermediate compartments of progenitors, such as common myeloid progenitors (CMPs) or granulocytemacrophage progenitors (GMPs) 10 , are largely composed of clusters of cells displaying lineageselective commitment. However, even though scRNA-seq experiments are necessary to reveal their heterogeneity 13 , the various cellular subpopulations are still identified by flow-sorting and immunophenotyping approaches.

In this scenario, conventional GMPs (cGMPs) represent the most restricted pool of myeloid progenitors in humans 14, 15 . In fact, based on flow cytometry approaches, cGMPs have been fractionated into three main compartments 16 , namely: i) granulocyte-monocyte-DC progenitors (GMDPs), that generate CD66b + granulocytes, CD14 + monocytes and the three dendritic cell (DC) subsets (DC1s, DC2s and pDCs); ii) monocyte-DC progenitors (MDPs), that generate DCs and CD14 + monocytes; iii) common dendritic progenitors (CDPs), that are exclusively committed to DCs.

By similar methodologies, but using a different panel of markers including CD64, common monocyte progenitors (cMoPs), exclusively differentiating into pre-monocytes and ultimately into CD14 + monocytes, were subsequently identified within cGMPs, while the existence of revised GMPs (rGMPs), generating only CD66b + granulocytes and CD14 + monocytes, was also postulated 17 . In this context, however, little progress has been made to characterize the ontogeny of human neutrophils, even though neutrophil -but not basophil or eosinophil -progenitors, are included in cGMPs 3, 4, 6, [18] [19] [20] . For instance, a CD34 + CD115 -CD64 + fraction able to originate granulocytes was identified within the fetal bone marrow (BM) in 1996 21 , without, however, discriminating their eventual eosinophil, basophil or neutrophil nature. Similarly, the phenotypic and/or morphological features of the CD66b + granulocytes originated by GMDPs 16 and rGMPs 17 were not exhaustively pursued. More recently, various human neutrophil progenitors have been described, precisely preNeus 22 , hNePs 23 , proNeus 24 and eNePs 25 . However, all of them express the CD66b and CD15 lineage markers, and thus resemble the promyelocyte (PM), or more mature neutrophil, stage. By contrast, in mice, CD34 + Ly6C + CD115 -proNeu1s have been found within GMPs 24 , thus representing the earliest identifiable progenitors along the neutrophil maturation trajectory currently identified.

Based on these premises, herein we report the identification, phenotypic characterization and single cell transcriptomic analysis of previously undescribed, human uni-lineage CD34 + and CD34 dim/neutrophil progenitors, which we named as neutrophil myeloblast (NMs).

To identify progenitors of human neutrophils at earlier stages than PMs, we focused on SSC low CD45 dim cells present in the low-density cells of BM (BM-LDCs) (panel III of Figure 1A , and panel V of Figure S1A ). SSC low CD45 dim cells are lineage negative ( Figure S1B ) and include CD34 + CD38 -HSCs, other than CD34 + and CD34 dim/myeloid/lymphoid progenitors 26, 27 . Moreover, although CD34 dim/cells represent neglected, but important, transitional progenitor stages 4 , they have been rarely investigated for a potential ability to generate neutrophils. Therefore, we stained BM-LDCs by a flow-cytometry antibody panel comprising key markers conventionally used to detect either BM or cord blood (CB) CD34 + myeloid and lymphoid progenitors, namely CD34, CD38, CD10, CD123, CD45RA, CD64 and CD115 14, 15, 21, 28 . By doing so, we could identify a lineage negative, SSC low CD45 dim CD10 -CD38 + region displaying variable CD34 and CD45RA levels (panel V of Figure 1A ), and including CD34 + CD45RA -, CD34 + CD45RA + , CD34 dim/-CD45RA + and CD34 dim/-CD45RAcell populations. It is noteworthy that CD34 + CD45RA + cells were also found CD135 + (data not shown), and thus represent cGMPs (panel V of Figure 1A ) 15 , while CD34 + CD45RAcells (panel V of Figure 1A ) include more immature progenitors, such as CMPs and MEPs 14, 15 . We then focused on the cGMP region and subdivided it into a total of five discrete cell populations based on their CD123, CD115 and CD64 expression. Precisely: i) CDPs (panel VI of Figure 1A ), which are strictly CD64 -(data not shown); ii) CD64 + CD115 + populations, resembling the monocyte-committed progenitor population (CFU-M) 21 , as well as cMoPs 17 , that we provisionally named cMoP-like cells (panel VII of Figure 1A ); iii) CD64 -CD115 + populations resembling MDPs 16 , that we provisionally named MDP-like cells (panel VII of Figure 1A ); (iv) CD64 -CD115populations resembling the conventional GMDPs 16 , that we provisionally named CD64 -GMDP-like cells (panel VII of Figure 1A ); and v) CD64 dim CD115cells, resembling previously described granulocyte progenitors 21 (panel VII of Figure 1A ). CD34 + CD45RA + CD64 dim CD115cells are neutrophil-restricted progenitors Recent studies reported that neutrophils, but not basophils/eosinophils, derive from cGMPs 4, [18] [19] [20] . Therefore, to investigate whether CD34 + CD45RA + CD64 dim CD115cells represent neutrophilrestricted progenitors, we seeded them on top of MS-5 cells 29 for seven days, in a medium containing a cocktail of SCF, Flt3L and G-CSF (SFGc), and, in turn, found them to give origin mostly to CD66b + cells, and poorly to monocytes ( Figure 1B and 1C, top panels). By contrast, SFGc-treated cMoP-like cells were found to differentiate mainly into CD14 + monocytes, and minimally to CD66b + cells and DCs ( Figure 1B and 1C, middle panels), confirming their commitment to monocytes 17 . Conversely, SFGc-treated MDP-like cells were found to originate CD141 + CD14 -cDC1, CD303 + CD14 -pDCs, CD14 dim/-CD1c + cDC2 and CD14 + monocytes ( Figure 1B and 1C, bottom panels), consistent with their predicted heterogeneous composition 16, 30 . Moreover, SFGc was found to fully maintain the viability of progenitor-generated cells after 7 days ( Figure S2A ), as well as to potently promote the expansion of CD34 + CD45RA + CD64 dim CD115cells compared to cMoP-like and MDP-like cells ( Figure S2B ), consistent with its high specificity for the neutrophil lineage, and with the highest expression of CD114/G-CSFR in CD34 + CD45RA + CD64 dim CD115cells ( Figure S2C ). Phenotypic ( Figure 1D ) and morphologic ( Figure 1E ) analysis of CD66b + cells derived from SFGc-treated CD34 + CD45RA + CD64 dim CD115cells revealed that they mostly consist of immature neutrophils, being composed of: CD11b -CD16 -PMs; CD11b dim/+ CD16myelocytes (MYs); CD11b + CD16 + metamyelocytes (MMs); CD11b + CD16 ++ CD10band cells (BCs); and CD11b + CD16 ++ CD10 + segmented neutrophils (SNs) (for the respective percentages, see also second row of Figure 2D ).

Consistently, CD34 + CD45RA + CD64 dim CD115cells incubated with SFGc for 14 days originated more mature stages of neutrophils, such as BCs and SNs ( Figure 1F ), as also confirmed by the morphology of their nuclei ( Figure 1G) . Moreover, specificity of lineage commitment to neutrophils was demonstrated by culturing CD34 + CD45RA + CD64 dim CD115cells with SCF, Flt3L and GM-CSF (SFG) ( Figure S2D ), a condition that induced monocytes or monocytes and DCs by, respectively, cMoP-like and MDP-like cells ( Figure S2E and F), as reported for cMoPs and MDPs 17, 29 .

Collectively, these experiments formally prove that CD34 + CD45RA + CD64 dim CD115cells represent neutrophil-restricted progenitors. Data also demonstrate that the CD34 + CD45RA + CD64 + CD115 + cMoP-like and the CD34 + CD45RA + CD64 -CD115 + MDP-like cells substantially resemble, respectively, the conventional cMoPs 17 and MDPs 16 , and thus they will be referred to as such.

Identification of additional CD34 + and CD34 dim/neutrophil-restricted progenitors

Since neutrophil-committed progenitors have been suggested to stand within the CD45RAcompartment 3, 20 , we then wondered whether other, not yet described, neutrophil-restricted progenitors could be identified within the SSC low CD45 dim region (panel I of Figure 2A ). Therefore, we searched for cells expressing the CD64 dim CD115phenotype in all the fractions delimited by the CD34/CD45RA marker combination (panel III of Figure 2A ), and ultimately found three more cell populations (orange, magenta and light blue gates in, respectively, panel IV, VI and VII of Figure   2A ). Moreover, we confirmed the presence not only of the previously described CD34 + CD45RA + CD64 dim CD115neutrophil-restricted progenitors (green gate), cMoPs and MDPs within cGMPs (panel V of Figure 2A ), but also of CD64 ++ CD115 + pre-monocytes 17 in CD34 dim/-CD45RA + cells (panel VI of Figure 2A ). Notably, by applying an identical flow cytometry gating strategy to either cord blood samples (CB) or spleen biopsies, we were able to identify SSC low CD45 dim CD123 dim/-CD10 -CD34 + and CD34 dim/cells, as well as to successfully uncover, even in these specimens, CD64 dim CD115neutrophil-restricted progenitors within the CD45RA + and CD45RAcells (data not shown). Hence, we renamed these CD34 + and CD34 dim/neutrophilrestricted progenitors as neutrophil myeloblasts (NMs), and subdivided them, on the basis of their CD34 and CD45RA expression, as NM1s (orange gate in panel IV), NM2s (green gate in panel V),

NM3s (magenta gate in panel VI) and NM4s (light blue gate in panel VII) for, respectively, CD34 + CD45RA -, CD34 + CD45RA + , CD34 dim/-CD45RA + and CD34 dim/-CD45RAcells.

Then, we sorted CD34 + and CD34 dim/-NMs to analyze their morphology and developmental potential. Morphologically ( Figure 2B ), the four NMs were found to display a high nuclear/cytoplasmic ratio compatible with a progenitor identity. Moreover, granules ( Figure 2B) were evident only in NM3s and NM4s, but at lower levels than in PMs. In addition, SFGc-treated NM1s and NM2s were found to display a comparable ( Figure 2C ), but higher proliferative potential than that exhibited by SFGc-treated NM3s and NM4s ( Figure 2C) . Notably, the four NMs were found to almost exclusively produce neutrophils at different stages of maturation, as determined by their phenotype and morphology ( Figure 2D ). By the latter criteria, in fact, we found no eosinophils and/or basophils among the SFGc-generated cells (data not shown), and even if the four NMs were incubated in SFG plus IL-3 and/or IL-5 18 . Moreover, all NMs generated no DCs when cultured in SFGc (data not shown), while exclusively NM2s and NM3s generated a few CD14 + monocytes (data not shown), which we subsequently found out to actually derive from contaminating unavoidably sorted cMoP and pre-monocyte populations (see paragraph on single cell RNA data). Importantly, neutrophils generated by NMs displayed respiratory burst ability ( Figure S3A ), as well as phagocytosis capacity ( Figure S3B ) comparable to those displayed by peripheral mature neutrophils.

Finally, we found that, unlike NM1s and NM2s, NM3s and NM4s express CD15, although at lower levels than PMs and mature SNs ( Figure S3C ). The latter data exclude that NMs could relate to the SSC int Lin -CD34 -CD15 int CD11b -CD16 -"early" PMs (EPMs) 31 . In our hands, in fact, EPMs were found to partially overlap not only with both NM3s and NM4s, but also with pre-monocytes (data not shown), which express CD15 at similar levels of those by cMoPs and mature monocytes ( Figure   S3D ). It is hence evident that that the sole CD15 does not function as a marker specifically identifying neutrophil progenitors.

In sum, our experiments have identified CD34 + and CD34 dim/-NMs standing prior PMs along the neutrophil maturation cascade, and specifically characterized by the CD64 dim CD115phenotype, but differentially expressing CD45RA. The features of CD34 + and CD34 dim/-NMs are summarized in Table S1 .

The relative content of NMs, MDPs, cMoPs and CDPs is altered in chronic-phase chronic myeloid leukemia (CP-CML) patients CML is a myeloproliferative disorder evolving in three clinical stages, known as CP-CML, accelerated phase CML (AP-CML) and blast phase CML (BP-CML) 32 . CP-CML stands out for a dramatically increased peripheral white blood count (WBC) (i.e., higher than 10 10 /L), which mainly reflects a remarkable rise of both the absolute number and the relative percentage of neutrophils.

Inexplicably, a dramatic decrease of the cGMP percentage has been reported to occur in CP-CML patients [33] [34] [35] , which is in contrast with the classical model of myeloid development centered on cGMPs as the main source of neutrophils.

To try clarifying such an issue, we set up a 14-color antibody panel allowing a more clear identification not only of the SSC low Lin -CD45 dim region, but also of the various neutrophil progenitors, starting from their earliest CD34 + ones, and up to mature, segmented neutrophils ( Figure   S4A and S4B). This approach turned out to be fundamental to analyze CP-CML pathological samples. By this analysis, we confirmed that cGMPs from CP-CML patients decrease ( Figure 3A) and result almost negative for CD115 expression ( Figure 3B ), therefore reflecting an altered ratio between neutrophil and mono-DC progenitors ( Figure 3C ). Precisely, we found the percentage of MDPs, cMoPs and CDPs as significantly lower in CP-CML patients than in HDs ( Figure 3C ), in the presence of a significant expansion of NM1s and of an unaltered NM2 frequency ( Figure 3C ).

Therefore, while confirming the presence of NMs in CML patients, these data imply that the downregulation of cGMP percentage in CP-CML patients depends on a marked reduction of monocyte/DC progenitors but not of NM2s. Data also support the notion that the selective expansion of neutrophils in CP-CML patients derives from NM1s.

RNA-seq experiments confirm that NMs precede PMs along the neutrophil maturation cascade.

NMs, as well as BM HSCs, PMs, MYs, MMs, BCs, SNs and mature neutrophils (PMNs) were then profiled by RNA-seq. Both PCA (shown in Figure 4A ) and hierarchical clustering analysis performed by optimal leaf ordering (OLO) 36 ( Figure 4B ) confirmed that NMs not only cluster by themselves, but are also placed along the maturation trajectory which from HSCs, via PMs, MYs, MMs, BCs and SNs, ends into PMNs. Interestingly, analysis of CD34 and PTPRC (encoding for CD45) transcript expression ( Figure 4C and 4D) corroborated the flow cytometry data. Furthermore, by performing K-means clustering of the differentially expressed genes (DEGs), ten main gene groups (g1-g10) were identified among all samples ( Figure 4E and Table S2 ). Genes encoding markers of immature cells (such as CD34, HOXA9, MYC, FLT3, SOX4 and KIT), expressed only by HSCs and NMs, were found present in g1 or g2 ( Figure 4E ). Interestingly, GO analysis of the g1 and g2 DEGs revealed the "MHC class II protein complex" term among the most significant overrepresented ( Figure S5A ), in line with a previously described MHC class II expression in very immature neutrophil progenitors 37 . Similarly, DEGs mainly involved in ribosome assembly, mitochondria formation and cell cycle regulation, characterizing g3, g4 and g5, respectively, were found expressed in NMs, PMs, MYs and MMs, but not in non-proliferating BCs and SNs ( Figure   4E ,F and Figure S5A ).

Importantly, g5 was also found to contain DEGs encoding the first granule mRNAs transcribed during granulopoiesis, namely those for the azurophilic granule (AG) proteins (such as MPO, AZU1, PRTN3 or ELANE), as well as transcription factors (TFs) known to be typically expressed in immature neutrophils, such as GFI1 and CEBPE 38 ( Figure 4E ). Focused/specific analysis on AG genes expression revealed that these genes start to be expressed in NM1s and NM2s, increase in NM3s and NM4s, are maximally transcribed in PMs and MYs and then gradually disappear ( Figure 4F ). By contrast, g6, g7 and g8 were found to include DEGs absent in NMs, being enriched for specific granule (SG) (LYZ, LTF, LCN2 and CEACAM8/CD66b), gelatinase granule (GG) (MMP9, ARG1 and CD177), secretory vesicles (SV) (e.g., FCGR3B/CD16B and ANXA1) and

cell membrane proteins (CM) (e.g., CXCR1, CXCR2 and ICAM3) mRNAs ( Figure 4E ,F) and thus associated to "neutrophil degranulation" and "neutrophil activation" GO terms ( Figure S5A ). Finally, g9 and g10 were found enriched in IFN-stimulated genes (ISGs), as well as other genes associated to the "NF-kB signaling" and "cytokine production" GO terms, and all of them present in BCs, SNs and PMNs ( Figure 4E and Figure S5A ). Consistent with the literature 39 , mRNAs encoding proteins involved in the production of reactive oxygen species (ROS) ( Figure S5B ), phagocytosis ( Figure   S5C ) and chemotaxis ( Figure S5D ) were found transcribed starting from the MM/BC stages, and highly expressed in mature neutrophils.

To get more insights into the specific transcriptomic differences among NMs only, we performed DEG analysis by using the likelihood ratio test (LRT) 40 and identified 1114 DEGs among them. As shown in Figure S5E , the four NMs were found to distinctly segregate from each other by PCA, indicating remarkable differences among their gene expression profiles. However, while PC1 differences were mostly determined by cell cycle and AG genes, which were more expressed in, respectively, NM1s/NM2s and NM3s/NM4s ( Figure S5E ), genes mostly contributing to the PC2 variations, namely PRG2, CLC, EPX and IL5RA mRNAs found in NM1s, or IRF8, ANXA2, SAMHD1, SLAMF7, LYZ and F13A1 mRNAs found in NM2s and NM3s ( Figure S5E ), were unexpected since typically expressed by progenitors of, respectively, eosinophils or monocytes. All in all, RNA-seq data not only confirm that NMs are committed to neutrophils, but also that they are placed prior to PMs along the neutrophil maturation cascade.

To unequivocally dissect their transcriptional heterogeneity, we performed scRNA-seq experiments of sorted NM1s, NM2s, NM3s, NM4s, and (to elucidate their relationship with NMs) cMoPs. We sequenced 17902 cells and, by performing dimensionality reduction by Uniform

Manifold Approximation and Projection (UMAP) 41 , it immediately emerged that NMs clearly segregate from and cMoPs, while NM1s and NM3s mostly overlap with, respectively, NM2s and NM4s ( Figure 5A and Figure S6A ). Then, by performing unbiased, graph-based, clustering by Seurat 42 , we could identify 9 discrete cell clusters ( Figure 5B ), of which three (c7, c8 and c9) were excluded from all subsequent analyses because consisting of: i) few eosinophil progenitors (i.e., c7, identified by EPX mRNA expression); ii) few megakaryocyte/erythroid progenitors (i.e., c8, identified by PF4 and HBB mRNA expression); iii) few cells with an unclear ontogeny and high expression of apoptosis-related genes (i.e., c9) ( Figure 5B ). Of the six remaining cell clusters, by far the most rich of cells, four were unequivocally attributable to the neutrophil lineage (i.e., c1, c2, c3 and c4), and two to the monocyte lineage (i.e., c5 and c6) ( Figure 5B ). Such segregation was also confirmed by hierarchical clustering analysis of their DEGs ( Figure S6B ), since genes typical of the neutrophil (such as CEBPE or GFI1) and monocyte (such as IRF8 or CSF1R) lineages were exclusively expressed in, respectively, c1-c4 and c5-c6 cells ( Figure 5C ). Moreover, distribution analysis of scRNA-seq cell clusters in NMs and cMoPs evidenced that c1-c4 cells were completely absent in cMoPs (thus proving that CD115 expression efficiently discriminates NMs). By contrast, a quote of c5 cells were found to contaminate both NM2s and NM3s, but not NM1s and NM4s ( Figure   5D ), therefore explaining the PCA results on bulk RNA-seq shown in Figure S5G .

By selectively focusing on c1-c4 cells, it resulted evident that their distribution in NM1s and

NM4s is specular to those of, respectively, NM2s and NM3s ( Figure 6A ). Moreover, while NM1s and NM2s were found enriched of c1 cells mainly, NM3s and NM4s were found to preferentially accumulate c3 and c4 cells ( Figure 6A ). By contrast, c2 cells were found to distribute among the four NMs at substantially similar levels ( Figure 6A ). Then, calculation of the pseudotime value for every cell by Destiny 43 ( Figure 6B ), to define their maturation state, uncovered that c1 contains the more immature, while c3 and c4 contain the relatively more mature, cells ( Figure 6C ). Accordingly, CD34

was found preferentially expressed in c1 and c2 cells ( Figure 6D ), consistent with their relatively more abundant presence in NM1s and NM2s ( Figure 6A ). c2 cells, instead, were found to display variable pseudotime values ( Figure 6C ), in line with their presence in all NMs ( Figure 6A) .

Assessment, by Seurat, of the DEGs characterizing c1-c4 cells resulted in the identification of 136 genes ( Figure 6E and Table S3 ). In such regard, c1 cells were found to express high levels of typical genes of immature proliferating cells (such as CD34, TOP2A, TUBB and HIST1H4C) ( Figure 6E and Figure S6C ,D), consistent with their lowest pseudotime values ( Figure 6C ). c1 cells were found to express also genes associated to the "MHC class II protein complex" GO term ( Figure S6E ), consistent with the g2 genes from bulk RNA-seq data ( Figure S5A ). By contrast, c3 and c4 cells were found to express high levels of AG genes ( Figure 6E and Figure S6C ,D), as well as to display GO terms mostly enriched for "neutrophil activation" and "neutrophil degranulation" (Figure S6E ), in accordance with their elevated pseudotime values ( Figure 6C) , and their prevalent correspondence with NM3s and NM4s ( Figure 6A) . Notably, c4 cells were also found to specifically express high mRNA levels of BEX1 ( Figure 6E and Figure S6C ), while unexpectedly, but in in line with "defense response to virus" as the GO term most enriched for them ( Figure S6E ), c2 cells were found to express elevated levels of interferon-stimulated genes (ISGs), such as ISG15, IFI6, IFIT3 and many more ( Figure 6E and Figure S6C ). No SG or GG genes were present in scRNA-seq datasets ( Figure   S6D ), confirming RNA-seq data ( Figure 4D ). Finally, by assessing their potential developmental trajectories [by using Destiny algorithm 43 ], we found that c1-c4 cells distribute along three branches, one of them including the majority of c1 cells ( Figure 6F ,G and Figure S6F ) other than a minor quote of the c2, c3 and c4 cells. The c1 branch was then found to continue into two different trajectories ( Figure 6F ): one, defined as "conventional trajectory", since mainly characterized by cells expressing elevated AG mRNA levels (i.e., c3 and c4 cells), but also including the BEX1 mRNApositive cells (c4 cells); the other one, unexpected, and defined as "ISG trajectory" ( Figure 6F ) since characterized by cells mainly expressing high ISG mRNA levels (i.e., c2 cells) ( Figure 6H ). In sum, scRNA-seq clustering analysis of NMs identified four clusters of neutrophil progenitors at different stages of maturation and distributed along two maturation routes. Data also uncovered that, even though phenotypically differing among themselves exclusively at the CD45RA level, NM1s and NM2 on the one hand, and NM3s and NM4s on the other hand, are identical in terms of cluster distribution.

In this study, by using the CD45/CD38/CD34/CD10/CD123/CD45RA/CD64/CD115 flow cytometry antibody panel to examine the SSC low CD45 dim region of human BM, we initially identified CD34 + CD45RA + CD64 dim CD115cell progenitors, within cGMPs, exclusively committed to the neutrophil lineage. These neutrophil progenitors may therefore be the long searched-for "missing piece in the puzzle" 44 that composes the cGMP region together with the uni-potent cMoPs and CDPs, and the heterogeneous MDPs and GMDPs. By the same approach, not only we confirmed cMoPs as CD64 + CD115 + 17, 21 , but also pointed both MDPs and GMDPs as CD64 -16 . Since our data are in line with results from other groups 4, 18-20 , we propose to rename GMDPs as NMDPs (i.e., neutrophilmonocyte-dendritic cell progenitors).

Subsequent flow cytometry and in vitro differentiation experiments uncovered, other than those initially found within cGMPs, three more CD34 + and CD34 dim/-CD64 dim CD115neutrophil progenitors, which we ultimately renamed as neutrophil myeloblasts (NMs), and based on their differential CD45RA expression, subdivided into NM1s for CD34 + CD45RA -NMs, NM2s for CD34 + CD45RA + NMs, NM3s for CD34 dim/-CD45RA + NMs and NM4s for CD34 dim/-CD45RA -NMs.

It is correct to point out that, since the CD34 and CD45RA fluorescence distributions in SSC low CD45 dim cells change gradually during their maturation, the separation of CD34 + and CD34 dim/-NMs into four phenotypical populations has been done to intentionally enrich for cells at similar maturation stages. However, since hematopoietic progenitors flow in continuum during their differentiation process, it is implicit that also NMs mature via gradual transitions at both phenotypical and transcriptional levels. The biological significance of the presence of both CD45RA + and CD45RA -NMs remains to be clarified. One hypothesis could be that CD45RA serves to localize CD34 + and CD34 dim/-NMs in specific BM niches. Another one could be that CD45RA has a

functional, yet unknown, role under discrete pathological conditions, as suggested by the selective expansion of CD45RA + immature cells in the majority of acute myeloid leukemia patients 45 .

That CD34 + and CD34 dim/-NMs represent very early progenitors has been unequivocally confirmed by RNA-seq experiments, which indeed proved that they do not express SG genes.

Moreover, clustering analysis of DEGs among CD34 + and CD34 dim/-NMs evidenced a remarkable enrichment of genes related to cell proliferation in NM1s/NM2s, and of AG-encoding genes in NM3s/NM4s. These findings exclude that our CD34 + and CD34 dim/-NMs correspond to the previously described human preNeus 22 , which instead express high levels of SG genes. NM2s, instead, might eventually correspond to the recently described murine proNeu1s, defined as early committed neutrophil progenitors preceding preNeus found within GMPs and not expressing SG genes 9, 24 .

However, the human counterpart of proNeus, identified by "Infinity Flow" in cord blood and fetal BM 24 , was found to be CD66b + /CD15 hi , unlike our CD34 + and CD34 dim/-NMs. Moreover, our NMs, other than being CD66b -, are included within the SSC low CD45 dim progenitor region of BM, which proves that they represent neutrophil progenitors more immature than the recently described CD66b + hNePs 23 , eNePs 25 Clusters of neutrophil progenitors, displaying levels of maturation apparently similar to our c1-c4 cells, have been also identified by scRNA-seq studies of HSPCs from human BM, even though no sorting strategy for their isolation was described. For instance, a cell cluster of myeloid progenitors expressing ELANE and PRTN3 has been found within CMPs 20 , but, as such, it might be also confused with cells belonging to the monocyte lineage. In another of these studies 3 , four clusters of neutrophil progenitors (i.e., N0-N3) were found in CMPs (i.e., N0) and cGMPs (i.e., N1, N2 and N3). N1, N2

and N3 might in part correspond to our c1 cells, given their CD34 protein, as well as AG mRNA, expression. However, N3 could be rather ascribed to a monocyte progenitor for its expression of LYZ, SAMHD1, CSF1R, ANXA2, KLF4 and IRF8 genes 3 , which we found to be peculiar to cMoPs (our unpublished data). Similarly, since shown to include only MPO among AG genes 3 , which can be also shared by cMoPs 9 , it remains uncertain whether N0 effectively corresponds to a neutrophil progenitor. None of these studies 3, 20 detected specific ISG-or BEX1-containing clusters, likely for the much lower number of neutrophil progenitors investigated than in our scRNA-seq set.

In summary, we identified CD34 + and CD34 dim/neutrophil-committed progenitors within human BMs, all of them characterized by a SSC low Lin -CD45 dim CD64 dim CD115phenotype, that in our view correspond to neutrophil myeloblasts. As such, CD34 + and CD34 dim/-NMs are easily sortable and manageable for further studies. We expect that future work could uncover specific membrane markers allowing the sorting of the cell clusters composing CD34 + and CD34 dim/-NMs, to perform a better characterization and also to evaluate if and how they expand under pathological conditions. 

The authors declare no competing interests.

Upon local ethical committee approval, BM samples were obtained from healthy donors (HDs) (n=29 , Table S4) Table   S6 . Anti-CD14 and anti-IL5R mAbs were always included in the sorting panel to exclude, respectively, mature monocytes and eosinophil-committed progenitors. Conversely, we omitted the use of antiCD38 and antiCD117 antibodies since NMs were found as CD38 + and CD117 + . Cells were then washed and resuspended at 30*10 6 /ml in staining buffer, to be ultimately filtered through a 0.35 µM nylon mesh. Cells were finally sorted by using a FACSAria Fusion (BD) cell sorter equipped with 85-μm nozzle, immediately centrifuged, resuspended in MEM medium, counted and used for experiments. Alternatively, sorted cells were lysed in RLT buffer (Qiagen) for RNA extraction 57 .

Sorted cell populations displayed a > 95 % purity, as verified by flow cytometry analysis.

To evaluate the differentiation potential of sorted BM progenitors, the latter cells were 

Either progenitor-derived neutrophils harvested from NMs cultured for 7 days, or freshly isolated HD neutrophils, were washed and resuspended at 0.25*10 6 /ml in HBSS supplemented with 10 % FBS, containing 1 mM CaCl2 and 5 mM glucose. O2production in response to 20 ng/ml PMA (Sigma-Aldrich) was assessed by the Cytochrome C reduction assay, as previously described 60 NextSeq 500 System (Illumina). This procedure was utilized for three different BMs, whose data were integrated as outlined below.

Computational analysis of transcriptome datasets generated by Smart-seq2 has been performed by using the bioinformatic pipeline, as previously described 64 . Briefly, after quality filtering, according to the Illumina pipeline, removal of contaminant adapters and base quality trimming were performed using Trim Galore! (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) script. Gene counts were normalized among various samples using DESeq2 40 , and only genes coding for protein and long non-coding RNA (lncRNA) were retained for downstream analysis. Only genes expressed above 1 fragment per kilobase of transcript per million mapped reads (FPKM), in at least one sample, were considered as "expressed" genes and thus included for downstream analysis. Differentially expressed genes (DEGs) were identified using DESeq2, by using as selection parameter adjusted Pvalue lower than 0.01 and likelihood ratio test (LRT) 40 . Batch effects were removed using the limma package's "removeBatchEffect" function before performing principal component analysis (PCA).

PCA was performed on DEGs by using Bioconductor/R package pcaExplorer v.2.10.0.

Hierarchical clustering was performed using the Euclidean distance and Ward aggregation as criteria.

To find a suitable linear order within the hierarchical clustering dendrogram (HSCs, NMs, PMs, MYs, MMs, BCs, SCs and mature blood neutrophils), we used the optimal leaf ordering (OLO) seriation method of R package seriation, version 1.2-9. 65 . The seriation algorithm is based on the function 'seriate' which tries to find a linear order for objects using data in form of a dissimilarity matrix.

Seven Bridges processing for scRNA-seq data Gene ontology (GO) of differentially expressed genes from bulk RNA-seq and scRNA-seq.

GO analysis was performed on DEGs from bulk RNA-seq and scRNA-seq for the cellular component, biological process and molecular function ontology domains by using the Bioconductor/R package clusterProfiler (version 3.14.3) 67 . Over-representation analysis was performed using the Benjamini-Hochberg (BH) procedure, with P value Cutoff = 0.01 and Q-value Cutoff = 0.05. Redundant GO terms were removed by using the "simplify" function in the clusterProfiler package, using the Wang similarity measure and a similarity cutoff of 0.5. The most significant terms were then plotted in R.

Trajectory analysis was performed by using the destiny algorithm v3.01 43 (C) Bar graphs depicting the fold expansion of purified NM populations after a 7 day-culture with SFGc (mean±SEM of live, CD45 + output cells, n=6; *p < 0.05; **p < 0.01; ***p < 0.001).

(D) Representative plots displaying the phenotype and morphology of CD66b + neutrophils derived from NMs cultured with SFGc for 7 days. Bar graphs report the percentages relative to live, CD45 + cells (mean±SEM, n=6, *p < 0.05) of the various neutrophil lineage progenitors generated by NMs. Figure   5B . Color scheme is based on z-score distribution from -4 (purple) to 4 (yellow).

(F,G) Trajectory plots showing the distribution of c1-c4 cells (F), and their pseudotime values (G).

In F, the arrows indicate both the "conventional", and the "ISG", developmental trajectories, along the neutrophil lineage.

(H) mRNA expression patterns of genes characteristic of c1 (i.e., TOP2A), c2 (i.e., IFI6), c3 (i.e., CTSG) and c4 (i.e., BEX1) cells, projected on the trajectory plots from (F). Figure S1 . Gating strategy to identify lineage-positive cells and immature SSC low CD45 dim myeloid/lymphoid progenitors within BM-LDCs. (D-F) Bar graphs depict the percentage (mean ± SEM, n=7-9, **p < 0.01; ***p < 0.001) of the various mature cell types (listed in the x axis) generated after a 7 day-culture with SFG, relative to live, CD45 + cells. (A) BM-LDCs were stained using a 14-color antibody panel specified in Table S6 . Steps shown in panels I-II were sequentially used to exclude doublets (I) and gate on CD45 + leukocytes (II). In steps III-V, analysis was performed on SSC low CD66bcells (III), CD3/CD19/CD1c/CD141(Lin)-negative cells (IV) and CD16/CD56-negative cells (V), to exclude lymphoid cells and mature monocytes/DCs.

Within the latter cells, the CD45 dim region (red gate in panel VI) defines the SSC low Lin -CD45 dim progenitor pool, from which we could subsequently exclude CD45RA + CD10 + lymphocyte progenitors (VII). Of note, the SSC low Lin -CD45 dim region substantially resembles the equivalent region shown in Figure 1 , Figure 2 , Figure S1 and Figure S4 , thus confirming that both 8-color and 14-color gating strategies are comparable. SSC low Lin -CD45 dim CD10progenitor cells (VII) were then displayed by the CD34/CD45RA marker combination, to separate them in the four region (VIII) equivalent to those shown in Figure 3A . Figure 6E .

(D) Expression patterns of cell-cycle, AG, SG, and GG genes projected on the UMAP plot restricted to neutrophil progenitor clusters (c1-c4) (E) Gene Ontology analysis of DEGs for c2-c4. For every cluster (x-axis), the top ten Gene Ontology terms with Benjamini-Hochberg-corrected P values <0.05 (one-sided Fisher's exact test) are shown.

For cluster 1 no enrichment of biological processes GO term was identified (F) Trajectory plots of c1-c4 cells as defined in Figure 5B . In each plot is depicted the density of cells according to the color bar placed at the right bottom corner of the panel

Distinct routes of lineage development reshape the human blood hierarchy across ontogeny

Transcriptional Heterogeneity and Lineage Commitment in Myeloid Progenitors

Human haematopoietic stem cell lineage commitment is a continuous process

A comprehensive single cell transcriptional landscape of human hematopoietic progenitors

Granulocyte-Monocyte Progenitors and Monocyte-Dendritic Cell Progenitors Independently Produce Functionally Distinct Monocytes

Molecular transitions in early progenitors during human cord blood hematopoiesis

Single-cell analysis reveals the continuum of human lympho-myeloid progenitor cells

Lineage tracing on transcriptional landscapes links state to fate during differentiation

Mouse models of neutropenia reveal progenitorstage-specific defects

Hematopoiesis: a human perspective

New insights into hematopoietic differentiation landscapes from single-cell RNA sequencing

From haematopoietic stem cells to complex differentiation landscapes

Haematopoiesis in the era of advanced single-cell technologies

Prospective isolation of human clonogenic common myeloid progenitors

Revised map of the human progenitor hierarchy shows the origin of macrophages and dendritic cells in early lymphoid development

Restricted dendritic cell and monocyte progenitors in human cord blood and bone marrow

Identification of a Human Clonogenic Progenitor with Strict Monocyte Differentiation Potential: A Counterpart of Mouse cMoPs

Identification of the human eosinophil lineage-committed progenitor: revision of phenotypic definition of the human common myeloid progenitor

Revision of the human hematopoietic tree: granulocyte subtypes derive from distinct hematopoietic lineages

Identification of two distinct pathways of human myelopoiesis

Granulocytic and monocytic differentiation of CD34hi cells is associated with distinct changes in the expression of the PU.1-regulated molecules, CD64 and macrophage colony-stimulating factor receptor

Developmental Analysis of Bone Marrow Neutrophils Reveals Populations Specialized in Expansion, Trafficking, and Effector Functions

Identification of an Early Unipotent Neutrophil Progenitor with Pro-tumoral Activity in Mouse and Human Bone Marrow

Combinatorial Single-Cell Analyses of Granulocyte-Monocyte Progenitor Heterogeneity Reveals an Early Uni-potent Neutrophil Progenitor

Coexpression of CD71 and CD117 Identifies an Early Unipotent Neutrophil Progenitor Population in Human Bone Marrow

Immunophenotypic differentiation patterns of normal hematopoiesis in human bone marrow: reference patterns for age-related changes and disease-induced shifts

Immunophenotypic pattern of myeloid populations by flow cytometry analysis

Dendritic cell ontogeny: a human dendritic cell lineage of myeloid origin

Defining human dendritic cell progenitors by multiparametric flow cytometry

Lineage specification of human dendritic cells is marked by IRF8 expression in hematopoietic stem cells and multipotent progenitors

Technical advance: immunophenotypical characterization of human neutrophil differentiation

Granulocyte-macrophage progenitors as candidate leukemic stem cells in blast-crisis CML

Heterogeneous leukemia stem cells in myeloid blast phase chronic myeloid leukemia

Molecular signature of CD34(+) hematopoietic stem and progenitor cells of patients with CML in chronic phase

Fast optimal leaf ordering for hierarchical clustering

Expression of cell surface markers during differentiation of CD34+, CD38-/lo fetal and adult bone marrow cells

Granulopoiesis and granules of human neutrophils

Biological Roles of Neutrophil-Derived Granule Proteins and Cytokines

Moderated estimation of fold change and dispersion for RNAseq data with DESeq2

Dimensionality reduction for visualizing single-cell data using UMAP

Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression

destiny: diffusion maps for large-scale single-cell data in R

Heterogeneity of neutrophils

Coexistence of LMPP-like and GMP-like leukemia stem cells in acute myeloid leukemia

Severe COVID-19 Is Marked by a Dysregulated Myeloid Cell Compartment

BEX1 acts as a tumor suppressor in acute myeloid leukemia

BEX1 promotes imatinib-induced apoptosis by binding to and antagonizing BCL-2

Single-cell transcriptome profiling reveals neutrophil heterogeneity in homeostasis and infection

Sex differences in neutrophil biology modulate response to type I interferons and immunometabolism

Single-Cell Transcriptomics of Human and Mouse Lung Cancers Reveals Conserved Myeloid Populations across Individuals and Species

Transcriptomic, epigenetic, and functional analyses implicate neutrophil diversity in the pathogenesis of systemic lupus erythematosus

Single-cell transcriptomics uncovers distinct molecular signatures of stem cells in chronic myeloid leukemia

The importance of being "pure" neutrophils

European LeukemiaNet 2020 recommendations for treating chronic myeloid leukemia

The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia

IFNalpha enhances the production of IL-6 by human neutrophils activated via TLR8

Optimal ex vivo expansion of neutrophils from PBSC CD34+ cells by a combination of SCF, Flt3-L and G-CSF and its inhibition by further addition of TPO

Large-scale ex vivo generation of human neutrophils from cord blood CD34+ cells

Effect of substance P on superoxide anion and IL-8 production by human PMNL

The MyD88-independent pathway is not mobilized in human neutrophils stimulated via TLR4

Human Neutrophils Produce CCL23 in Response to Various TLR-Agonists and TNFalpha

Full-length RNA-seq from single cells using Smart-seq2

Deciphering the fate of slan(+) -monocytes in human tonsils by gene expression profiling

Getting things in order: An introduction to the R package seriation

Integrating single-cell transcriptomic data across different conditions, technologies, and species

clusterProfiler: an R package for comparing biological themes among gene clusters