A population of innate myelolymphoblastoid effector cell expanded by inactivation of mTOR complex 1 in mice

Adaptive autoimmunity is restrained by controlling population sizes and pathogenicity of harmful clones, while innate destruction is controlled at effector phase. We report here that deletion of Rptor in mouse hematopoietic stem/progenitor cells causes self-destructive innate immunity by massively increasing the population of previously uncharacterized innate myelolymphoblastoid effector cells (IMLECs). Mouse IMLECs are CD3-B220-NK1.1-Ter119- CD11clow/-CD115-F4/80low/-Gr-1- CD11b+, but surprisingly express high levels of PD-L1. Although they morphologically resemble lymphocytes and actively produce transcripts from Immunoglobulin loci, IMLECs have non-rearranged Ig loci, are phenotypically distinguishable from all known lymphocytes, and have a gene signature that bridges lymphoid and myeloid leukocytes. Rptor deletion unleashes differentiation of IMLECs from common myeloid progenitor cells by reducing expression of Myb. Importantly, IMLECs broadly overexpress pattern-recognition receptors and their expansion causes systemic inflammation in response to Toll-like receptor ligands in mice. Our data unveil a novel leukocyte population and an unrecognized role of Raptor/mTORC1 in innate immune tolerance.


Introduction
Autoreactive T and B lymphocytes are controlled by regulation of population sizes and pathogenicity through clonal deletion (Burnet, 1957), clonal anergy (Nossal and Pike, 1980) and regulatory T cells (Sakaguchi et al., 1995). Through broadly-reactive pattern recognition receptors (PRRs) (Janeway, 1989;Medzhitov et al., 1997), innate immunity protects host against infections by both direct effector function and, indirectly, by induction of adaptive immunity (Liu and Janeway, 1991;Liu and Janeway, 1992;Wu and Liu, 1994;Pasare and Medzhitov, 2005). Since innate immune responses triggered by host components can also cause fatal tissue damage (Stetson et al., 2008;Chen et al., 2009), it must be properly regulated to protect hosts against self-destruction. Although a number of mechanisms have been proposed to prevent self-destructive innate effector functions Liu et al., 2011;Ljunggren and Kärre, 1990;Liew et al., 2005;Takeuchi and Akira, 2010), it is less clear if population sizes of innate effectors are suppressed to limit selfdestruction.
The significance of protective self-tolerance mechanisms in adaptive immunity are revealed only when they have gone awry. For example, the significance of clonal deletion was elucidated when it was prevented by blocking either costimulation or antigen-expression in the thymus (Anderson et al., 2002;Gao et al., 2002), while mice with Foxp3 mutation informed us of the consequence of defective regulatory T cells (Hori et al., 2003;Fontenot et al., 2003). Since removal of the forbidden autoreactive T and B cells is achieved during T and B cell development (Nemazee and Bürki, 1989;Sha et al., 1988;Kisielow et al., 1988), it is intriguing whether certain parallel mechanisms in innate immunity might also remain to be uncovered through genetic inactivation of key regulators in development of innate effector cells.
The hematopoietic system is among the best characterized of all tissue/systems in mammalians, with cell types and lineages clearly defined in the context of developmental stages and localization (Metcalf, 1999;Weissman, 1994). Hematopoiesis in bone marrow (BM) is responsible for generation of major lineages of innate effectors, including NK cells, granulocytes, monocytes and dendritic cells. While genetic switch in generation of innate immune system has been identified (Weissman, 1994), we are not aware of defects that predispose host to innate immune attack through increasing population sizes of self-destructive innate effectors.
The mammalian target of rapamycin (mTOR) pathway, which couples energy and nutrient abundance to the execution of cell growth and division, has emerged as a major regulator of hematopoiesis. Thus, activation of mTOR complex 1 (mTORC1) by deleting Tsc1, which encodes a negative regulator for mTORC1 (Inoki et al., 2002), causes loss of hematopoietic stem cell (HSC) function and renders mice prone to leukemiogenesis in conjunction with loss of tumor suppressor Pten (Yilmaz et al., 2006;Zhang et al., 2006;Chen et al., 2008). More recently, two groups reported that deletion of Rptor, which encodes a critical component of mTORC1 (Hara et al., 2002; eLife digest The cells of the immune system defend us from bacteria, viruses and other microbes that might cause harm to the body. Immune cells develop from stem cells in the bone marrow. The stem cells first develop into one of two types of progenitor cell before specializing further into the different types of mature immune cell. Researchers often categorize immune cells as either myeloid or lymphoid, depending on which progenitor cell they developed from.
A protein complex called mTORC1 in the stem cells helps to guide immune cell development. One of the proteins in the mTORC1 complex is called Raptor. In mice that lack the Raptor protein, cells with particular markers on their surface accumulate in the bone marrow. The exact identity of these cells and why they appear was not known. Tang et al. have analyzed these cells in mice that lacked Raptor in their bone marrow stem cells. This revealed that the cells have features of both myeloid and lymphoid cells: although they develop from myeloid progenitor cells, they are shaped like lymphoid cells. The cells also have a surface marker normally found on myeloid cells. Tang et al. have

named the cells innate myelolymphoblastoid effector cells (IMLECs).
Further investigation showed that the lack of the Raptor protein caused another gene in the stem cells, called Myb, to become less active than normal. Tang et al. suggest that this lack of activity causes more IMLECs to develop from the stem cells. The overproduction of IMLECs also causes inflammation in the mice.
IMLECs bridge the gap between myeloid and lymphoid cells, challenging the current categorization of these as separate cell types. Targeting this new cell population could help researchers to develop new methods to control the immune response, for example, during autoimmune disorders where the immune system is overactive and damages the body's own cells. Kim et al., 2002), dramatically perturbed hematopoiesis in mice (Hoshii et al., 2012;Kalaitzidis et al., 2012), as evidenced by defects in production of mature lymphoid and myeloid cells. Remarkably, cells with CD11b + Gr-1surface markers massively accumulated in BM following Rptor deletion in HSCs (Hoshii et al., 2012;Kalaitzidis et al., 2012). The nature of this population and consequences of their accumulation, however, remains a mystery.
Here we systematically analyzed the gene expression signature, cell surface markers, morphology and functions of the CD11b + Gr-1population in the Rptor-deficient BM and other organs and sought for their physiological counterpart in the normal mice. We found that these cells can be identified in both normal and Rptor-deficient hosts by CD3 -B220 -NK1.1 -Ter119 -CD11c low/-CD115 -F4/ 80 low/-Gr-1 -CD11b + PD-L1 + markers, lymphoid morphology and actively transcribed Ig loci. Interestingly, these cells broadly express essentially all TLRs along with many other pattern recognition receptors and mount a greatly exacerbated response to all TLR ligands tested. We name this population IMLEC for innate myelolymphoblastoid effector cell that can be derived from common myeloid progenitors. Because their expansion and broad distribution render the host vulnerable to TLR ligands, we suggest that mTORC1-mediated repression of IMLEC expansion represents a new mechanism of immune tolerance in the innate immunity. Our study also raises an intriguing perspective that while repressing mTOR over-activation suppresses leukemia, a functional mTORC1 must be maintained to limit generation of IMLECs to avoid innate immune destruction.

Results
Raptor suppresses accumulation of a previously uncharacterized subset of leukocytes with features of both myeloid and lymphoid cells As germline deletion of Rptor (which encode the Raptor protein) is embryonic-lethal, we crossed mice harboring homozygous loxp-flanked Rptor exon 6 (Polak et al., 2008) to those with interferoninducible Mx1-Cre recombinase transgene, which allows inducible deletion of target genes effectively in the hematopoietic system upon treatment of interferon or its inducers (Kühn et al., 1995). We treated the 6-8 weeks old Rptor F/F and Rptor F/F ,Mx1-Cre mice with polyinosinic: polycytidylic acid (pIpC) every other day for 2 weeks to induce the deletion of Rptor. Hereafter, we refer to the pIpC-treated Rptor F/F mice as Ctrl (control) mice, while the Rptor F/F , Mx1-Cre mice as cKO (conditional knockout) mice ( Figure 1A and Figure 1-figure supplement 1). As has been reported by others (Hoshii et al., 2012;Kalaitzidis et al., 2012), Rptor deletion causes broad defects in all lineages of hematopoietic cells (see also Figure 1-figure supplements 1, 2 and 3). However, the number of hematopoietic stem/progenitor cells (HSPCs) increased ( Figure 1-figure supplement 4). Most notably, CD11b + Gr-1cells, which amount to nearly 50% of BM cells in our model, emerge at the expense of CD11b + Gr-1 + granulocytes from the cKO mice ( Figure 1B,C). Importantly, we also observed the massive accumulation of CD11b + Gr-1cells in the BM of Rptor F/F , Cre-ER mice after tamoxifen induced targeted mutation of Rptor, which clearly excludes the role of pIpC in the generation of these cells (Figure 1-figure supplement 5).
The CD11b + Gr-1cells were smaller and had reduced granularity when compared to CD11b + Gr-1 + granulocytes, but were larger and more granular than the CD11b -Gr-1cells (Figure 1-figure supplement 2E,F). Surprisingly, despite the expression of myeloid marker CD11b on the expanded population of BM cells, histological analysis of BM section revealed pervasive expansion of lymphoblastoid cells ( Figure 1D). The cKO BM contained markedly decreased erythroid and myeloid lineage cells and markedly increased lymphocytes. Lymphocytes were predominantly small-to-medium sized and had normal cytological features. There was also an increased population of large blast-like cells with prominent nucleoli and perinuclear clearing resembling lymphoblasts. Plasma cells were present in small numbers. The myeloid: erythroid ratio was within normal range (3.04) but the overall number of erythroid and myeloid cells was very low. In particular, very few erythroid cells were present. In the myeloid lineage there was also maturation disruption since immature ring form neutrophils (neutrophilic metamyelocytes) predominated over mature neutrophils (condensed chromatin) ( Figure 1D). Giemsa staining of BM smear revealed a massively increased lymphoblast population and severe depletion of both immature erythroid cells and granulocytes in the cKO mice ( Figure 1E). These cells were replaced by cells with prominent nucleoli and perinuclear clearing resembling lymphoblasts. To confirm that the lymphoblasts were the CD11b + Gr-1cells identified Figure 1. Targeted mutation of Rptor in hematopoiesis led to massive accumulation of IMLEC. (A) Schematic of experimental design. Sex-matched 6-8 weeks old Ctrl (Rptor F/F ) and cKO (Rptor F/F , Mx1-Cre) mice were treated with pIpC for seven times. The phenotypes were analyzed on day 7 or day 30 after the complement of pIpC treatment. (B) Representative flow cytometric analysis of myeloid cells from mice BM by CD11b and Gr-1 on day seven after pIpC treatment. Similar data were obtained on day 30. (C) Frequencies of various hematopoietic cell populations in BM based on CD11b and Gr-1 markers. n = 7 for Ctrl mice; n = 9 for cKO mice. Data are pooled from three independent experiments. (D) Expansion of lymphoblast population in cKO BM revealed by histology. Left, H&E staining showing prominence of cells with lymphoid morphology in the cKO BM; Right, normal BM histology from Ctrl mouse showing normal myeloid and erythroid lineages. Scale bar, 20 mm. (E) Expansion of lymphoblast revealed by cytology. Left, cytological preparation of BM smear from cKO mouse showing preponderance of lymphocytes (short arrows) and severe depletion of myeloid and erythroid lineages. There were some promyelocytes and myelocytes present (larger cells) but no mature neutrophils. Right, normal cytological preparation of BM from Ctrl mouse. There were numerous erythroid precursors with intensely basophilic, condensed chromatin (arrowheads) and a few lymphocytes (short arrow) with less condensed chromatin. Additionally, in Ctrl BM there were numerous mature neutrophils (ringed nucleus with constrictions) which were severely depleted in the cKO mouse. Scale bar, 10 mm. (F) Giemsa staining of FACS-sorted Rptor-deficient CD11b + Gr-1 -BM cells. Scale bar, 10 mm. Similar morphology was observed in three independent experiments. (G) Principal component analysis (PCA) of gene expression in CD11b + Gr-1 -BM cells (IMLECs) and other hematopoietic cells. Numbers along axes indicate relative scaling of the principal variables. RNA-seq data from IMLECs obtained in our study were compared with those deposited in public database by others. Datasets are from known lymphoid (Pro-B, Pre-B, Naïve B, CD4 T, CD8 T, Treg and NK) or myeloid (macrophage, dendritic cell, granulocyte, neutrophil, eosinophil, basophil and erythrocyte) subsets. DOI: https://doi.org/10.7554/eLife.32497.003 The following figure supplements are available for figure 1:   Figure 1B,C), we FACS-sorted the subset based on CD11b + Gr-1surface markers and validated its lymphoid morphology ( Figure 1F).
The spleen was greatly enlarged due to expansion of the follicular centers and periarteriolar sheaths within the splenic white pulp (lymphoid areas) of the cKO mice ( Figure 1-figure supplement 1C). The expanding white pulp populations consisted of lightly stained, large cells that morphologically resembled germinal center lymphocytes. In some areas, these populations expanded within the marginal zones while in others, they involved the periarteriolar sheaths. The cells had an increased amount of pale eosinophilic cytoplasm and mild pleomorphism with both centroblast-like cells (larger cells with large ovoid nuclei and 1-2 prominent nucleoli per cell) and centrocyte-like cells (smaller cells with cleaved or elongated nuclei and unapparent nucleoli).
Since cells with such combination of morphology and surface markers had not been identified previously, we FACS-sorted the CD11b + Gr-1cells from the Rptor cKO BM and carried out nextgeneration RNA sequencing (RNA-seq). Using principal component analysis (PCA), we compared gene expression profiles of these CD11b + Gr-1 -cKO BM cells with other known subsets of hematopoietic cells, including B cells, T cell subsets, NK cells, myeloid cell subsets, dendritic cells and erythroid cells. This analysis demonstrates that the CD11b + Gr-1 -cKO BM cells were distinct from all known blood cell types, although they appear to be closely related to B lymphocytes and macrophages ( Figure 1G). We hereafter refer these cells as innate myeloidlymphoblastoid effector cells (IMLECs).
We identified subset specific genes using a threshold of 4-fold changes and an adjusted FDRadjusted p-value<0.01. As shown in Figure 2A, by comparing RNA-seq data-based gene expression signature of all known leukocyte subsets, a unique gene expression signature was identified in the CD11b + Gr-1 -cKO BM IMLECs. The signature consists of 48 genes that are up-regulated by more than 4-fold. The 48 genes over expressed in the CD11b + Gr-1 -cKO BM IMLECs are listed in Figure 2B. Among their diverse functions, these genes are involved in intracellular signaling cascades (such as Arhgap31, Rab20, Gna12, Mink1 and Prkch) and metabolic processes (such as Naga, Atf3, Aoah, Chst14 and Gns). The uniqueness of cKO BM IMLEC is also supported by pair-wise comparisons between IMLEC and peritoneal macrophage or other closely related cell types that are prominent in BM (Figure 2-figure supplement 1).
A defining feature of the B cell lineage is activation of Ig gene loci, as evidenced by 'sterile' transcripts transcribed from the unarranged loci of Ig heavy chain (Igh) and light chains (Igk and Igl) (Sleckman et al., 1996). RNA-seq data revealed high levels of sterile transcripts within the Igh locus ( Figure 2C) and Igk and Igl loci (Figure 2-figure supplement 2) from both B cell lineages and cKO BM IMLECs. This is a significant difference from macrophage, which had no detectable expression of the sterile transcripts, as expected. Another defining feature of developing B-lymphocytes is Ig gene rearrangement, a unique mechanism of genetic recombination that occurs only during the early stages of B cell maturation. This process is strictly dependent on recombinases genes Rag1 (Mombaerts et al., 1992;Schatz et al., 1989) and Rag2 (Shinkai et al., 1992;Oettinger et al., 1990). As shown in Figure 2D, no expression of Rag1 was detectable by quantitative PCR (qPCR), although a detectable but extremely low level of Rag2 was observed. Consistent with lack of Rag1 expression, no gene rearrangement was found in the Ig loci ( Figure 2E). Taken together, our data so far demonstrate that Raptor suppresses accumulation of a previously uncharacterized leukocyte subset with features of both myeloid and lymphoid cells.

Characterization of IMLECs in Rptor cKO and WT mice
IMLECs identified in the Rptor cKO mice did not express surface markers that are used to define other lymphocytes, such as B220 for B cells, CD3 for T cells and NK1.1 for NK cells ( Figure 3A). The high levels of CD11b indicated that these cells are distinct from the recently identified innate lymphoid cells (ILCs) that are CD11b negative (Walker et al., 2013). In addition, they lack surface markers for progenitor cells, such as c-Kit and Sca-1 ( Figure 3A). Although the cKO IMLECs retained a high level of myeloid marker CD11b, they expressed a very low level of F4/80 macrophage marker and lacked CD115 monocyte marker ( Figure 3B). To identify a positive marker for IMLEC, we searched our RNA-seq database for overexpression of genes that encode cell surface CD (cluster of differentiation) markers. Among 317 CD markers (Supplementary file 1), the most up-regulated gene in the cKO BM IMLECs over Ctrl whole BM cells was Cd274 (also called B7h1 or Pdl1). As shown in Figure 3B and Figure 3C, PD-L1 was expressed on the vast majority of CD11b + Gr-1 -BM cells from Rptor cKO mice.
It should be noted that although cKO IMLECs also over-express Cd11c gene (Supplementary file 1), IMLEC gene expression profiles are distinct from dendritic cell (DC) based on gene signature ( Figure 1G and Figure 2-figure supplement 1E,F). In cKO IMLECs, the CD11c levels were somewhat lower than the PD-L1 -DC ( Figure 3D). In WT mice, greater than 90% of Lin -CD11b + PD-L1 + IMLEC in BM, lung and peripheral blood mononuclear cells expressed only low levels of CD11c, while those in spleen and mesenteric lymph nodes consisted of two major subsets: CD11c high and CD11c low/-( Figure 3E). IMLECs were also found among the leukocytes isolated from lung and in peripheral lymphoid organ ( Figure 3F, G and H), and this population was greatly expanded in the cKO mice ( Figure 3I).
To further confirm that this subset is the IMLEC in normal BM, we FACS-sorted the Lin -(CD3 -B220 -NK1.1 -Ter119 -Gr-1 -F4/80 -CD115 -) CD11b + PD-L1 + cells from wild type (WT) BM and characterized their morphology and levels of sterile Ig transcripts. As shown in Figure 4A, the sorted cells had a lymphoid morphology as did the cKO IMLECs. They also displayed comparable size and granularity as cKO IMLECs ( Figure 4B). Moreover, Lin -CD11b + PD-L1 + cells from WT BM expressed sterile transcripts of Ig loci identified by RNA-seq ( Figure 4C). Furthermore, subsequent validation of IMLECs in WT BM was undertaken by comparing the expression of other top candidate markers (CD14, CD16) and MHC-I/MHC-II (Figure 4-figure supplement 1A), as well as population-specific transcription factors Mitf, Atf3 and Zdhhc1 ( Figure 4D,E). The largely comparable expression levels of these surface markers and transcription factors between WT and cKO IMLECs provide additional lines of evidence for these cells to be naturally occurring IMLECs. Therefore, a small fraction of normal leukocytes in lymphoid and non-lymphoid tissues have the IMLEC phenotype, and this subset is massively expanded after Rptor deletion.

Altered differentiation is responsible for accumulation of IMLECs
Theoretically, expansion of IMLECs in cKO BM may be caused by increased proliferation and/or reduced apoptosis. To test this possibility, we analyzed the proliferation of IMLECs by Ki-67 staining and BrdU incorporation. Remarkably, Lin -CD11b + PD-L1 + IMLECs from both Raptor Ctrl and cKO BM had much fewer Ki-67 + cells ( Figure   other BM cell types effectively rules out rapid proliferation as an explanation for IMLECs accumulation in Raptor cKO BM. Likewise, the massive increase of IMLEC in cKO mice over those in the Ctrl mice cannot be due to proliferation, as the percentage of Ki-67 + or BrdU + cells is not increased in cKO mice. Furthermore, based on cell surface Annexin V staining, IMLECs from cKO BM were more   proliferation and increased apoptosis of IMLECs, our exhaustive efforts to demonstrate self-renewal of IMLEC through transplantation of massive numbers of IMLEC have all been unsuccessful (data not shown).
As an alternative hypothesis, we evaluated whether IMLECs accumulated because of altered differentiation of hematopoietic stem and progenitors (HSPCs). As the first step to test this hypothesis, we evaluated if IMLEC accumulation in cKO BM was cell-intrinsic. Briefly, we mixed either Rptor F/F or Rptor F/F , Mx1-Cre (both CD45.2 + ) BM cells with recipient type CD45.1 + WT BM cells at a 2:1 ratio. At six weeks after BM transplantation, Rptor was deleted from the Rptor F/F , Mx1-Cre donor-derived cells by pIpC treatment ( Figure 5A). As shown in Figure 5B, the accumulation of CD11b + Gr-1 -IMLECs was intrinsic to Rptor -/-BM cells. Since our earlier data suggested that IMLECs accumulated at the expense of granulocytes ( Figure 1B,C), we tested if granulocytes were converted to IMLECs following Rptor deletion. We produced Rptor F/F , Lyz2-Cre +/+ mice that should have myeloid lineagespecific deletion of Rptor. However, despite the effective deletion of the Rptor gene in the granulocytes ( Figure 5C), the percentages of CD11b + Gr-1 + granulocytes and CD11b + Gr-1 -IMLECs were unchanged ( Figure 5D). These data suggest that accumulation of IMLEC in the cKO mice was not due to trans-differentiation from granulocytes.
Next, we use both in vitro co-culture and in vivo BM transplantation to identify the progenitor that may give rise to IMLECs. We co-cultured OP9 stromal cells with FACS-sorted BM LSK (Lin -Sca-1 + c-Kit + ), CMP (Lin -Sca-1c-Kit + CD34 Medium CD16/32 Medium ) and CLP (Lin -CD127 + Sca-1 Medium c-Kit Medium ) populations from Ctrl and cKO mice that had been treated with pIpC ( Figure 5E). As shown in Figure 5F, both LSK and CMP populations from Rptor -/-BM gave rise to CD11b + Gr-1 -PD-L1 + IMLECs. As expected, Rptor-sufficient CLPs were not able to give rise to CD11b + myeloid cells. Interestingly, Rptor-deficient CLPs generated progenies with a small portion exhibiting immunophenotypes of IMLECs. We also transplanted sorted LSK and CMP populations and induced Rptor deletion in the donor cells by treating recipients with pIpC ( Figure 5G) to confirm their ability in giving rise to IMLECs. Due to lack of self-renewal activity of progenitor cells and rapid apoptosis of IMLEC, we used a much shorter timeline than the whole bone marrow transplantation studies in order to capture progenitor-derived IMLEC. As shown in Figure 5H, deletion of Rptor in either LSKs or CMPs was sufficient to induce the generation of CD11b + Gr-1 -PD-L1 + cells in recipients BM. The shorter timeline explained relative paucity of LSK-derived IMLEC when compared with long-term bone marrow transplantation ( Figure 5A,B). As expected, since CMPs do not have self-renewal capability, only a small number of progeny cells were produced. However, since IMLEC can be generated from CLP in vitro, their potential to do so under physiological conditions cannot be ruled out. Taken together, our data demonstrate that the massive accumulation of IMLECs in cKO mice can be caused by altered differentiation of CMPs, although other differentiation pathway cannot be ruled out.

Reduced c-Myb expression is responsible for accumulation of IMLECs
A previous study demonstrated that heterozygous Myb mutation leads to an expansion of BM CD11b + Gr-1cells (García et al., 2009). Although expression of PD-L1 was not evaluated in the earlier study, we were intrigued by the possibility that down-regulation of Myb may be the underlying mechanism for the massive production of IMLECs. Since both LSKs and CMPs are able to give rise to CD11b + Gr-1 -PD-L1 + IMLECs, we evaluated expression of c-Myb in both LSK and CMP populations. Indeed, the Myb transcripts were significantly reduced in both LSK and CMP populations sorted from Rptor cKO mice ( Figure 6A). Moreover, our intracellular staining also revealed reduced levels of c-Myb protein in both LSKs and CMPs from cKO BM ( Figure 6B). Interestingly, induced deletion of c-Myb in mice with homozygous floxed c-Myb (Myb F/F , Mx1-Cre), but not heterozygous floxed c-Myb (Myb F/+ , Mx1-Cre), showed obvious increase of PD-L1 expression in CD11b + Gr-1 -BM cells ( Figure 6C,D,E and Figure 6-figure supplement 1A,B,C), despite of significant decrease of BM whole leukocytes due to Myb-reduction induced cell apoptosis. To avoid excessive apoptosis and test if the down-regulation of Myb was necessary and sufficient to cause accumulation of IMLECs, we transplanted BM cells from either Myb F/F , Rag2 -/or Myb F/F , Rag2 -/-, CreER mice to CD45.1 + recipients, which then received tamoxifen to achieve deletion of Myb specifically in donor-derived hematopoietic cells after their full reconstitution ( Figure 6F). Consistent with essential role for c-Myb in hematopoiesis, surviving leukocytes appeared heterozygous for Myb deletion ( Figure 6G). As early as 1 week after the first injection of tamoxifen, a significant decrease in CD11b + Gr-1 + Linc-Kit + HSPCs were isolated from Raptor Ctrl and cKO mice at 10 days after pIpC treatment. miRNA levels were measured by miRNA microarray. The y-axis shows the log 2 ratio of signal, while the x-axis shows the -log 10 P value. The doted boxes show the numbers of significantly (p<0.05) up-regulated (fold change >2) or down-regulated (fold change <0.5) miRNAs among 50 miRNAs with mirSVR score <À1.0. Each dot represents the mean value of a unique miRNA from three independent samples. Figure 6 continued on next page granulocytes and an increase in CD11b + Gr-1 -PD-L1 + IMLECs were observed in Myb cKO BM ( Figure 6H). Moreover, the provision of heterologous c-Myb significantly diminished the generation of IMLECs from Raptor-deficient LSK cells in our in vitro OP9 co-culturing experiments (Figure 6figure supplement 1D,E,F,G). Therefore, down-regulation of Myb is necessary for production of IMLECs.
We recently reported that deletion of Rptor caused up-regulation of miRNA biogenesis in HSPCs (Ye et al., 2015). We searched our miRNA microarray database and mirSVR score database (Betel et al., 2010) for potential impact of miRNAs in down-regulation of Myb expression. Using the stringent criteria of mirSVR score <À1.0, we identified 50 miRNAs that presented in the HSPCs (Supplementary file 2). Among them, 13 miRNAs showed >2.0 folds up-regulation in the Rptor-deficient HSPCs (p<0.05), while none showed statistically significant down-regulation ( Figure 6I). Significant up-regulation of miR-150 (1.4 folds increase, p=0.05), which was previously demonstrated to inhibit Myb expression Xiao et al., 2007), was also observed. Up-regulation of Myb-targeting miRNAs provides a plausible mechanism for down-regulation of Myb by Rptor deletion. However, the broad spectrum of the up-regulated miRNAs suggests that it is unlikely that a single miRNA is responsible for the overall reduction of Myb expression.
Expansion of IMLECs associated with lethal inflammatory response to TLR ligands RNA-seq data indicated that IMLECs broadly up-regulate pattern recognition receptors (PRRs) genes. TLRs, the first family of PRRs identified, were broadly over-expressed in cKO IMLECs as determined by RNA-seq ( Figure 7A). RT-PCR confirmed that IMLECs from both cKO and WT mice over-expressed essentially all TLRs tested, particularly Tlr2, Tlr3, Tlr4, Tlr6, Tlr7, Tlr8 and Tlr9 ( Figure 7B). In addition to TLRs, expression of other PRRs, including NLRs, ALRs and RLRs, was also broadly elevated (Figure 7-figure supplement 1). When the BM from Ctrl and Rptor -/mice were compared for their responses to TLRs agonists, including synthetic tripalmitoylated lipopeptide Pam3CysSerLys4 (Pam3CSK4, TLR1/2 agonist), heat-killed Listeria monocytogenes (HKLM, TLR2 agonist), synthetic analog of double-stranded RNA poly I:C (pIpC, TLR3 agonist), lipopolysaccharide (LPS, TLR4 agonist), flagellin from Salmonella typhimurium (FLA-ST, TLR5 agonist), synthetic lipoprotein derived from Mycoplasma salivarium (FSL-1, TLR2/6 agonist), single-stranded RNA Double-Right complexed with LyoVec (ssRNA-DR, TLR7 agonist) and synthetic oligonucleotides containing unmethylated CpG dinucleotides (ODN1826, TLR9 agonist), it is clear that Rptor -/-BM increased production of TNF-a ( Figure 7C) and MCP-1 ( Figure 7D) in response to all TLR ligands tested. To determine the cellular basis for the enhanced cytokine production, we used intracellular and cell surface staining to identify cells that produced the inflammatory cytokines. As shown in Figure 7E, IMLECs from both WT and Rptor cKO BM were potent producers of TNF-a when stimulated by LPS.
Consistent with exacerbated responses to TLR ligands in BM cells, deletion of Rptor in cKO mice resulted in massive increase in inflammatory cytokines in serum ( Figure 8A). Approximately 40% of cKO mice died within 2 months after 7 pIpC treatments ( Figure 8B). Histological analyses revealed extensive inflammation in the liver with associated tissue injuries ( Figure 8C). A substantial proportion of the leukocytes in the liver of cKO mice were IMLECs, as demonstrated by cell surface markers CD11b + Gr-1 -PD-L1 + F4/80 low/-( Figure 8D,E). To test if the mice with expanded IMLECs were more sensitive to endotoxin, we challenged the Ctrl and cKO mice with low doses of LPS (5 mg/kg body weight). While all Ctrl mice survived the LPS challenge, all cKO mice succumbed within 36 hr ( Figure 8F). The dramatically increased mortality due to endotoxic shock was associated with remarkably elevated levels of inflammatory cytokines. As shown in Figure 8G and Figure 8H, a more than 500-fold increase in TNF-a and an approximately 10-fold increase of MCP-1 were detected in the serum of cKO mice at 6 hr after LPS injection. It has been demonstrated that  acetaminophen-triggered liver necrosis induces HMGB-1-mediated inflammatory responses to danger-associated molecular patterns (DAMPs) Scaffidi et al., 2002). As shown in Figure 8I and Figure 8J, cKO mice mounted a significantly elevated inflammation to challenge by experiments. n = 14 for Ctrl mice; n = 11 for cKO mice. Inflammatory cytokines TNF-a (G) and MCP-1 (H) levels from mice serum before (0 h) and at 6 hr after (6 h) LPS injection are shown. n = 10 for Ctrl mice; n = 9 for cKO mice. Note greater than 500-fold increase in plasma TNF-a levels. (I, J) cKO mice mounted enhanced inflammatory response to acetaminophen-triggered liver necrosis. Serum TNF-a (I) and MCP-1(J) levels at 6 hr upon acetaminophen (3.2 mg/mouse) treatment are shown. n = 4 for Ctrl mice; n = 6 for cKO mice. Mann-Whitney test was used for statistics analysis, and lines indicate Mean ±SD. Similar trends were observed in another independent experiment. DOI: https://doi.org/10.7554/eLife.32497.020 low doses of acetaminophen. Therefore, amplification of IMLEC also leads to elevated response to tissue injuries.
Since Mx1-Cre was broadly activated after pIpC treatment, it is less certain whether the increased sensitivity of the cKO mice to TLR ligands is due to immunological abnormality. To address this issue, we produced chimeric mice in which pIpC induces deletion of the targeted gene exclusively in hematopoietic cells by transplanting Rptor F/F , Mx1-Cre (CD45.2 + ) BM cells into lethally irradiated CD45.1 + recipients. After hematopoietic reconstitution, the recipients were treated with 3 doses of pIpC to induce deletion of Rptor exclusively in the hematopoietic cells. After 10 days of pause, the Ctrl and cKO chimera mice were challenged with new pIpC injection and monitored for survival ( Figure 9A). As shown in Figure 9B, while a large portion of the cKO chimera mice died progressively starting within a week of the second round of pIpC treatment, all Ctrl chimera mice survived the observation period of more than 45 days. Massive leukocytes infiltration was observed in the liver of cKO chimeric mice ( Figure 9C). Cell surface phenotyping of the donor-type leukocytes in BM ( Figure 9D), spleen ( Figure 9E) and liver ( Figure 9F) revealed accumulation of CD11b + Gr-1 -PD-L1 + F4/80 low/-IMLECs. Collectively, the data in Figure 8 and Figure 9 demonstrate that over-expansion of IMLECs, as a result of Rptor deletion, renders the host highly vulnerable to TLR ligands.

Discussion
We have characterized a leukocyte population, which we called IMLECs, that has a strong innate effector function but with features of both myeloid and lymphoid cells. Furthermore, our data reveal an unexpected function of mTORC1 in suppressing IMLEC expansion. The high vulnerability to TLR ligands after IMLEC expansion highlights a new consequence of defective hematopoiesis and a new mechanism of immune tolerance.
Two groups have previous reported that inactivation of mTORC1 by deletion of Rptor leads to massive accumulation of CD11b + Gr-1cells in the BM (Hoshii et al., 2012;Kalaitzidis et al., 2012). Similar results were obtained in mice with Mtor deletion in the hematopoietic cells (Guo et al., 2013). We have demonstrated here that despite expression of a myeloid cell marker CD11b, the CD11b + Gr-1 -BM cells have lymphoid morphology. The active production of sterile transcripts at Ig loci suggests that these cells have partially committed to the B-cell lineage, while lack of VDJ rearrangement and cell surface B cell markers suggests that the differentiation toward the B cell lineage is limited. Importantly, this cell population expresses high levels of PD-L1 but does not express markers for other lymphoid cells including T cells (CD3) and NK cells (NK1.1) as well as for myeloid cells including F4/80, Gr-1 and CD115. Expression of CD11b also distinguishes IMLECs from ILCs (Walker et al., 2013), which are CD11b -. PCA analysis demonstrated that IMLECs are distinct from but close to B cells and macrophages in gene expression profiles. While at much lower frequencies, cells with the same phenotypes and functional properties were also identified in normal BM, peripheral lymphoid and non-lymphoid organs.
It is of interest to note that the hallmark of IMLECs is the high expression of cell surface PD-L1. First identified as B7-H1 (Dong et al., 1999), PD-L1 has been shown to be involved in tumor evasion of T cell immunity, both by inducing exhaustion of effector T cells and by shielding tumor cells from effector T cells (Hirano et al., 2005;Barber et al., 2006). With the induction by cytokines such as IFN-g and hypoxic tumor microenvironment, PD-L1 has been found on both tumor cells and host inflammatory cells such as myeloid derived DCs (Curiel et al., 2003;Dong et al., 2002), tumor-infiltrating myeloid derived suppressor cells (Noman et al., 2014). Recent studies have demonstrated that PD-L1 is an important biomarker and therapeutic target in cancer immunotherapy (Garon et al., 2015;Brahmer et al., 2012). Since IMLECs constitutively express high levels of PD-L1, it will be of interest to investigate their function in cancer immunity. It's worth noting that IMLECs of Raptor-deficient BM have an overall higher expression of PD-L1 than that of Raptor-sufficient BM, perhaps this reflects the indirect consequence of reduced mTORC1-mediated translation of PD-L1 negative regulators.
Since the IMLEC population expanded at the expense of granulocytes in the BM, we tested if IMLECs were derived by trans-differentiation of the granulocytes. Our genetic analyses demonstrated that inactivation of mTORC1 in the granulocytes, using Lyz2-Cre, failed to produce this subset, suggesting that loss of mTORC1 in granulocytes does not cause their trans-differentiation into IMLECs. Furthermore, since inactivation of mTORC1 in CD11b + Gr-1 + granulocytes did not affect their abundance in BM, loss of granulocytes in the mice with broad deletion of Rptor in all hematopoietic cells was not due to a cell-intrinsic requirement for mTORC1 in survival of granulocytes. Since IMLECs generally do not undergo active proliferation and are prone to apoptosis, their massive accumulation is most likely due to continuous production rather than self-renewal. Indeed, while LSKs and CMPs differentiate into IMLECs both in vitro and in vivo, IMLECs are not able to propagate in vivo (data not shown). Since Lyz2-Cre mediated Rptor deletion does not result in accumulation of IMLECs, it is obvious that IMLEC differentiation pathway is initiated before Lyz2-Cre expression which also occurs in some CMPs, and not all CMPs can give rise to IMLEC. On the other hand, since CLPs give rise to IMLEC in vitro, the possibility that other progenitor cells may give rise to IMLECs has not been ruled out.
Since heterozygous mutation of Myb causes an increase of CD11b + Gr-1 low/cells in BM (García et al., 2009), we tested if conditional deletion of Myb is sufficient to cause accumulation of IMLEC. Our data demonstrate that cell-intrinsic reduction of Myb in BM resulted in accumulation of CD11b + Gr-1 -PD-L1 + IMLECs. It is of interest to note that mice with inactivation of Rptor share many hematopoietic phenotypes with mice harboring targeted disruption of Myb, such as an increase in HSPCs (Sandberg et al., 2005) and defective B-lymphopoiesis (Fahl et al., 2009;Thomas et al., 2005;Greig et al., 2010). Moreover, consistent with the proposed roles of c-Myb in regulating precise hematopoietic commitments, CMPs with deletion of either Rptor or Myb favored differentiation to CD11b + Gr-1cells (Lieu and Reddy, 2012). Since our data show that the majority of the CD11b + Gr-1cells generated from CMPs are IMLECs, these data support the notion that IMLEC accumulation in Rptor cKO BM is caused by reduced Myb expression in CMPs.
While the mechanism by which mTORC1 down-regulates Myb remains to be fully elucidated, we have found a general up-regulation of putative miRNAs targeting Myb in HSPCs after Rptor deletion. Our data suggest that mTORC1 inactivation in HSPCs expands IMLECs by down-regulation of Myb, perhaps through increased miRNA biogenesis.
Previous studies by us and others have demonstrated a critical role for regulated mTOR signaling in hematopoiesis (Yilmaz et al., 2006;Zhang et al., 2006;Chen et al., 2008). Thus, while mTOR activation by deletion of Tsc1 complex expands the numbers of hematopoietic stem cells, it causes reduced hematopoietic stem cell function but induced leukemic stem cells in conjunction with Pten deletion. This is consistent with the generally accepted association between defective differentiation and leukemiogenesis, which forms the foundation for treatment of leukemia through induction of differentiation (Nowak et al., 2009). The expansion of IMLECs caused by mTORC1 inactivation did not lead to leukemiogenesis as IMLECs are non-dividing cells that undergo a high rate of apoptosis. Instead, our data demonstrate a new consequence of defective differentiation, namely generation of a new population of cells with distinct effector function, as discussed below.
Our RNA-seq data suggest that IMLECs broadly over-express pattern recognition receptors for innate immunity. Corresponding to a broad, although not universal TLR elevation, IMLECs mount drastically exacerbated responses to all TLR ligands tested in vitro. Intracellular cytokine staining revealed that both normal and Rptor-cKO IMLECs were among the most active producers of inflammatory cytokines. Surprisingly, although Rptor -/-BM cells have normal TLR5 levels, they produce 10-20 folds more inflammatory cytokines, such as TNF-a and MCP-1 in response to TLR5 ligand. It is therefore likely that beyond TLRs, IMLECs have acquired other features that enhance their response to TLR ligands.
Apart from BM, expanded IMLECs are broadly distributed in lymphoid and non-lymphoid organs such as lung and liver. Our data demonstrate that an increase in IMLEC numbers makes the host highly vulnerable to TLR ligands such as LPS and polyI:C, as indicated by rapid demise of mice following pIpC and low doses of LPS injections. Since necrosis of normal tissues leads to release of endogenous TLR ligands, such as HMGB1 (Scaffidi et al., 2002) and HSP70 (Millar et al., 2003), and since unregulated host response to HMGB1 leads to fatal inflammation , it is conceivable that expansion of IMLECs may also make the host more vulnerable to tissue injuryinduced inflammation. This is evidenced by elevated inflammation in acetaminophen-triggered liver necrosis model.
The severe pathological consequences may explain the paradoxical functions of mTOR in regulation of inflammation. While mTOR signaling is known to be activated by inflammatory cytokines (Chen et al., 2010) and rapamycin has been shown to inhibit production of inflammatory cytokines (Abraham and Wiederrecht, 1996), perhaps through its inhibition of mTOR-mediated NFkB activation (Lee et al., 2007), rapamycin has been shown to induce inflammation in a small number of transplantation patients (Diekmann et al., 2012;Thaunat et al., 2005;Dittrich et al., 2004). Likewise, we have observed that rapamycin increases production of inflammatory cytokines in autoimmune Scurfy mice and in mice treated with high doses of endotoxin (Chen et al., 2010). It would be of great interest to determine if rapamycin can expand IMLECs in transplantation patients and in mice that either receive high doses of endotoxin or are genetically predisposed to autoimmune disease. It is also reported that persistent mTORC1 inhibition can result in elevated inflammation, activation of STAT3 and enhanced hepatocellular carcinoma development (Umemura et al., 2014). Our findings focused on characterizing IMLECs and inflammation support the provocative findings of this report, and might provide alternative and complementary explanations. However, despite an extensive effort, we have failed to induce IMLECs by mTOR kinase inhibitor Torin2 or rapamycin in WT mice (data not shown). Therefore, under normal circumstances, pharmaceutical inhibition of mTORC1 alone cannot achieve comparable levels of inflammation as those achieved by genetic inactivation of either Mtor or Rptor. Additional conditions must be met for mTOR inhibitors to cause inflammation. While these unknown barriers have ensured safety of mTOR inhibitors in most circumstances, their breakdown may explain paradoxical induction of inflammation by rapamycin.
The concept of immune tolerance has traditionally been reserved for adaptive immunity to avoid autoimmune diseases. A multitude of mechanisms, including clonal deletion (Sha et al., 1988;Kisielow et al., 1988;Kappler et al., 1987), clonal anergy (Nossal and Pike, 1980;Schwartz et al., 1989) and dominant regulatory T cells (Sakaguchi et al., 1995), have been described to reduce selfreactive T and B cell clone sizes to avoid autoimmune diseases. On the other hand, innate immunity is known to be regulated at levels of cellular activation (Kärre et al., 1986) and cellular recruitment (Springer, 1994). However, we and others have reported that innate immune effectors, especially NK cells, have features of adaptive immunity as their immune protective function against cancer and viruses is amplified through increased population sizes (Gao et al., 2003;Sun et al., 2009). However, a regulatory mechanism to control innate effector population size for the sake of preventing self-destruction has not been described. Our discovery of a developmentally regulated mechanism to control the population size of IMLECs to avoid unwanted self-destruction, as described herein, reveals a parallel between adaptive and innate immunity to avoid potentially life-threatening inflammation and tissue damage. It is therefore of interest to consider the concept of immune tolerance in the area of innate immunity.
An important consideration is whether IMLEC is a normal population of hematopoietic cells or a population that arises after pathogenic mutations. While we have identified cells of similar phenotypes and functional properties in normal mice, they are extremely rare and thus have no obvious physiological functions unless they are substantially expanded. We have demonstrated that these cells do not undergo proliferation and are prone to apoptosis, and that their expansion depends on abnormal hematopoiesis. Therefore, the pathological consequence observed herein is only known to manifest itself if the mTORC1-Myb pathway is genetically inactivated, resulting in disruption of normal hematopoiesis. Further studies are needed to identify conditions that can lead to accumulation of IMLEC short of these known mutations.
In summary, our data demonstrate that inactivation of mTORC1 in hematopoietic stem/progenitor cells leads to generation of IMLECs, a new cell population that shares features with myeloid and lymphoid lineages. The greater than 500-fold increase in population size of IMLECs in mTORC1-or Myb-defective BM highlights the critical role for mTORC1 and c-Myb in repressing the development of sufficient number of IMLECs to cause serious inflammation and tissue damage. Our study reveals a new consequence of defective hematopoiesis and may help to extend the concept of immune tolerance to innate immunity. Mice and induction of gene deletion Rptor F/F mice (Bentzinger et al., 2008) were crossed to the C57BL/6 background for more than 10 generations. Myb F/F and Myb F/F , Rag2 -/-, CreER mice were reported previously (Fahl et al., 2009;Thomas et al., 2005;Bender et al., 2004). The interferon-inducible Mx1-Cre transgenic mice (RRID: IMSR_JAX:003556) (Kühn et al., 1995), tamoxifen-inducible Cre-ER T2 transgenic mice (RRID:IMSR_ JAX:007001) (Indra et al., 1999) and Lyz2-Cre knock-in mice (RRID:IMSR_JAX:004781) (Clausen et al., 1999) with C57BL/6 background were purchased from the Jackson Laboratory. Rptor F/F mice were crossed with Mx1-Cre mice, Lyz2-Cre mice or CreER mice to produce Rptor F/F , Mx1-Cre (cKO) mice, Rptor F/F , Lyz2-Cre +/+ mice or Rptor F/F , CreER mice, respectively. Myb F/F mice were crossed with Mx1-Cre mice to generate Myb F/F (Ctrl) and Myb F/F , Mx1-Cre (cKO) mice. Offspring of these mice were genotyped by PCR-based assays with genomic DNA from mouse tail snips. Mice were cared for in the Unit of Laboratory Animal Medicine (ULAM) at the University of Michigan, where these studies were initiated, or Research Animal Facility (RAF) of Children's National Medical Center, where the studies were completed. All procedures involving experimental animals were approved by the University Committee on the Use and Care of Animals (UCUCA) at the University of Michigan or Children's National Medical Center. Raptor Ctrl and cKO mice used in each experiment were sex-matched littermates. Mice were given 2 mg/kg body weight of pIpC (GE Healthcare Life Sciences) or 400 mg pIpC (Sigma-Aldrich) every other day for consecutive 3 to 7 times as specified by intra-peritoneal (i.p.) injection to induce Cre expression as in previous study (Tang et al., 2012). Deletion of target genes were confirmed as previously described (Bentzinger et al., 2008;Bender et al., 2004). Wild type C57BL/6 (CD45.2) mice and congenic C57BL/6 (CD45.1) mice were purchased from the Charles River Laboratories. Tamoxifen (Sigma-Aldrich) was dissolved in corn oil (Sigma-Aldrich) to 20 mg/ml and injected i.p. at 150 mg/kg/day for five consecutive days.

Histology, cytology and complete blood cell count
Ctrl and cKO mice were euthanized by CO 2 inhalation on day 30 of last pIpC treatment. For histology, tissues were fixed in 10% neutral buffered formalin for 24-48 hr and the sternums were then decalcified in Immunocal (formic acid-based decalcifier, Decal, Tallman, NY) for 24 hr. Tissues were trimmed and cassetted and processed to wax on an automated processor using standard methods. Sections were cut at 5 mm thickness and hematoxylin and eosin-stained slides prepared on an automated stainer. For cytology, BM was collected from the femoral marrow cavity with a fine diameter paintbrush dipped in sterile PBS with 5% fetal bovine serum and cytology smears were prepared by gently brushing the collected cells in parallel lines on a glass slide. Cytology slides of BM smear and cytospins of FACS-sorted BM cells were stained using a Romanowsky-based stain (Diff-Quik, Hema 3 Manual staining system, Fisher Scientific).
Histological and cytological parameters were evaluated using an Olympus BX45 light microscope at total magnifications ranging from 40 X to 100 X (oil). Histological alterations were descriptively identified. Cytological alterations were descriptively identified and quantitative BM differential counts were made using a manual differential counter and standard criteria for cell identification. Images were taken using a 12.5 megapixel microscope-mounted Olympus DP72 digital camera and accompanying software (Olympus). Complete blood cell count was performed using the Hemavet 950 Hematology System (Drew Scientific Inc.) by the Animal Diagnostic Laboratory of ULAM Pathology Cores for Animal Research in the University of Michigan.

Cells preparation and bleeding
BM cells were flushed out from the long bones (tibiae and femurs) by a 25-gauge needle with staining buffer (1XHanks Balanced Saline Solution without calcium or magnesium, supplemented with 2% heat-inactivated fetal bovine serum). Single cell suspensions of spleen, thymus, lung and lymph nodes were generated by gently squashing with frosted slides in a small volume of staining buffer. Cells from mouse peritoneal cavity were harvested as described before (Ray and Dittel, 2010). For isolation of mouse liver mononuclear cells, liver fragments were pressed through 70 mm round cell strainer (Becton Dickinson). Single-cell suspensions in a 35% Percoll solution (GE Healthcare) were centrifuged for 20 min at 800 g with brake off at room temperature. Pellet was collected and washed with staining buffer. Peripheral blood was collected by retro-orbital bleeding with heparinized capillary tubes or by submandibular bleeding with a lancet.

Flow cytometry
For surface staining, cells were stained with the indicated antibodies (Abs) in staining buffer for 20 min at 4˚C. In the characterization of surface markers for CD11b + Gr-1cells, Fcg receptors were pre-blocked by incubating cells with culture medium from hybridoma 2.4G2 (Kurlander et al., 1984) for 20 min at 4˚C. For intracellular staining, cells were first stained with the indicated surface markers Abs and then fixed with Cytofix/Cytoperm buffer (BD Biosciences) for 1-2 hr at 4˚C, followed by incubation with Cytoperm Plus buffer (BD Biosciences) for 15 min at room temperature (R.T.). After refixing for 15 min at R.T., cells were incubated with antibodies or isotype controls for 20 min (anti-TNF/IgG, anti-Ki-67Abs) or overnight (anti-c-Myb/IgG Abs) and further stained with the secondary Ab if necessary. BrdU labeling experiments were performed per the manufacture's instruction (BD Biosciences), as previously reported (Chen et al., 2008;Tang et al., 2012). Apoptosis assays by 7-AAD and Annexin V (BD Biosciences) were according to manufacturer's instructions. All FACS analyses were performed on a BD LSR II or a Canto II Flow Cytometer, and data were analyzed with FlowJo software (Tree Star, Inc.). The enrichment of Lin -BM cells was performed using MACS beads from mouse Lineage Cell Depletion Kit (Miltenyi Biotec). CD11b + Gr-1 -BM cells from Rptor F/F , Mx1-Cre mice, Lin -CD11b + PD-L1 + IMLECs / LSK/CLP/ CMP populations from Raptor Ctrl/cKO or WT mice, CD11b + Gr-1 + / CD11b -Gr-1 -BM populations from Rptor F/F , Lyz2-Cre +/+ mice, CD11b high Gr-1 -F4/80 + peritoneal macrophages were sorted using FACSAria II or Influx cell sorter (BD Biosciences). The detailed information on Abs used in this study is in Supplementary file 3.

BM cells transplantation
C57BL/6 Ly5.2 (CD45.1 + ) recipient mice at the age of 6-12 weeks old were lethally irradiated for total 900-1,100 rads with a Cs-137 g-ray source or a RS 2000 X-ray irradiator (Rad Source Technologies, Inc.). Indicated donor BM cells (whole BM or FACS-sorted BM cells) were transplanted into recipients through the mice tail by intra-venous (i.v.) injection within 24 hr after irradiation (Tang et al., 2012). At different time points post-transplantation, peripheral blood from the recipient mice was analyzed by flow cytometry to test the reconstitutions.

Conventional PCR and quantitative PCR
Genomic DNA was isolated from BM cells by DNeasy Blood and Tissue Kit (Qiagen) as per manufacturer's instructions. Total RNA was isolated using TRIzol (Invitrogen) or ReliaPrep RNA Cell Miniprep System (Promega). Reverse transcription was carried out using random hemaxmer primers and SuperScript II Reverse Transcriptase (Invitrogen). Conventional PCR was performed using GoTaq Green Master Mix (Promega). Quantitative PCR (q-PCR) was performed by the 7500 real-time PCR system using Power SYBR Green Master Mixture (Applied Biosystems). Fold changes were calculated according to the DDCT method (Livak and Schmittgen, 2001). The primers used for conventional PCR and q-PCR are listed in Supplementary file 4.

Ig gene rearrangement test by V(D)J analysis
Immunoglobulin (Ig) gene recombination was determined using genomic DNA as previously described (Riddell et al., 2014). For Heavy chain, a semi-nested PCR strategy was employed to amplify the framework regions of VH to specific sites of JH. First round amplification of 25 cycles was performed with primers FR/JH1 (70˚C annealing/20 s extension). Second round amplification of 35 cycles was with primers FR/JH2 (65˚C annealing/30 s extension). Light chain (Igk and Igl) recombination was tested by primers V k /J k 5 and V l 1/J l 1,3 following previous report (Cobaleda et al., 2007). Sequences of primers used are listed in Supplementary file 4.

TLR stimulation and inflammatory cytokine assay
For in vitro TLR stimulation, fresh BM cells were seeded in a 12-well plate with a density of 4 Â 10 6 cells/ well(1 ml medium/well) or in a 48-well plate with a density of 1 Â 10 6 cells/ well (200 ml medium/well). The culture medium was RPMI 1640 (Life Technologies) supplemented with 10% heatinactivated fetal bovine serum (Hyclone), 100 units/ml of penicillin and 100 mg/ml of streptomycin (Gibco). BM cells were stimulated with 1 mg/ml LPS (Sigma-Aldrich, from O111:B4 E.coli) or a panel of TLR agonists (InvivoGen) for 16 hr. For intracellular cytokine staining, the protein transport inhibitor Brefeldin A (eBioscience) was added to the culturing medium during the last 4 hr of incubation.
cDNA library preparation and RNA sequencing FACS-sorted CD11b + Gr-1 -BM cells from Raptor cKO mice and whole BM cells from Raptor Ctrl mice were used for RNA isolation with TRIzol Reagent (Life Technologies) per manufacturer's instructions. The cDNA libraries were constructed following the standard Illumina protocols by TruSeq RNA and DNA sample preparation kits (Illumina). Briefly, beads containing oligo (dT) were used to isolate poly(A) mRNA from total RNA. Purified mRNA was then fragmented in fragmentation buffer. Using these short fragments as templates, random hexamer-primers were used to synthesize the firststrand cDNA. The second-strand cDNA was synthesized using buffer, dNTPs, RNase H and DNA polymerase I. Short double-stranded cDNA fragments were purified for end repair and the addition of an 'A' base. Next, the short fragments were ligated to Illumina sequencing adaptors. DNA fragments of a selected size were gel-purified and amplified by PCR. The amplified cDNA libraries were quality validated and then subjected to 50 nt single-end sequencing on an Illumina HiSeq 2000 at the University of Michigan DNA Sequencing Core.

RNA-seq gene expression analysis
The reference sequences used were genome and transcriptome sequences downloaded from the UCSC website (version mm10). Clean reads were respectively aligned to the reference genome and transcriptome using Tophat (RRID:SCR_013035) (Kim et al., 2013). No more than two mismatches were allowed in the alignment for each read. Reads that could be uniquely mapped to a gene were used to calculate the expression level. The gene expression level was measured by the number of uniquely mapped reads per kilobase of exon region per million mappable reads (RPKM) and was calculated by DEGseq (RRID:SCR_008480) (Wang et al., 2010). The formula was defined as below: RPKM ¼ 10 6 C NL 10 3 in which C was the number of reads uniquely mapped to the given gene; N was the number of reads uniquely mapped to all genes; L was the total length of exons from the given gene. For genes with more than one alternative transcript, the longest transcript was selected to calculate the RPKM.

PCA, differential and cell-specific expression analysis
The RPKM method eliminates the influence of different gene lengths and sequencing discrepancies on the gene expression calculation. Therefore, the RPKM value can be used for comparing the differences in gene expression among samples. The RPKM value of all RNA-seq raw data were calculated according to the same workflow as stated above.
A function was implemented in the R software to perform principal component analysis (PCA). This function computes the eigenvalues and eigenvectors of the dataset (23498 genes) using the correlation matrix. The eigenvalues were then ordered from highest to lowest, indicating their relative contribution to the structure of the data. The projection of each sample defined by components was represented as a dot plot to generate the PCA figures.
Selected samples were then pooled by subtypes and a two sided t-test with FDR (False discovery rate) of 0.05 and fold change of 4 was performed to identify differentially expressed genes between IMLECs and other subtypes (mean RPKM values of genes in two subtypes both below five were deleted). For subtype-specific genes identified, a one-sided t-test (null hypothesis is greater) was performed with FDR of 0.01 and fold change of 4 contrasting each subtype in turn versus all other subtypes pooled, and statistically significant genes were assigned to the respective subtype

Genome browser display of immunoglobulin expression
To facilitate the global viewing of transcript structure and gene expression quantity of immunoglobulin, an interface in which RNA-seq gene expression of immunoglobulin can be viewed in Genome Browser display was constructed. For viewing and analysis, the UCSC Genome Browser (http://www. genome.ucsc.edu, RRID:SCR_005780) (Kent et al., 2002) with the mm10 version of the mouse genome was used. For each base in each cell type, the normalized number of aligned reads count was defined as below: Normalized number of aligned reads count ¼ A* P n k¼0 N k n in which N was the number of reads uniquely mapped to the given base in each cell type; n was the number of replicates for each cell type samples; A was a constant t in which takes the value of 3.5E + 7. The normalized number of aligned reads count for each base in each cell type was stored as WIG file and was uploaded to UCSC genome browser. For each track, y-axis depicted the normalized number of aligned reads count and x-axis depicts physical distance in bases along the chromosome.

Statistical analysis
All the data are presented as mean ±SD. Unless otherwise indicated, two-tailed, unpaired student's t tests were used for comparison between two experimental groups. In Figure 8I and Figure 8J, where the data do not follow normal distribution, Mann-Whitney tests were used. The log-rank tests were used for the Kaplan-Meier survival analysis. Statistical significance was determined as p<0.05 (*p<0.05; **p<0.01; ***p<0.001). The following previously published datasets were used: