Endothelial cell-specific expression of serine/threonine kinase 11 modulates dendritic cell differentiation

In the bone marrow, classical and plasmacytoid dendritic cells (DC) develop from the macrophage-DC precursor (MDP) through a common DC precursor (CDP) step. This developmental process receives essential input from the niche in which it takes place, containing endothelial cells (EC) among other cell types. Here we show that targeted deletion of serine/threonine kinase 11 (Stk11) encoding tumor suppressor liver kinase b1 (Lkb1) in mouse ECs but not DCs, results in disrupted differentiation of MDPs to CDPs, severe reduction in mature DC numbers and spontaneous tumorigenesis. In wild type ECs, Lkb1 phosphorylates polypyrimidine tract binding protein 1 (Ptbp1) at threonine 138, which regulates stem cell factor (Scf) pre-mRNA splicing. In the absence of Lkb1, exon 6 of Scf is spliced out, leading to the loss of Scf secretion. Adeno-associated-virus-mediated delivery of genes encoding either soluble Scf or the phosphomimetic mutant Ptbp1T138E proteins rescued the defects of MDP to CDP differentiation and DC shortage in the endothelium specific Stk11 knockout mice. In summary, endothelial Stk11 expression regulates DC differentiation via modulation of Scf splicing, marking the Stk11-soluble-Scf axis as a potential cause of DC deficiency syndromes.

D endritic cells (DC) are a major type of antigen-presenting cell. DCs effectively link the innate with the adaptive immune systems by inducing the activation and differentiation of various immune effector cells, such as naïve B and T lymphocytes 1,2 . These unique properties place DCs at the center of immune homeostasis 3,4 . Insufficient numbers of DCs lead to defective or dysregulated tumor immunosurveillance, or conversely, attack self-antigens to induce tissue inflammation and autoimmune disease 5,6 . A better understanding of the intrinsic and extrinsic cellular mechanisms regulating DC differentiation will identify new therapeutic targets for both cancer and autoimmune diseases.
DC developed from macrophage-DC precursor (MDP) into common DC precursor (CDP), then to pre-classical DCs that finally give rise to classical DCs and plasmacytoid DCs [7][8][9] . In mouse bone marrow, DCs and monocytes share a common progenitor, the MDP, whereas a distinct progenitor, the CDP, is dedicated to DC production 7 . Although this cascade of DC progenitors has been identified, it remains unknown if the bone marrow niche regulates DC differentiation. Bone marrow niche is a local perivascular microenvironment, created mainly by endothelial cells (ECs) and mesenchymal stromal cells, maintaining or regulating hematopoietic stem cell (HSC) and hematopoietic progenitor cell (HPC) development and function 10 . Particular HSCs or HPCs may occupy distinct bone marrow niches in vivo 11 , as most HSCs lie in direct contact with bone marrow microvessels, whereas the HPCs are less likely than HSCs to be immediately adjacent to sinusoidal endothelium [12][13][14][15] . Although extensive work has been carried out to identify and characterize the perivascular niche of HSC quiescence and self-renewal 10,16,17 , it remains unknown if the bone marrow niche regulates the differentiation of hematopoietic DC precursors, such as of the MDPto-CDP, and how disorders of the bone marrow niche function trigger DC dysfunction-related diseases.
Stem cell factor (Scf; also known as KitL) is a key cytokine in hematopoiesis and binds to the c-Kit tyrosine receptor kinase [18][19][20] . ECs and leptin receptor-expressing perivascular stromal cells are the main sources of Scf required for hematopoiesis in normal mouse bone marrow 17 . There are two isoforms of Scf, membrane-bound and soluble, generated by alternative mRNA splicing 21 . Membrane-bound Scf seems to be much more important for HSC maintenance and self-renewal, as HSCs are depleted in compounds of steel and steel Dickie mutant mice (Kitl Sl /Kitl Sl-d ; also called Sl/Sl d ) 22,23 , which express normal levels of soluble Scf but lack the membrane-bound form 24 . In contrast, soluble Scf is more important for the normal development of restricted myeloid lineage 25 . Although recent data demonstrate that Scf is strictly required for the long-term expansion of DC precursors 26 , and c-Kit inhibitor fully blocks maturation of DCs from progenitors 27,28 , how Scf regulates the proliferation and retention of the restricted DC progenitor population in the bone marrow is unknown.
Here, we report that Stk11 ec−/− mice spontaneously develop tumors with splenomegaly and disorganized immune activation characterized by decreased number of DCs and myeloid proliferation. Mechanistically, Lkb1 phosphorylates Ptbp1 to promote Scf exon 6 retention; thus, switches alternative splicing of Scf mRNA to favor soluble Scf, which regulates DC differentiation through cell-detached means of communication between bone marrow ECs and DC precursors in vivo.

Results
Endothelial cell-specific Stk11-deleted mice (Stk11 ec−/− ) spontaneously develop tumors. Endothelial cell (EC)-specific Stk11deleted (Stk11 ec−/− , Stk11 fl/fl Cdh5 Cre ) mice were generated as described previously 38 . Over time, natural death (found dead or moribund and euthanized) in Stk11 ec−/− mice occurred and was frequently accompanied by clinical signs attributable to tumor burden, including weight loss, dyspnea, or observable masses ( Fig. 1a and Tables 1 and 2). Hematoxylin-and-eosin (H&E) staining and histopathological examination further confirmed tumor development with different pathological diagnoses, such as carcinoma, adenocarcinoma, lymphoma, leukemia, and endometrial carcinoma (Fig. 1c). Tumor lesions were observed in multiple tissues of Stk11 ec−/− mice, including lung, uterus, kidney, ureter, abdominal cavity, bone, and lymph node (Supplementary Table 3). The overall incidence of tumor-bearing mice increased from 19.12% at 14 months of age, to 73.91% at 24 months of age in Stk11 ec−/− mice (Fig. 1b and Supplementary Table 3). The most common neoplasm at 14 months of age found in Stk11 ec−/− mice was an abdominal tumor (7.35%), and the most common neoplasm at 24 months and 30 months of age was lung tumor (21.74% and 14.29%, respectively) (Supplementary Table 3).
To explore whether tumors developed in Stk11 ec−/− mice originated from Stk11 deleted ECs, we performed immunochemical staining of Lkb1 (encoded by Stk11) in tumor sections from Stk11 ec−/− mice. As shown in Fig. 1c, Lkb1 expression in cancer cells and cells in para-cancer areas were comparable (both Lkb1 positive). To validate this observation, we further generated Stk11 ec (EYFP) −/− mice (Stk11 fl/fl ROSA EYFP Cdh5 cre ) to explore tumor origin and biology in vivo. In these mice, Stk11-depleted Cdh5 + cells and their progeny were irreversibly labeled with an enhanced yellow fluorescent protein (EYFP). In-depth microscopic analysis revealed that ECs were abundantly labeled with EYFP (>90%), whereas EYFP staining in non-ECs in the spontaneously developed tumor from Stk11 ec (EYFP)−/− mice was negligible (Fig. 1d). Given the evidence that tumor cells are Lkb1-positive and strict EYFP expression in tumor ECs (identified by positive staining for vWF [vWF + ]) in the lineagetracking mice (Stk11 ec (EYFP)−/− ), we concluded a non-cellautonomous effect of EC Stk11 deficiency on tumorigenesis or tumor suppression.
Stk11 ec−/− mice exhibit disorganized immune activation. In addition to spontaneous tumor development in Stk11 ec−/− mice, spleens of 12-week-old Stk11 ec−/− mice weighed approximately three times more than those of littermate controls (Fig. 2a). Correlated with increased organ size and weight, the number of whole spleen cells in Stk11 ec−/− mice was significantly higher than that in littermate controls (Fig. 2a). Increased number of spleen cells was associated with extensive megakaryocyte infiltration (identified by positive staining for CD61 [CD61 + ]) in spleen red pulp (Fig. 2b) lymphadenopathy and enlarged thymus (Fig. 2a). Histopathological analysis revealed massive lymphocyte infiltration into multiple organs, including intestine, liver, pancreas, and lung in Stk11 ec−/− mice ( Supplementary Fig. 1). Overall, these results indicate disorganized immune activation in Stk11 ec−/− mice.
Differentiation of MDP into CDP is suppressed in Stk11 ec−/− mice. Next, we sought to identify the key reason for DC depression and myeloid proliferation in Stk11 ec−/− mice. Firstly, we checked the number of DC progenitors in bone marrow (preclassical DCs: Lin − CD45 + CD11c + MHC II -SIRP-α int Flt3 + ; MDP: Lin − c-Kit high CD115 + CX3CR1 + Flt3 + ; CDP: Lin − c-Kit low CD115 + Flt3 + ). The number of pre-classical DC (Pre-cDC) and CDP showed an 80% reduction (Fig. 3d, e), whereas the number of MDP increased (Fig. 3e). We also noticed that there was no significant difference in the number of HSC, multipotent progenitor (MPP), lymphoid-primed multipotent progenitor (LMPP), common lymphoid progenitor (CLP), common myeloid progenitor (CMP), and granulocyte-monocyte progenitor (GMP) between Stk11 ec−/− mice and littermate control mice (Supplementary Fig. 3). Taken together, these data suggest that Stk11 ec−/− mice underwent a myeloid shift toward macrophages, monocytes, and granulocytes, associated with depressed DC lineage and MDP-to-CDP differentiation, independent of HSC or other hematopoietic progenitor cell (HPC) maintenance and differentiation ( Supplementary Fig. 4). Next, we sought to determine whether Stk11 deletion in ECs regulates terminal differentiated DC function. To this end, we challenged Stk11 ec−/− mice with B16 tumors expressing the model antigen ovalbumin (B16-OVA). Compared with WT mice, tumor growth was significantly increased in Stk11 ec−/− ( Supplementary Fig. 5a). Immunization of DCs from either WT or Stk11 ec−/− mice at same dose delayed B16 tumor growth in Stk11 ec−/− mice ( Supplementary Fig. 5a). Additionally, B16-OVA bearing Stk11 ec−/− mice immunized with either WT or Stk11 ec−/− derived DCs showed similar OVA epitope (SIIN-FEKL) complex expression ( Supplementary Fig. 5b). Altogether, these results demonstrated that Stk11 deletion in ECs does not affect the function of terminal differentiated DCs. Selective EC-deletion of Stk11 in adult mice also show obstructed MDP-to-CDP differentiation. Recent studies proposed that HSC and EC progenitors originate from a common mesodermal precursor 41,42 . These populations diverge before the onset of definitive hematopoiesis. After the onset of definitive hematopoiesis, Cdh5 (VE-cadherin) is considered a specific marker of ECs 43 . In this regard, we checked whether Cdh5-Cre recombinase leaked to DCs in vivo. As showed in Supplementary  Fig. 6, Cdh5-Cre driven Stk11 exon 3 to exon 6 excision mainly happens in ECs, not DCs.
Additionally, we generated tamoxifen-inducible Stk11 ec−/− mice by cross-breeding Stk11 fl/fl mice with VECad-Cre ERT2 mice 44 45 . Interestingly, the size of spleen, lymph node, and thymus showed no significant difference between Stk11 dc−/− mice and littermate control mice ( Supplementary Fig. 8a). Moreover, the spleen structure of Stk11 dc−/− mice is comparable to littermate control mice, without accumulated megakaryocyte infiltration in spleen red pulp ( Supplementary Fig. 8b). Importantly, depletion of Stk11 in DC lineage does not affect numbers of classical DCs, plasmacytoid DCs, macrophages, or the MDP/CDP ratio in vivo (Supplementary Fig. 8c-e). Taken together, we conclude that selective deletion of Stk11 in DC does not affect MDP-to-CDP differentiation.

Cell-detached means of communication between bone marrow
ECs and DC precursors in vivo. Given the evidence of obstructed MDP-to-CDP differentiation in Stk11 ec−/− mice, but not Stk11 dc−/− mice, we next sought to examine the fate and distribution of MDP and CDP in mice bone marrow. We used green fluorescent protein (GFP) Cx3cr1 knock-in mice (hereafter designated as Cx3cr1 GFP ). In these mice, MDP and their progeny cells are irreversibly labeled with GFP. Interestingly, the distribution of Cx3cr1-GFP + c-Kit + cells in bone marrow indicating MDP/CDP were more common in the bone marrow perivascular area but not directly in contact with vessels ( Fig. 4a). Approximately 85% of MDP/CDP were 20 μm or far away from a sinusoidal blood vessel (Supplementary Movie 1), suggesting a cell-detached means of communication between bone marrow ECs and DC precursors.
Soluble Scf is a key extrinsic perivascular niche cytokine for DC differentiation in mouse bone marrow. Deeper analysis of the most differentially expressed genes between primary ECs isolated from WT and Stk11 ec−/− mice highlighted extracellular protein binding as the most changed pathway based on the number of changed genes and statistical significance ( Supplementary  Fig. 9a). Among the factors involved in extracellular protein binding, we found that Stk11 deletion in ECs reduced soluble Scf, whereas increased membrane-bound Scf mRNA and protein levels (Fig. 4b). Stk11 ec−/− mice showed a 50% reduction in serum soluble Scf level, compared with littermate control mice (Fig. 4c). Additionally, Jak2 and Akt phosphorylation decreased in MDP, but not terminal differentiated DC (classical DC and plasmacytoid DC) in Stk11 ec−/− mice compared with their littermate controls ( Supplementary Fig. 9b, c).
To explore whether Scf is required for MDP differentiation into CDP, unfractionated bone marrow cells from Stk11 ec−/− mice and littermate control mice were treated with or without ISCK03, a cell-permeable inhibitor of the Scf receptor c-Kit. In the presence of ISCK03 in culture, both DC production and colony formation reduced sharply in bone marrow cells of littermate control mice, associated with an increased number of megakaryocytes (Fig. 4d). Interestingly, ISCK03 did not further impair DC differentiation in the Stk11 ec−/− mouse bone marrow cultures compared with vehicle (Fig. 4d).
Next, we cultured the unfractionated bone marrow cells from Stk11 ec−/− mice and littermate control mice in the presence or absence of recombinant murine soluble Scf (rmScf). Relative to littermate control mice, Stk11 ec−/− mouse bone marrow cells showed a 60% reduction in the percentage of colonies and DC production without rmScf supplementation (Fig. 4e). Addition of rmScf markedly increased colony number and DC production in littermate control mice and Stk11 ec−/− mouse bone marrow cultures (Fig. 4e), suggesting a compensatory function of rmScf in DC differentiation in Stk11 ec−/− mouse bone marrow cells. Similarly, bone marrow cells from the Stk11 ec−/− mice showed a ten-fold induction of megakaryocytes compared with the unfractionated bone marrow cells of WT mice, whereas in the presence of rmScf, there is negligible megakaryocyte production in either the littermate control mice or the Stk11 ec−/− bone marrow in culture (Fig. 4e).
To further distinguish whether DC shortage in Stk11 ec−/− mice is due to hematopoietic cells or defective niche cells, Stk11 ec−/− mice were transplanted with bone marrows from WT or Stk11 ec−/− mice, and vice versa. As shown in Supplementary  Fig. 9d, there are no difference in the number of classical DCs and plasmacytoid DCs between WT mice receiving WT or Stk11 ec−/− mice bone marrows. In contrast, Stk11 ec−/− mice receiving WT mice bone marrow still showed lower number of classical DCs and plasmacytoid DCs than those WT mice ( Supplementary  Fig. 9d). Consistent with this observation, MDPs isolated from either WT or Stk11 ec−/− mouse bone marrow showed no difference on progenitor cell maintenance when cultured on standard AFT024 feeder cells ( Supplementary Fig. 9e). Taken together, those results indicated an extrinsic perivascular niche effect of soluble Scf on DC differentiation.
Lkb1-dependent phosphorylation of Ptbp1 inhibits Scf alternative splicing. Given the evidence of decreased soluble Scf and increased membrane-bound Scf mRNA and protein in Stk11deleted ECs, we next thought to identify the splicing element that directly regulates Scf alternative splicing. The particular high degree of sequence conservation of Scf exon 6 during evolution raises the hypothesis that exon 6 might be the target of a sequencespecific RNA-binding protein 21 , and bioinformatic analysis revealed several candidates (Supplementary Fig. 11a). Among these, only polypyrimidine tract binding protein 1 (Ptbp1) knockdown enhanced Scf alternative splicing (Fig. 5a). We next designed and generated four different peptides nucleic acid (PNA) isoforms based on the Scf exon 6 sequence to inhibit RNA-binding protein binding (Supplementary Fig. 11b). Among these, only PNA3 which prohibited Ptbp1 binding to the mRNA (Fig. 5b), dramatically reduced the soluble Scf/membrane-bound Scf ratio (Fig. 5c). Interestingly, an efficient knockdown of Stk11 inhibited Ptbp1 binding to mRNA (Fig. 5d), suggesting that Lkb1 may promote Scf alternative splicing through Ptbp1. We next tried to determine how Lkb1 regulates Ptbp1 binding. Because phosphorylation of Ptbp1 dramatically decreases upon Stk11 deficiency ( Supplementary Fig. 11c, d), Ptbp1-driven Scf alternative splicing is likely due to Lkb1 phosphorylation. To this end, we performed in vitro kinase assay and found that Lkb1 directly phosphorylates Ptbp1 (Fig. 5e). Additionally, Stk11 deletion in ECs decreased Ptbp1-Lkb1 binding (Supplementary Fig. 11e). Computer alignment with an optimal Lkb1 substrate motif showed that Ptbp1 residue Thr138 is a potential Lkb1 phosphorylation target site  Supplementary Fig. 11f). Therefore, we generated two sitedirected mutant constructs of Ptbp1 (Ptbp1 T138A and Ptbp1 T138E ). Phosphorylation of Ptbp1 at Thr138 was abolished in ECs expressing Ptbp1 T138A (Fig. 5f), implying that Thr138 is the target of Lkb1-mediated phosphorylation. We further determined whether Ptbp1 Thr138 phosphorylation was required for the alternative splicing of Scf. To this end, we found that Ptbp1 T138A mutation decreased the soluble Scf/membrane-bound Scf ratio, whereas Ptbp1 T138E mutation increased soluble Scf/membranebound Scf ratio (Fig. 5g) and increased PTBP1-Scf (exon 6) binding ( Supplementary Fig. 11g). Deeper analysis of the most differentially expressed genes among ECs overexpressed Ptbp1 T138A , Ptbp1 T138E , and Ptbp1 wt mutant further highlighted protein binding and adheshion as the most changed pathyway based on the number of changed genes and statistical significance ( Supplementary Fig. 11h). Notably, the protein binding and adheshion pathway also stood out in the RNAseq data set with Stk11 deficiency (Supplementary Fig. 9a). Taken together, these findings imply that Lkb1 phosphorylated Ptbp1 at Thr138, initiating Ptbp1 binding to Scf exon 6 and alternative splicing.
Administration of Ptbp1 T138E restores MDP-to-CDP differentiation and DC lineage in Stk11 ec−/− mice in vivo. Given the evidence that phosphorylation of Ptbp1 Thr138 promotes soluble Scf generation, we next sought to identify the function of Ptbp1 Thr138 phosphorylation in DC lineage differentiation in vivo. To this end, we generated recombinant AAV encoding Ptbp1 T138E (AAV-Ptbp1 T138E ) or acGFP (AAV-acGFP). SCF expression and MDP-to-CDP differentiation was restored in the bone marrow of Stk11 ec−/− mice treated with AAV-Ptbp1 T138E ( Supplementary  Fig. 10a, Fig. 5h). The frequencies of classical DC and plasmacytoid DC significantly increased in Stk11 ec−/− mice treated with AAV-Ptbp1 T138E (Fig. 5i). Consistent with this observation, spleen size and megakaryocyte infiltration decreased with AAV-Ptbp1 T138E treatment in Stk11 ec−/− mice compared with AAV-acGFP transduction (Supplementary Fig. 10b, c).

Discussion
In this study, we showed that Stk11-guided alternative splicing of Scf in bone marrow ECs is strictly and non-cell-autonomously required for MDP differentiation into CDP. Mechanistically, The most important finding from this study is that we have for the first-time presented evidence that Stk11 in bone marrow ECs plays a key role in DC differentiation, and thus regulates immune homeostatic equilibrium. Stk11 ec−/− mice showed a disruption of MDP-to-CDP differentiation and reduced DC number. Consequently, inadequate DCs led to defective or dysregulated tumor immune surveillance, manifested as spontaneous tumor development in Stk11 ec−/− mice. Furthermore, the reduced number of DCs led to aberrant lymphocyte activation, inducing tissue inflammation, and disorganized immune activation. This also resulted in compensatory hematopoiesis, as myeloid proliferation disorder and splenomegaly in the Stk11 ec−/− mice. These phenomena (high premature death rate, spontaneous tumor development, aberrant lymphocyte activation, and myeloid proliferation) in Stk11 ec−/− mice are extremely similar to the clinical DC deficiency syndrome 46,47 . Some patients with DC deficiency syndromes become seriously ill and die in their early 30s. Those who survive show a high incidence of solid tumors, autoimmune disease, and myeloproliferation 46,47 . Notably, there is no clear identified pathogen or cause for the syndromes, except two mutations, GATA-binding factor 2 (GATA2) and interferon regulatory factor 8 (IRF8) 46,47 . Our findings constitute an important example of how the loss of tumor suppressor in bone marrow ECs or niche cells drives specific DC precursor cell differentiation and spontaneous tumor development. We also suggest that the Stk11-Ptbp1-Scf axis may be a potential new therapeutic avenues for DC deficient syndrome for which there are no treatment options right now.
Another significant finding of this study is that disrupted MDP-to-CDP differentiation after Stk11 deletion in ECs appears to depend on soluble Scf, as shown by the evidence that Stk11 deletion in ECs decreased soluble Scf and increased membranebound Scf. The addition of soluble Scf restored DC differentiation and the MDPs/CDPs ratio both in vivo and in vitro. Scf has long Fig. 4 Stk11 in ECs regulates soluble stem cell factor (sScf) to drive DC differentiation. a Deep imaging of Cx3cr1-GFP + cells (green), hematopoietic progenitors (c-Kit + , red), and blood vessels (laminin, gray) in digitally reconstructed bone marrow (300 μm thick). Scale bar: 20 μm. Images are representative of three independent experiments. b, c Supernatant of cultured primary bone marrow ECs from WT or Stk11 ec−/− mice was subjected to soluble Scf (sScf) ELISA (b, upper) (n = 4 each group). *P = 0.005 versus WT by Student's t-test (two-sided). Lysates of primary bone marrow ECs from WT or Stk11 ec−/− mice were analyzed by western blotting for Scf, Lkb1, and β-Actin (b, lower), and by PCR for Scf (c, upper). The serum Scf level from WT or Stk11 ec−/− mice was detected by ELISA (c, lower; 12-weeks-old, mixed-gender, WT, n = 14; Stk11 ec−/− , n = 13). *P < 0.001 versus WT by Student's t-test (two-sided). The blots are representative results of three independent experiments. d Bone marrow cells (1 × 10 6 cells/per well) isolated from WT or Stk11 ec−/− mice were cultured with or without c-Kit inhibitor ISCK03 (1 µM) for 6 days to generate DCs. Primary DCs were harvested from each dish by collecting non and loosely adherent cells and stained for CD11c, detected by flow cytometry. DCs were defined as CD11c + cells. The colonies (red head) and megakaryocytes (green head) were stained with Giemsa. Scale bar: 1000 μm. Bar graph summarizes the numbers of DCs and megakaryocytes per cm 2 in culture. *P < 0.001 versus WT + vehicle by Student's t-test (two-sided). e Bone marrow cells (1 × 10 6 cells/per well) isolated from WT or Stk11 ec−/− mice were cultured with or without recombinant murine stem cell factor (rmScf, 50 ng/mL) for 6 days to generate DCs. Primary DCs were harvested from each dish by collecting nonadherent and loosely adherent cells and stained for CD11c, detected by flow cytometry. DCs were defined as CD11c + cells. The colonies ( Red head) and megakaryocytes (green head) were stained with Giemsa. Scale bar: 1000 μm. Bar graph summarizes the numbers of DCs and megakaryocytes per cm 2 in culture. *P < 0.001 versus WT + Vehicle and † P < 0.001 versus Stk11 ec−/− + vehicle by Student's t-test (two-sided). f Flow cytometry analysis of spleen/lymph node cells isolated from WT or Stk11 ec−/− mice treated with either AAV-acGFP or AAV-sScf and stained for CD11c and MHC II or CD11c and B220. Bar graph summarizes frequency of spleen and lymph node cDCs and pDCs (12-weeks-old, mixed-gender, n = 5-6 each group). *P < 0.001 versus WT and † P < 0.001 versus Stk11 ec−/− by nonparametric Mann-Whitney U test (two-sided). g Flow cytometry analysis of bone marrow cells isolated from WT or Stk11 ec−/− mice treated with either AAV-acGFP or AAV-sScf and stained for Lin, c-Kit, CD115, CX3CR1, and Flt3. Bar graph summarizes the ratio of MDP/CDP in bone marrow (12-weeks-old, mixed-gender, n = 5-7 each group). *P < 0.001 versus WT and † P < 0.001 versus Stk11 ec−/− by nonparametric Mann-Whitney U test (two-sided).  Blots are representative of three independent experiments. g PCR analysis was carried out to detect soluble Scf/membrane-bound Scf ratio in BAECs transfected with Ptbp1 wt and mutant plasmid. *P = 0.003; † P = 0.007 by nonparametric Mann-Whitney U test (two-sided). h Flow cytometry analysis of bone marrow cells isolated from WT or Stk11 ec−/− mice treated with either AAV-acGFP or AAV-Ptbp1 T138E and stained for Lin, c-Kit, CD115, CX3CR1, and Flt3. Bar graph summarizes the ratio of MDP/CDP in bone marrow (12-weeks-old, mixed-gender, n = 5-6 each group). *P < 0.001 versus WT and † P < 0.001 versus Stk11 ec−/− by nonparametric Mann-Whitney U test (two-sided). i Flow cytometry analysis of spleen/lymph node cells isolated from WT or Stk11 ec−/− mice treated with either AAV-acGFP or AAV-Ptbp1 T138E and stained for CD11c and MHC II or CD11c and B220. Bar graph summarizes the frequency of cDCs and pDCs in spleen, and lymph node (12-weeks-old, mixed-gender, n = 5-6 each group). *P < 0.001 versus WT and † P < 0.001 versus Stk11 ec−/− by nonparametric Mann-Whitney U test (two-sided).

Vehicle
been suggested to be required for HSC survival, as HSCs are depleted in Tie2-Cre-mediated EC-Scf-deleted mice (Scf fl/-Tie2 cre mice) 17 . In our case, disrupted MDP-to-CDP differentiation in Stk11 ec−/− mice seems independent of HSC and other lineagerestricted HPC, as HSC, MPP, LMPP, CLP, CMP, GMP, and other restricted-lineage cells were unaffected. Membrane-bound Scf is not reduced (actually increased) in Stk11-deleted ECs, while membrane-bound Scf is more important for HSC maintenance and self-renewal 22,23 , whereas soluble Scf is more important for the normal development of restricted lineages, e.g., the DC lineage in this case. The different distribution of membrane-bound and soluble Scf raised the possibility that physical distance within the niche between HSC/HPC and ECs may affect the regulation and fate of stem cells and progenitor cells. Deep imaging of mouse bone marrow further supports this hypothesis, as shown by the fact that MDP/CDP reside mainly in perivascular niches adjacent to, but not in contact with, bone marrow ECs.
In the current study, we present evidence that Lkb1 in ECs regulates membrane-bound Scf/soluble Scf expression through Ptbp1-phosphorylation-mediated Scf alternative mRNA splicing. It has been reported that an acidic microenvironment might be responsible for the switch of Scf mRNA splicing in favor of membrane-bound Scf 21 . But it is unclear how this alternative mRNA splicing happens. In this study, using four different sequence-specific PNAs to compete with the target Scf exon 6 sequence, we found that Ptbp1 is the key to Scf exon 6 skipping and alternative splicing. Lkb1-dependent phosphorylation of Ptbp1 at Thr138 is required for Ptbp1 binding to Scf pre-mRNA exon 6. The binding thus switches Scf mRNA alternative splicing in favor of soluble Scf.
In sum, our data demonstrates that Stk11 in bone marrrow ECs is strictly and non-cell-autonomously required for DC lineage and MDP-to-CDP differentiation through Ptbp1-phosphorylationdependent Scf alternative splicing. These newly recognized niche features of DCs open up the possibility of expanding or reducing DC numbers by lineage-restricted cytokines or pathways in vivo, and provide new therapeutic opportunities for DC deficiency syndromes.
Two-photon deep imaging of bone marrow. Freshly dissected tibias from 8 to 12week-old mice were fixed for 6-8 h in 4% paraformaldehyde in phosphate-buffered saline (PBS) at 4°C. Tibias were sectioned perpendicular to the long axis into 300μm thick sections using a Leica VT100 S vibratome. Sections were blocked in a staining solution containing anti-CD16/32 mouse Fc-blocking antibody overnight on a rotator at room temperature. The staining solution contained 10% dimethyl sulfoxide (Sigma, #276855), 0.5% IgePal630 (Sigma, # I3021), and 5% donkey serum (Jackson Immuno, # 017-000-121) in PBS. Bone sections were stained with green fluorescent protein (GFP), c-Kit, and laminin antibodies for 3 days, washed overnight in several changes of PBS then incubated for 3 days in a staining solution containing secondary antibodies. Antibody dilutions are listed in Supplementary  Table. 7. The stained bone sections were dehydrated in a methanol series and cleared in benzyl alcohol: benzyl benzoate 1:2 mix (BABB clearing). 3D microscopy of the bone marrow was performed on a Confocal/Multiphoton Zeiss Leica SP8 resonant scanning confocal microscope with two-photon excitation. Confocal tiled Z-stack images were rendered in 3D and analyzed using Imaris v. 7.7.1 software.
Flow cytometry sorting of bone marrow endothelial cells. Bone marrow endothelial cells (ECs) were stained by intravenous injection of 10 μg anti-VEcadherin antibody (BD Biosciences, PE Rat Anti-Mouse CD144, #562243) 5-10 min before euthanizing the mice. Tibias and femurs were gently crushed using a mortar and pestle and then digested with DNase I (200 U/mL; ThermoFisher, # EN0521), Liberase DL (250 mg/mL; Sigma, # 5401160001), type IV collagenase (1 mg/mL; ThermoFisher, # 17104019), and collagenase D (500 μg/mL; Sigma, # 11088858001) with agitation for 30 min at 37°C. The cells were dissociated to a single-cell suspension by passing through a 25G needle several times and filtered with a 70-µm nylon mesh to generate a single-cell suspension. Cells were sorted in two successive rounds to ensure high purity using a FACS Aria II flow cytometer.
Phos-tag TM western blotting assay. For Phos-tag TM western blot, 8% acrylamide gel was mixed with 50 μmol/L Phos-tag™ acrylamide (Fujifilm, #AAL-107) and 100 μmol/L ZnCl 2 (Sigma, #229997). After sample separation on a Zn 2+ -Phos-tag SDS/PAGE gel, the gel was incubated with gentle agitation in transfer buffer supplemented with 1 mM EDTA, 3 times for 10 min each, followed by washing with transfer buffer without EDTA for another 10 min. Proteins were transferred to nitrocellulose membranes using a wet-tank method and analyzed by western blotting. Ptbp1, His, and β-actin signals were detected simultaneously using enhanced chemiluminescence (ECL) western blotting detection reagents (Thermo Fisher Scientific).
Histological analysis. Mouse tumor, femurs, spleen, and lymph node were dissected and cleaned, followed by fixation in 10% formalin and paraffin embedding. Femurs were decalcified in Decalcifying Solution-Lite (Sigma, #D0818). For morphological analyses, tissue sections (4-μm) were prepared and stained with hematoxylin and eosin (H&E). The histological features of the tissues were observed and images captured using a light microscope (Olympus, Tokyo, Japan).
Immunofluorescence analysis. Snap-frozen mouse tissues were embedded in Optimal Cutting Temperature compound (OCT) and 8-μm sections were prepared. For immunofluorescent staining, tissue sections were incubated with antibodies against Von Willebrand factor (vWF) or Lkb1 and DAPI Staining Solution, followed by incubation with secondary antibody. Antibody dilutions are listed in Supplementary Table. 7. All sections were imaged with identical exposure settings, and the images were quantified using Image-Pro Plus 6.0 software. Quantification of the positive signal in regions of interest was performed as described above.
Immunohistochemical staining. The immunohistochemical staining of paraffinembedded tissue sections was completed as described above. Following heatinduced antigen retrieval and incubation with blocking buffer (DAKO), tissue sections were incubated with antibodies against Lkb1 or CD61. Antibody dilutions are listed in Supplementary Table. 7. The tissues were then incubated sequentially with labeled horseradish peroxidase (Dako Real EnVision-HRP, Rabbit-Mouse, #K4063), 3,3' diaminobenzidine (DAB; Dako, #K4061), and hematoxylin (Sigma, St. Louis, MO). Positive and negative controls were included in each experiment. Tissue sections were mounted and visualized with an Olympus BX53 microscope. Images were captured and quantified with Image-Pro Plus 6.0 software, as described above.
Enzyme-linked immunosorbent assay (ELISA). Serum was obtained from peripheral blood after centrifugation. Scf levels in mouse serum or cell culture medium were determined using an ELISA kit (Abcam, #ab100740) according to the manufacturer's instructions.
Reverse transcription-polymerase chain reaction (RT-PCR) and quantitative PCR (qPCR). Total mRNA was isolated using TRIzol reagent (Invitrogen, #15596018) and reverse-transcribed using the cDNA synthesis kit (Promega, #A5001) according to the manufacturer's instructions. RT-PCR was performed using SYBR green on a C1000 Touch™ Thermal Cycler (Bio-Rad). PCR was performed using AccuPower® HotStart PCR PreMix from Bioneer on an S1000™ Thermal Cycler (Bio-Rad). Primers are listed in Supplementary Table 5. Sample RNA levels were normalized to that of Gapdh. Calculations were performed by the comparative method 2 −ΔΔCT .
RNA sequencing and data processing. Total RNA was extracted from endothelial cells isolated from 8 to 12-week-old mice using TRIzol (Invitrogen, USA), followed by sequencing at LC Science (Houston, TX). High-quality reads were aligned to Mus GRCm38 genome by HISAT2. The expression levels of each gene were normalized to fragments per kilobase million (FKPM) reads. Paired differential gene expression analyses were performed using String Tie by addition of fold change >2 and P-value < 0.05. GO or KEGG enrichment analyses of differential genes were performed using Ballgown R package.
Western blotting analysis. Cell and tissue extracts were separated by SDS-PAGE and transferred onto a nitrocellulose membrane. The membrane was then sequentially incubated with primary and secondary antibodies. Antibody dilutions are listed in Supplementary Table. 7. Protein signals were detected using enhanced chemiluminescence (ECL) western blotting detection reagents (Thermo Fisher Scientific).
Plasmid construction and transfection. Ptbp1 cDNA clone (pQE30, Ptbp1 iso-form1) was purchased from Addgene (#108591). Phosphorylation mutant T138A and phosphomimetic mutant T138E were generated using a QuikChange II sitedirected mutagenesis kit (Agilent, # 200523) according to the manufacturer's instructions. Primers used for point mutation are listed in Supplementary Table 6. All mutations were verified by DNA sequencing.
Statistical analyses. Data were analyzed with SPSS version 18.0 (IBM). All values are expressed as the mean ± standard error of the mean (SEM) unless otherwise stated. All experiments were performed at least three times unless otherwise stated.
When assessing the variance within two groups, the F-test was used. In the results describing more than two groups, the Brown-Forsythe test was applied to analyze the equal variance assumption. The Shapiro-Wilk test was performed to assess data normality. When analyzing the difference between the two groups, Student's t-test with a two-tailed distribution was performed when the assumptions (equal variance and normal distribution) were satisfied. The nonparametric Mann-Whitney U test was applied to analyze the difference between two groups when the assumptions of equal variance and normally distributed data were not met. P < 0.05 was considered statistically significant unless otherwise stated.