Mixed lineage kinase (MLK) controls tumor development and angiogenesis

Abstract

Cancer is among the leading causes of death in the USA and worldwide. Solid tumors require the formation of new blood vessels (angiogenesis) for their growth. The endothelium plays a crucial role in angiogenesis and tumor progression. Hypoxic stress generated by tumors can activate stress kinases such as mixed lineage kinases (MLKs). Publicly available datasets on lung adenocarcinoma, along with our experimental findings, indicate that MLK2 and MLK3 are expressed in human lung tumors. In this study, using three distinct mouse models of tumor development, we demonstrated that MLK2 (MAP3K10) and MLK3 (MAP3K11) are essential for tumor growth and angiogenesis. Furthermore, MLK2 and MLK3 are highly expressed in the endothelium and are necessary for endothelial proliferation, migration, and angiogenesis. In the endothelium, MLKs regulate the expression of angiogenic growth factors and metalloproteinases, including Pgf, Vegfa, Angptl4, Adam8, and Mmp9. Additionally, the MLK family of kinases acts through the long noncoding RNA (lncRNA) H19 to control the expression of these pro-angiogenic factors in the endothelium. Collectively, these findings suggest that the MLK-H19 axis coordinates endothelial function, angiogenesis, and tumor growth.

Introduction

More than 20 million new cancer cases are diagnosed worldwide, resulting in over 600,000 deaths in the United States alone and approximately 10 million deaths globally each year [1–3]. Tumor angiogenesis is pivotal for primary tumor growth and metastasis, as solid tumors cannot grow beyond a few millimeters without angiogenesis [4]. Hypoxic stress in solid tumors activates Hypoxia-Inducible Factor (HIF)-1 and other stress-induced pathways [5,6] that can induce endothelial cells to proliferate, migrate, and form new blood vessels. Endothelial responses to stress, including hypoxia, are dictated, in part, by the mitogen-activated protein kinase (MAPK) system, which comprises the extracellular signal-regulated kinase (ERK), p38 MAP kinase (p38MAPK), and c-Jun N-terminal kinase (JNK) [7]. JNK and p38 are well-studied signaling cascades that are essential for a wide range of vital cellular activities, including coordinating cellular responses to stress stimuli such as cytokines, inflammation, and reactive oxygen species (ROS) [8–10]. Therefore, inhibition of JNK or p38 is likely to be accompanied by severe side effects, making them poor drug targets.

More than fifteen MAP3K (Mitogen-Activated Protein Kinase Kinase Kinase) family members have been identified in mammals, including the mitogen-activated protein kinase kinase kinases (MEKK1–4), the apoptosis signal-regulating kinases (ASK1–2), the transforming growth factor-β-activated kinase 1 (TAK1), and the mixed-lineage kinases (MLK1–4) [11,12]. Among these upstream regulators of the MAP kinase cascade, members of the MAP3K tier, which includes the MLKs and others, have been shown to activate both the JNK and p38 pathways [11]. The diverse array of upstream activators is likely responsible for the specificity of stimulus-induced pathway activation. Our previous findings highlight MLK2/3 as selective regulators of JNK activation in response to free fatty acids (FFAs) [13,14]. Others have shown that ASK1 is the main activator of the JNK pathway during ROS activation [15]. These specificities suggest that targeting MLK2/3 could offer a more precise therapeutic approach while minimizing unintended perturbations of other JNK- and p38-mediated cellular responses.

The MLK protein family comprises serine/threonine kinases [11] that, under specific conditions, can activate all three main MAPK subfamilies (JNK, p38, and ERK) via MAP2Ks [16,17]. Structurally, the MLK proteins are characterized by an SRC homology 3 (SH3) domain at their N-terminus that precedes their kinase domain. Adjacent to the kinase domain is a leucine-zipper domain that is followed by a Cdc42/Rac interactive binding (CRIB) [16]. Full activation of MLK and its subsequent downstream targets requires the CRIB domain to bind with the GTPases of Cdc42/Rac, which leads to a conformational change to expose the ATP binding pocket of the kinase domain in MLK [16,18]. Within mammals, the MLK family comprises four family members, MLK1 through MLK4, corresponding to the genes MAP3K9, MAP3K10, MAP3K11, and MAP3K21 [11]. This family of kinases plays a role in insulin resistance, sepsis, and fatty acid signaling [13,14,18–20]. Notably, in humans, MLK3 has been implicated in the pathogenesis of metabolic and cardiovascular diseases through genome-wide association studies (GWAS) [21,22] and in vitro cancer cell migration [23]; however, the role of MLKs in post-natal and tumor angiogenesis remains unknown.

The long noncoding RNA (lncRNA) H19 is evolutionarily conserved among mammals and is notably expressed in fetal tissues and adult muscles [24,25]. H19 has been implicated in human genetic disorders and cancer [24,26,27]. It is known to be paternally imprinted [28] and upregulated by hypoxia [29,30]. It functions through various mechanisms, such as acting as a molecular ‘sponge’ for microRNA let-7 [31], modifying chromatin [32,33], and stabilizing RNA binding proteins [34]. In endothelial cells, H19 plays a critical role in limiting senescence and inflammation [35,36]. However, its role in regulating the expression of angiogenic growth factors, metalloproteinases, and tumor angiogenesis is not yet fully understood.

In this study, we sought to examine the role of stress-responsive MLKs in tumor development and angiogenesis. Animals lacking both MLK2 and MLK3 (Mlk2−/−Mlk3−/−), as well as our newly generated inducible MLK2 knockout in the endothelium (iECKO^Mlk2), showed reduced tumor growth and vascular density. This impaired tumor angiogenesis ultimately led to significantly smaller tumors in both the single and double knockout mice of the MLK family. We discovered that the MLK2 and MLK3 family members exert profound influences on normal endothelial function and the maintenance of vascular homeostasis. Moreover, the absence of MLKs in the endothelium was associated with decreased expression of angiogenic growth factors and metalloproteinase genes, promoting a maladaptive phenotype. Furthermore, MLK regulation of angiogenic growth factors and metalloproteinase gene expression in the endothelium occurs via long non-coding RNA H19. Collectively, these data suggest that MLK2/3 could serve as potential therapeutic targets for tumor angiogenesis.

Results

MLK is expressed in human lung tumors and partially overlaps with the endothelium.

To investigate the role of the MLK family in lung cancer, we first examined the expression of two ubiquitously expressed family members, MLK2 and MLK3 (MAP3K10 and 11), using the previously published RNA-sequencing data from human patients with lung adenocarcinoma [37]. We found that both MLK2 and MLK3 are expressed in lung adenocarcinoma tumors from these patients (Fig. 1A). To further investigate MLK expression in human lung tumors, we obtained human lung adenocarcinoma samples and first assessed their architecture via hematoxylin and eosin (H&E) staining (Fig. 1B). We then evaluated MLK2 and MLK3 protein expression by immunofluorescent staining in these human lung adenocarcinoma samples. Consistent with the mRNA data, our results showed that MLK2 and MLK3 proteins are also present in these lung adenocarcinoma tumor samples (Fig. 1C). Next, we performed a co-staining of endothelial marker CD31 (PECAM1) alongside MLK2 or MLK3 in human lung adenocarcinoma samples and observed a partial colocalization of CD31 with MLK2 and MLK3 proteins (indicated by white arrows; Fig. 1D). These findings suggest that both MLK2 and MLK3 are expressed in both endothelial and non-endothelial cells within human lung adenocarcinoma tissues.

Fig. 1: Loss of MLK suppresses tumor growth and tumor angiogenesis.

(A) (A) MLK2 and MLK3 (MAP3K10 and MAP3K11) gene expression was examined by RNA sequencing in lung adenocarcinoma tumors (n=77–87). (B) Human normal lung or lung adenocarcinoma tumors were stained for H&E staining. (C-D) Human lung adenocarcinoma tumors were either stained for MLK2 and MLK3 proteins (Green; C) or costained for MLK2 and MLK3 proteins (Green) with endothelial marker CD31 (Red; D; Scale bar 20 μm). (E-G) Representative images of LLC1 tumors in mice (E), quantification of tumor length (F) and tumor volume (G) after 12 days of subcutaneous implantation (n=5–8). (H) H&E staining was performed on LLC1-induced tumors from both wild-type (WT) and MLK2/3 knockout mice. (I-J) Angiogenesis was assessed in tumors generated by LLC1 subcutaneous injection in WT and MLK2/3 knockout mice using the endothelial marker CD31, either conjugated with alkaline phosphatase (I) or via immunostaining (J; Scale bar 50 μm). (K-L) A representative picture of a subcutaneous LLC1 tumor model in WT mice transplanted with either WT or Mlk2−/− Mlk3−/− bone marrow or Mlk2−/− Mlk3−/− mice transplanted with WT bone marrow (K; n=7/grp) or quantification of tumor volume (L).

MLK is required for promoting tumor growth and angiogenesis in mouse models.

Next, to investigate the role of MLK in vivo in tumor development and angiogenesis, we utilized three different mouse tumor models in WT and MLK2 and MLK3 double knockout (Mlk2−/−Mlk3−/−) mice. In the first xenograft model of tumor formation, Lewis Lung Carcinoma 1 (LLC1) cells were subcutaneously injected into mice to induce subcutaneous tumor formation. We found that after 12 days of subcutaneous injection of LLC1 cells, MLK-deficient mice developed significantly smaller tumors in length and volume compared to WT control mice, exhibiting an approximately 60% reduction (Fig. 1E–G). We collected tumors from WT and Mlk2−/−Mlk3−/− mice to examine the tumor morphology via H&E staining (Fig. 1H). Additionally, tumor samples from WT and Mlk2−/−Mlk3−/− mice were stained for endothelial marker CD31 via alkaline phosphatase-conjugated and immunostaining methods to investigate the role of MLK family during angiogenesis. We found a significant reduction in CD31 staining in Mlk2−/−Mlk3−/− mice, indicating that tumors from MLK-deficient mice exhibited reduced vascularity compared to those from WT controls (Fig. 1I–J). MLK family members are expressed in a variety of cell and tissue types, including endothelial cells, neurons in the brain, liver, fat, and immune cells, among others (Fig. 1D [11]). Therefore, we performed bone marrow transplantation studies to distinguish the roles of MLK2 and MLK3 proteins in vascular cells versus immune cell compartments. Transplanting WT or Mlk2−/−Mlk3−/− bone marrow into WT mice did not impair LLC1 tumor growth in the subcutaneous xenograft tumor model (Fig. 1K, L). However, even with WT bone marrow transplantation, Mlk2−/−Mlk3−/− mice exhibited significantly smaller tumor formation (Fig. 1K–L). This observation suggested that the presence of vascular MLK2 and MLK3 is required for tumor development in a manner independent of hematopoietic cells.

Next, to determine the role of the MLK family in lung tumors, we intravenously injected LLC1 cells via the tail vein to establish lung tumors in WT and Mlk2−/−Mlk3−/− mice (Fig. 2A). Consistent with findings from the xenograft model results, MLK2/3-deficient mice developed fewer tumors (Fig. 2B), and the tumors that did form were significantly smaller in both length and area (Fig. 2C–D) within the lung tissue compared to WT controls, suggesting that MLK plays an important role in tumor formation and growth in the lung tissue. In WT mice, as observed using H&E staining, the tumors spread throughout the whole lung, and the tumor volume increased compared to that in Mlk2−/−Mlk3−/− mice, where the lung structure was still intact (Fig. 2E). Moreover, the production of a proangiogenic growth factor called placental growth factor (PGF or PLGF) was impaired in the tumors of Mlk2−/− Mlk3−/− mice compared to tumors developed in WT mice (Fig. 2F). Macrophage infiltration within tumors has been associated with the promotion of tumor angiogenesis, as tumor-associated macrophages secrete pro-angiogenic factors and MMPs which contribute to endothelial cell activation and neovascularization [38]. Therefore, we used F4/80 staining to determine the macrophage infiltration in tumors in WT and Mlk2−/−Mlk3−/− mice. We found that macrophage infiltration in tumors of WT mice was higher compared to tumors from MLK2/3 knockout mice (Fig. 2G). These data showed that MLK plays a critical role in tumor initiation and progression in lung tissue.

Fig. 2: MLK is required for tumor growth in lung and liver.

(A) A representative picture of an LLC1 tumor in the lung of WT or Mlk2−/− Mlk3−/− mice following tail-vein injection of LLC1 cells. (B-F) Quantification of LLC1 tumor number (B), tumor length (C), tumor area (D), H&E staining (E), and immunoblot analysis (F) of lung tumors from WT or Mlk2−/− Mlk3−/− mice after 12 days of LLC1 tail vain injection (n=7–10). (G) A representative image of F4/80 staining of lung tumor from WT or Mlk2−/− Mlk3−/− mice (H) A representative image of hepatocellular carcinoma (HCC) tumor growth in WT or Mlk2−/− Mlk3−/− mice nine months after injection of carcinogen diethyl nitrosamine (DEN). (I-M) Quantification of HCC tumor number (I), tumor length (J), tumor area (K), immunoblot analysis (L), and H&E staining (M) of liver tumors in WT or Mlk2−/− Mlk3−/− mice nine months after DEN injection (n=7).

To further investigate the role of MLK in liver tumors, we used another mouse model for tumor development where we utilized a diethylnitrosamine (DEN)-induced hepatocellular carcinoma (HCC), in which tumor development in part depends on tumor angiogenesis [39]. DEN was administered intraperitoneal in both WT and MLK2/3-deficient mice at three weeks of age and formation of the hepatocellular carcinoma in the liver was assessed after a nine-month period. In the HCC model, Mlk2−/−Mlk3−/− mice exhibited both a reduced tumor number (indicated by white arrows, Fig. 2H–I) and size (Fig. 2J–K) compared to WT control mice. Moreover, HCC tumors from Mlk2−/−Mlk3−/− mice produced less PGF protein than WT mice tumors (Fig. 2L). We also performed H&E staining of tumors from both WT and MLK2/3 knockout to analyze the morphology of the tumor. We found that in WT mice, tumors are spread in the larger area, whereas in MLK2/3 knockout, they were small, and large portions of normal liver tissue were intact (Fig. 2M). The above data further indicates that MLK proteins are required for tumor initiation and progression in the mice.

MLK is required for endothelial cell proliferation and migration.

Our in vivo data in the mouse tumor model showed that MLK protein plays a critical role in tumor angiogenesis (Fig. 1I–J). The endothelial cells play an important role during tumor angiogenesis. Therefore, to examine the role of the MLK family of kinases in endothelial cell function, we first evaluated the expression of all four family members of MLK, MLK1–4 (MAP3K9, 10, 11, and 21), in human aortic endothelial cells and mouse lung endothelial cells (HAECs and MLECs). We found that in both humans and mice, endothelium predominantly expressed MLK2 and MLK3 among all four family members (Fig. 3A and Supplementary Fig. 1A). Solid tumors develop hypoxic conditions that can activate stress kinase-like MLK family members, we examined MLK3 activation in the endothelium and found that MLK3 phosphorylation was increased after hypoxia treatment in human endothelium. (Supplementary Fig. 1B; phosphor-MLK2 antibody is not available). Endothelial cell proliferation and migration are critical hallmarks of angiogenesis and endothelial cell function. Therefore, to determine the role of MLK2 and MLK3 in endothelial cell migration and proliferation, we used primary MLECs isolated from either WT and Mlk2−/−Mlk3−/− mice. To our surprise, unlike primary fibroblasts, which we have published previously [18], MLK2/3-deficient endothelial cells have a reduced ability to proliferate compared to WT cells (~40%; Fig. 3B). Moreover, compared with WT endothelial cells, MLK-deficient endothelial cells exhibited a reduced migratory capacity (Fig. 3C–D). These data suggest that the MLK family plays an important role in endothelial cell proliferation and migration.

Fig. 3. MLK is required for pro-angiogenic factors and metalloproteinase gene expression in endothelium.

(A) mRNA expression of MLK1 (MAP3K9), MLK2 (MAP3K10), MLK3 (MAP3K11), and MLK4 (MAP3K21) in HAECs (n=3). (B-D) Proliferation assay (B; n=12), scratch wound assay (C) and its quantification (D) in MLECs isolated from either WT or Mlk2−/− Mlk3−/− mice (E-F) Heat map of differentially expressed genes (E) and KEGG analysis for differential expressed pathways (F), observed by RNA sequencing, between WT or Mlk2−/− Mlk3−/− mouse endothelium (n=3). (G-H) mRNA levels (G) and immunoblot analysis (H) differentially expressing angiogenesis or metalloproteinase-related genes in MLECs isolated from WT or Mlk2−/− Mlk3−/− mice after 90 minutes (for mRNA expression) and 24 hours (for protein expression) post-hypoxia treatment respectively (n=6). (I-J) Sprouting assay on aorta rings (I) and its quantification (J) isolated either from WT or Mlk2−/− Mlk3−/− mice (n=6). (K-L) WT or Mlk2−/−Mlk3−/− MLECs were transfected either with control, MLK2 (K), or MLK3 (L) vector and assessed for gene expression by RT-qPCR (n=4).

MLK is required for autocrine endothelial expression of proangiogenic growth factors and metalloproteinases: PGF, VEGF-A, ANGPTL4, ADAM8, and MMP9.

To determine the mechanism by which MLK plays an important role in endothelial cell function, we performed unbiased RNA sequencing (RNA-seq) of primary endothelial cells (MLECs) exposed to either normoxia or hypoxia, isolated from wild-type (WT) and compound Mlk2−/−Mlk3−/− mice (Fig. 3E). Our bioinformatics analysis revealed that the pathways most strongly associated with differentially expressed genes between wild-type and MLK2/3-deficient endothelial cells were the “Pathways in cancer” and “HIF1α signaling pathway” (Fig. 3E–F). Both pathways are known to regulate growth factors and metalloproteinases, which are required for endothelial growth and survival, including vascular endothelial growth factor A (Vegfa), angiopoietin-like 4 (Angptl4), and matrix metalloproteinase 9 (Mmp9) [40–42]. Therefore, we performed RT-qPCR for growth factors and metalloproteinases in primary endothelial cells isolated from WT and MLK2/3 knockout mice. Our RT-qPCR results revealed that the mRNA expression of proangiogenic factors and metalloproteinases, such as placental growth factor (Pgf or Plgf), vascular endothelial growth factor A (Vegfa), angiopoietin-like 4 (Angptl4), A disintegrin and metalloproteinase domain-containing protein 8 (Adam8), and matrix metalloproteinase 9 (Mmp9), were decreased in MLK2/3-deficient MLECs compared to WT during hypoxic conditions (Fig. 3G). Additionally, using western blot analysis, we observed that, similar to mRNA data, the protein levels of PGF, ADAM8, and MMP9 were significantly lower in the MLK2/3 deficient endothelium than in the WT endothelium (Fig. 3H). The downregulation of the Pgf, Vegfa, Angptl4, Adam8, and Mmp9 genes has been implicated in reduced endothelial migration and angiogenesis [43–51]. Therefore, these data suggest that a lack of endothelial MLK2/3 may impair endothelial angiogenesis responses.

Pro-angiogenic factors such as PGF and VEGF-A can promote angiogenesis. Therefore, we investigated the role of endothelial MLK family members in an ex vivo angiogenesis model using a vessel sprouting assay with aortic ring segments isolated from either wild-type or Mlk2−/− Mlk3−/− mice. We found that, compared to wild-type mice, the MLK-deficient aortic rings exhibited significantly reduced sprouting angiogenesis in Matrigel (Fig. 3I–J).

MLK2 and MLK3 can have overlapping functions [18]. To determine the individual contributions of MLK family members to proangiogenic gene expression, we conducted a rescue experiment by introducing either MLK2 or MLK3 expression vectors into Mlk2−/− Mlk3−/− MLECs. We observed that the re-expression of either Mlk2 or Mlk3 in Mlk2−/− Mlk3−/− MLECs was sufficient to restore the mRNA expression levels of Pgf and Angptl4 (Fig. 3K–L). These gene expression data suggest that MLK proteins are essential for maintaining the expression of proangiogenic factors and metalloproteinase genes in endothelial cells, which is required for angiogenesis.

MLK-controlled lncRNA H19 expression is required for endothelial gene expression.

Examination of our RNA-seq dataset (Fig. 3E) for differentially expressed RNAs in Mlk2−/−Mlk3−/− vs. wild-type endothelium indicated that the lncRNA H19 expression was significantly downregulated in Mlk2−/−Mlk3−/− mice compared to wild-type mice (Fig. 4A). Recent studies have implicated the role of lncRNA H19 in cancer development [24,26,27] and our data showed that MLK2/3 plays a role in regulating pro-angiogenic factors in endothelium (Fig. 3). Therefore, we investigated if lncRNA H19 play any role in the regulation of in endothelial cell function and angiogenesis. We found that lncRNA H19 suppression in human endothelial cells (HAECs) led to decreased protein expression of PGF, ADAM8, and MMP9 (Fig. 4B), as well as a significant reduction in PGF and ANGPTL4 mRNA expression (Fig. 4C) under hypoxic conditions in endothelial cells. Next, we examined the role of lncRNA H19 in endothelial proliferation. We used BrdU assay in human endothelial cells either treated with control or H19 siRNA for 72 hours. We found that the proliferation of endothelial cells was significantly reduced in H19 siRNA-treated cells compared to control-treated cells (Fig. 4D). These findings demonstrate that lncRNA H19 is required for endothelial cell proliferation and gene expression.

Fig. 4. LncRNA H19 is needed for angiogenesis-related endothelial gene expression.

(A) LncRNA H19 expression by RNA-seq as mean homology fragments/kilobase/million reads in MLECs (n=3). (B-C) Impact of H19 inhibition on protein expression (B) and mRNA expression (C) of pro-angiogenesis and metalloproteinases genes in HAECs treated either with control or H19 siRNA for 72 hours (n=6). (D) Human aortic endothelial cells (HAECs) were either treated with control or H19 siRNA, and a proliferation assay was performed by using BrdU (n=5). (E) Mlk2−/−Mlk3−/− MLECs were transfected with either a control or H19 vector. mRNA was isolated, and RT-qPCR was performed to assess endothelial gene expression (Pgf1, Vegfa, Angptl4, Adam8, Mmp9, and H19) (n=6). (F) Mlk2−/−Mlk3−/− MLECs were transfected either with control or H19 vector and assessed for protein expression as indicated by immunoblot. (G) Immunostaining was used to evaluate CD31 (PECAM1) protein expression in MLK2/3 knockout MLECs transfected with either the control or H19 plasmid (scale bar: 20μm). (H) H19 overexpression vector or control vector was added to Mlk2 and Mlk3 siRNA treated HAECs, and scratch wound closure assay was performed. (I) Quantification of scratch wound closure assay for migration in Mlk2/3 siRNA treated HAECs transfected either with control or H19 overexpression vector (n=6).

Taking the converse approach, we performed a rescue experiment by introducing an H19 overexpression vector into Mlk2−/−Mlk3−/− MLECs. We found that H19 overexpression rescued the expression of the endothelial cell markers and angiogenesis genes in MLECs at the protein and mRNA levels (Fig. 4E–F). In MLK-deficient tumors, the CD31 stained cells were reduced (Fig. 1I–J); therefore, we stained for the endothelial marker proteins CD31 (PECAM1) in Mlk2−/−Mlk3−/− MLECs treated with either the control plasmid or the H19 expression vector. We found that MLECs overexpressing H19 had increased expression of CD31 compared to control (Fig. 4G). Next, we used a scratch wound healing assay to examine the role of H19 overexpression during endothelial cell migration in MLK2/3 deficient cells. We observed that H19 overexpression significantly increased the migration of MLK2/3 deficient endothelial cells during scratch wound healing assay compared to the control (Fig. 4H–I). These observations suggest that H19 plays an important role in endothelial proliferation and migration and that, at least in part, MLK controls downstream gene expression via H19.

MLK controls lncRNA H19 expression via the ERK pathway.

Next, to investigate how MLK regulates H19 expression, we focused on three canonical downstream targets of MLK: ERKs, JNKs, and p38 MAPK [11]. A recent publication has demonstrated that JNK does not influence postnatal endothelial function [52]. Consequently, we examined the role of MLK2/3 in activating two downstream targets, ERK and p38 MAPK. Because VEGF-A signaling is critical for endothelial function during angiogenesis [53,51,54], we treated wild-type (WT) and MLK2/3 knockout MLECs with VEGF-A for 30 minutes. Our immunoblot data indicate that MLK2/3 is required for ERK activation but not for p38 MAPK, as phosphor-ERK1/2 levels were reduced in MLK2/3 knockout cells, whereas p38 MAPK remained unchanged (Fig. 5A). These findings suggest that MLK2/3 primarily control ERK1/2 activation in endothelial cells. Next, we utilized inhibitors for ERK and p38 MAPK pathway in endothelial cells and discovered that, unlike p38, ERK regulates H19 expression in the human endothelial cells (Fig. 5B–C). Subsequently, treatment with ERK pathway inhibitor (MEK inhibitor, U0126) led to reduced mRNA expression of Pgf, Vegfa, Angptl4, Adam8, and Mmp9 genes in human endothelium (Fig. 5D). These observations showed that MLK controls H19 and downstream gene expression via ERK pathway.

Fig. 5. ERK controls H19 expression.

(A) An immunoblot was performed on MLECs isolated from either WT or MLK2/3 knockout mice, with or without a 30-minute VEGF-A treatment. (B, C) mRNA was isolated from HAECs (human aortic endothelial cells) and treated with either a control or an ERK pathway inhibitor (MEK inhibitor, U0126; B) or with a p38 MAPK inhibitor (SB203580; C). RT-qPCR was then performed to measure H19 lncRNA expression (n=4). (D) mRNA was isolated from human endothelial cells treated with either control or U0126. After 90 minutes of hypoxia treatment, RT-qPCR was performed to analyze various angiogenesis-related genes (n=3).

Endothelial-specific MLK2 deficiency mimics the whole body MLK2/3 phenotype in vitro and in vivo.

We found that among all members of the MLK family, MLK2 expression was highest in human endothelial cells (Fig. 3A). Therefore to determine the role(s) of individual MLK family members, such as MLK2, we generated an inducible endothelial cell model in which MLK2 was knocked out (Supplementary Fig. 3). Mlk2 conditional mice were generated by targeting exons 3 and 4 of the Mlk2 gene with flanking LoxP sites. These Mlk2^fl/fl mice were then bred with the tamoxifen-inducible Cdh5-CreERT2 driver line (iECKO^Mlk2; Supplementary Fig. 3; [55]). The conditional MLK2 knockout mice were injected with LLC1 cells subcutaneously. We observed that iECKO^Mlk2 mice exhibited a reduction in subcutaneous LLC1 tumor size compared with control (Mlk2^fl/fl) mice (Fig. 6A–C). Next, we immunostained LLC1 tumors from control and iECKO^Mlk2 mice for CD31 (an endothelial marker for angiogenesis) and observed a higher CD31 staining in tumors from control mice compared with those from iECKO^Mlk2 mice (Fig. 6D). We also observed morphological differences in H&E stained tumors from control and iECKO^Mlk2 mice (Fig. 6E). We then isolated MLECs from both control and iECKO^Mlk2 mice and performed a scratch wound healing assay. MLECs isolated from iECKO^Mlk2 mice exhibited reduced migration compared with control MLECs (Fig. 6F–G). To determine the role of MLK2 in human endothelial cell lines, we used control siRNA or Mlk2 siRNA. We found that MLK2 knockdown significantly reduced both proliferation (Fig. 6H) and migration (Fig. 6I–J) compared with control siRNA-treated cells. We also observed that MLK2 deficiency in mouse endothelium, achieved by infecting MLECs isolated from Mlk2^f/f mice with Cre virus, led to decreased expression of several angiogenic growth factors, and metalloproteinase genes, including Pgf, Angptl4, and Mmp9 compared to GFP-infected controls (Fig. 7A). Next, we performed RT-qPCR for other endothelial markers and found that CD31 (Pecam1), vWF, and Tie1 levels were lower in Cre-infected Mlk2^f/f MLECs than in those infected with GFP virus (Fig. 7B). Immunostaining confirmed that MLK2 deficiency decreased CD31 protein levels in endothelial cells (Fig. 7C). Thus, endothelial-specific ablation of Mlk2 mimicked the phenotype of Mlk2−/−Mlk3−/− mice, as indicated by reduced endothelial gene expression, proliferation, migration, and tumor formation.

Fig. 6. MLK2 in the endothelium is required for angiogenesis and tumor growth.

(A) A representative picture of subcutaneous LLC1 tumor development in control or endothelial-specific iECKO^Mlk2 mice after 10 days of LLC1 injection. (B-C) Quantification of the length (B) and volume (C) of subcutaneous LLC1 tumors in control or iECKO^Mlk2 mice (n=6). (D) Immunostaining of LLC1 tumors was done to evaluate CD31 (PECAM1) protein expression in control (Mlk2^fl/fl) and endothelial-specific iECKO^Mlk2 mice (scale bar 20 μm). (E) H&E staining of LLC1 tumors from control and MLK2 knockout mice (scale bar 20 μm). (F-G) A representative image of scratch wound healing assay (F) and its corresponding migration quantification (G) in MLECs isolated from Mlk2^fl/fl mice infected with either a GFP or Cre virus (n=5). (H) Human endothelial cells were treated with either control or Mlk2 siRNA, followed by a Brdu-based proliferation assay (n=5). (I-J) A representative image of the scratch wound healing assay (I) and its migration quantification (J) in human endothelial cells treated with either control or Mlk2 siRNA (n=4).

Fig. 7. MLK2 in the endothelium controls endothelial gene expression.

(A) Angiogenesis-related gene (Pgf, Vegfa, Angptl4, Adam8, and Mmp9) expression was assessed by RT-qPCR in MLECs isolated from Mlk2^fl/fl mice either treated with GFP or Cre virus (n=6). (B) The endothelial cell gene expression (Pecam1, vWF, and Tie1) was measured by RT-qPCR in MLECs isolated from Mlk2^fl/fl mice either treated with GFP or Cre virus (n=6). (C) Immunostaining for the endothelial marker CD31 (PECAM1) was performed on MLECs isolated from Mlk2^fl/fl mice infected with either a GFP or Cre virus (scale bar 20 μm). (D) A schematic model of the role of MLK and its downstream effector in endothelial function and tumor development.

Discussion

Endothelial cells play a crucial role in angiogenesis, a fundamental process in solid tumor progression that facilitates sustained tumor growth by ensuring a continuous supply of oxygen and nutrients [56–58]. Consequently, targeting tumor-associated angiogenesis has been extensively explored as a therapeutic strategy for cancer prevention and treatment. Several angiogenesis inhibitors, including bevacizumab, sunitinib, and sorafenib, have been developed [59,60], yet their clinical efficacy is often hindered by resistance mechanisms and limited responses. One possible explanation is that these inhibitors primarily target VEGF-A, whereas our data indicate that MLK-controlled PGF may also play a crucial role in angiogenesis and tumor formation. To improve therapeutic outcomes, a deeper understanding of the molecular pathways regulating tumor angiogenesis is essential.

Mixed-lineage kinases (MLKs), a subgroup of MAP3Ks, are known to activate the JNK and p38 signaling pathways, which mediate cellular responses to environmental stressors such as hypoxia, inflammation, and oxidative stress. The MLK family of proteins is most highly expressed in the neural compartment of mammals [11]. Earlier studies proposed that MLK plays a significant role in neurological diseases [16,11], and our previous work showed MLK involvement in insulin resistance in vivo [13]. More recently, in vitro data have suggested that MLK contributes to cancer cell proliferation and migration [11,61–63]. However, investigations into MLK’s role in angiogenesis in vivo remain limited.

In this study, we clearly demonstrate that vascular MLK plays a crucial role in tumor development by regulating tumor vascularity. We found that human lung tumors express both MLK2 and MLK3, with their expression partially overlapping with endothelial cells in human lung tumors. Using MLK2/3 knockout and endothelial-specific MLK2 knockout mice, we showed that the MLK family of proteins is essential for tumor growth in three different mouse models, including lung and liver tumors models. Tumor size was significantly reduced in the absence of MLK family members, suggesting their role in tumor progression.

Endothelial cell migration and proliferation are essential for tumor angiogenesis, enabling the formation and extension of new blood vessels toward the tumor. In response to proangiogenic signals, endothelial cells migrate toward hypoxic tumor regions and proliferate to expand the vascular network. Endothelial cells lacking MLK2 and MLK3 exhibit impaired proliferation and migration, as well as a marked reduction in the expression of proangiogenic factors and metalloproteinases, including Pgf, Vegfa, Angptl4, Adam8, and Mmp9. Additionally, our findings underscore the pivotal role of MLK signaling in angiogenesis ex vivo, as demonstrated by diminished sprouting in aortic rings from Mlk2−/−Mlk3−/− mice. Using MLK2/3 knockout and endothelial-specific MLK2 knockout mice, we further showed that the absence of MLK family members leads to a significant reduction in tumor vascularity and tumor growth.

Hematopoietic cells are attracted to tumor sites by chemotactic signals secreted by tumor and stromal cells, including VEGF-A, CXCL12, and CCL2. After mobilizing from the bone marrow to the tumor microenvironment, these cells contribute to angiogenesis by releasing proangiogenic factors such as matrix metalloproteinases (MMPs) and cytokines, thereby activating endothelial cells and promoting blood vessel formation [64–66]. For instance, macrophage infiltration within tumors creates a proangiogenic microenvironment, as tumor-associated macrophages secrete VEGF-A, FGF2, and MMPs, enhancing both angiogenesis and immune modulation [38].

Our data indicate that MLK2 and MLK3 deficiency reduces macrophage infiltration in tumors, underscoring the role of MLKs in establishing a proangiogenic microenvironment and sustaining tumor growth. Furthermore, bone marrow transplantation from wild-type (WT) mice into Mlk2−/−Mlk3−/− mice did not affect tumor formation, indicating that vascular MLK2 and MLK3 are required for tumor progression independently of hematopoietic cells. Overall, our findings demonstrate that the MLK family regulates tumor vascularity by controlling endothelial cell migration, proliferation, and the expression of proangiogenic factors. However, future studies are needed to determine whether the MLK family also regulates vascular permeability, expression of adhesion molecules, and chemotactic signaling, thereby influencing macrophage infiltration into tumors.

A critical question is how MLK modulates tumor growth and angiogenesis in various tumor models. Our RNA sequencing analysis indicates that MLK signaling regulates the expression of the long noncoding RNA (lncRNA) H19, which influences endothelial proliferation, migration, and the expression of proangiogenic factors and metalloproteinases. H19 has been associated with human genetic disorders and cancer [24,26]. It can control cellular functions through multiple mechanisms, including chromatin modification [32,33], stabilization of RNA-binding proteins [34], and acting as a molecular “sponge” for let-7 miRNA [31].

Previous studies suggest that H19 may play a role in blood vessel development and differentiation, particularly in smooth muscle cells [67,68]. Northern blot analysis has shown that H19 is highly expressed during the developmental stages of the rat aorta but declines in differentiated vascular tissue. Interestingly, H19 reappears following vascular injury in rat vascular smooth muscle cells [68]. However, its specific role in regulating endothelial cells during blood vessel formation remains poorly understood.

Our study is the first to establish the role of H19 in promoting endothelial cell proliferation, migration, and the expression of proangiogenic factors, thereby regulating angiogenesis. We demonstrate that H19 is a critical downstream target of MLK signaling and that the MLK-H19 axis plays a key role in tumor angiogenesis. Nevertheless, the underlying mechanisms by which this lncRNA regulates endothelial cell phenotype and drives proangiogenic gene expression remain largely unknown and warrant further investigation.

ERKs, JNKs, and p38 MAPK are three canonical targets of the MLK family. Mechanistically, we found that MLK signaling modulates H19 expression via the ERK pathway, whereas p38 MAPK plays a limited role in endothelial regulation. Furthermore, our data suggest that MLK2/3 are essential for ERK activation but do not affect p38 signaling in endothelial cells. We have also previously published that JNK plays an important role in collateral vessel formation in the endothelium. However, endothelial JNK plays no role in postnatal endothelial function, as evidenced by the fact that no changes were observed in proliferation, migration, sprouting angiogenesis, or tumor formation in mice lacking JNK family members [52]. Therefore, all the postnatal endothelial functions of MLK may be JNK and p38 independent and possibly regulated by the ERK pathway.

The ERK/MAPK signaling pathway has been shown to enhance VEGF transcription and suppress thrombospondin-1, thereby promoting tumor angiogenesis, growth, and metastasis. Additionally, ERK1/2 regulates IL-8 and VEGF expression, further supporting tumor vascularization. Notably, targeting the MEK/MAPK and PI3K/AKT pathways with inhibitors can suppress VEGF expression, reduce angiogenesis, and hinder tumor progression [69].

One possible explanation for why MLK regulates ERK activation but not p38 MAPK in endothelial cells is that different stimuli produce distinct activation profiles, and specificity arises from the formation of particular signaling complexes with various upstream and downstream kinases. More than fifteen MAP3K family members have been identified in mammals, and their diversity appears to dictate which downstream pathway is activated in response to specific signals. For example, free fatty acids primarily activate JNK via the MLKs, whereas reactive oxygen species (ROS) rely on ASKs. Meanwhile, TAK1 is a key mediator of JNK activation during cytokine signaling [70]. Therefore, in tumor-mediated hypoxia, MLK may signal through the ERK-H19 axis independently of the p38 and JNK pathways.

Additionally, the JNK and p38 signaling pathways are critical for many essential cellular processes, particularly in helping cells respond to stress signals such as cytokines, inflammation, and ROS [8–10]. Because JNK and p38 play such vital roles in these responses, their inhibition may lead to severe side effects, making them less suitable as drug targets. Conversely, MLKs exhibit stimulus-specific functions, positioning them as promising targets for future drug development.

Overall, our findings established the MLK-ERK-H19 axis as a key regulator of proangiogenic factor expression, endothelial function, angiogenesis, and tumor progression, providing new insights into its potential as a therapeutic target in tumor-associated angiogenesis (Fig. 7D). Although many tyrosine kinase inhibitors have been employed to curb tumor growth, targeting serine/threonine kinases in cancer therapy has lagged behind. In this study, using three different tumor models, we clearly demonstrate the critical role of MLK members in regulating angiogenic factors and promoting tumor development, suggesting that the MLK pathway could serve as a promising target for solid tumor therapies.

Methods

Animals

Our study used both male and female animals for in vivo and ex vivo experiments. C57BL/6J (#000664), Tie2-Cre (#008863), and VE-cadherin-Cre (#006137) mouse strains were obtained from Jackson Laboratory (Bar Harbor, ME) Cdh5-CreERT2+/− mice were a kind gift from Ralf Adam’s laboratory at the Max Planck Institute (Saarbrucken, Germany) [55]. Mice with Mlk2 and Mlk3 gene disruptions were generated as described previously [18]. Mice with an Mlk2 conditional allele were generated with the help of the International Knockout Mouse Consortium and The European Mouse Mutant Archive (Muchen, Germany). In brief, an Mlk2 conditional allele was created using a targeting vector with an Frt-Neo-Frt cassette within intron 2 of the Mlk2 gene and intronic LoxP sites flanking exons 3 and 4 (Supplementary Fig. 3). This vector was used to create chimeric mice via standard techniques, germline transmission was confirmed by RT-qPCR, and the Frt–Neo-Frt cassette was removed by crossing animals with Flp mice. The resulting Mlk2^fl/fl mice were bred to homogeneity (>10 generations) on the C57BL/6J background, and appropriate LoxP incorporation was demonstrated via RT‒qPCR (Supplementary Fig. 3). Furthermore, the endothelial Mlk2 gene was excised by crossing Mlk2^fl/fl mice with the Cdh5-CreERT2+/− driver line Field 53 to generate an inducible endothelial cell-specific knockout of MLK2 (iECKO^Mlk2; Supplementary Fig. 3) upon tamoxifen injection. Tamoxifen (T5648, Sigma, St. Louis, MO) was dissolved in ethanol and subsequently diluted in corn oil (C8267, Sigma, St. Louis, MO) to a concentration of 20 mg/ml for use, after which the mice were injected for 5 consecutive days. All mouse experiments were performed according to the relevant ethical regulations. Mice were housed in facilities accredited by the American Association for Laboratory Animal Care. Animal protocols were approved by the Institutional Animal Care and Use Committee of the Brigham and Women’s Hospital, Boston, MA

Tumor formation and tumor angiogenesis

Lewis lung carcinoma 1 cells (LLC1, ATCC, CRL-1642, Manassas, VA) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 4.5 g/L D-glucose, L-glutamine, 10% heat-inactivated fetal bovine serum and 1% penicillin/streptomycin (Life Technology Thermo, Waltham, MA). To assess the role of MLKs in tumor angiogenesis in vivo, LLC1 cells (2×10⁶ cells/mouse) were injected subcutaneously into the dorsal flanks of Mlk2−/−Mlk3−/− mice and wild-type (WT) controls. Mice were monitored for weight, well-being, tumor formation, and tumor diameter (by calipers) on days 7, 10, 12, and 14. The tumors were allowed to grow for 10–14 days before they reached 1–2 cm. The skin of the dorsal flank was then removed, and the tissue around the tumor was gently removed for further analysis.

Additionally, LLC1 cells (1×10⁶ cells/mouse) were also injected intravenously via the tail vein to establish lung tumors in WT and Mlk2−/− Mlk3−/− mice. After 12 days, mouse lungs were excised, and lung tissues were examined for solid masses. Tumor vascularity and architecture were examined by CD31/DAPI immunofluorescence.

We used a caliper to measure the tumor length, width, and height of all subcutaneous LLC1 tumors in the mice and used the standard formula 0.5 × length × width × height for volume calculation. For Lung and liver tumors, we used histological sections of every 100–200 μm number, and the area was quantified. All tumor experiments were repeated 3–6 times.

Bone marrow transplant

Donor WT and Mlk2−/−Mlk3−/− mice were humanely euthanized, and the femurs and tibias were collected. The bones were flushed with sterile phosphate-buffered saline (PBS) using a 25-gauge needle. The extracted bone marrow was dissociated mechanically, filtered through 70 μm gauze, and pelleted by centrifugation at 500 × g for 10 minutes. Pooled bone marrow cells were resuspended in PBS and kept on ice before they were injected into irradiated recipient mice.

Recipient mice were placed on antibiotics (Sulfamethoxazole, 220 mg/kg/day and Trimethoprim 40 mg/kg/day) water 2 days prior to irradiation was performed and remained on antibiotics for 12–14 days post-irradiation. On day 0, 12-week-old recipient mice were irradiated at the UMASS Chan Medical School gamma irradiation facility two times at 550 Rad (5.5Gy) + 550 Rad (5.5Gy) doses with a 4-hour interval. Twenty-four hours postirradiation, the mice were injected via the tail vein at a concentration of 5×10⁶ cells/injection in RPMI for Mlk2−/−Mlk3−/− donor mice and 7×10⁶ cells/injection in RPMI for WT donor mice. Five weeks postirradiation, the mice received a dorsal flank injection of LLC1 cells at a concentration of 1×10⁶ cells/mouse in PBS. Tumors were harvested at 12–14 days before they reached 1–2 cm in length.

Sprouting angiogenesis

Descending aortae were isolated from wild-type or Mlk2−/−Mlk3−/− mice, and 1 mm thick aortic rings were prepared using a #12 surgical scalpel. Ex vivo sprouting angiogenesis was performed using Matrigel (Cultrex, R&D Systems, #3536–005-02, Minneapolis, MN), and aortic rings were cultured in a humidified tissue culture incubator (37°C/5% CO₂). Matrigel-aorta ring sandwiches were prepared: (1) 100 μL of cold liquid Matrigel was added to a 48-well plate and incubated at 37°C for 20 minutes to solidify; (2) the aorta rings were transferred to Matrigel (one ring/well) and allowed to settle for 20 minutes at 37°C; and (3) 100 μL of cold liquid Matrigel was added to the top of the aorta ring and incubated at 37°C for 20 minutes to solidify. After the Matrigel-aorta rings were embedded, 150 μL of cell culture growth medium was added to the culture plate, and endothelial cell sprouting was monitored for 7 consecutive days. The growth medium was replaced every 24 hours without disturbing the Matrigel. Aortic sprouts were observed on a 4x phase contrast inverted microscope (Eclipse Ti, Nikon, Melville, NY), and images were acquired on a digital camera (Axiocam 305 mono, Carl Zeiss, White Plains, NY).

Human tumor

Lung adenocarcinoma tumors were collected by Dr. Hassan Khalil’s laboratory and fixed in 10% formalin, embedded in paraffin, and sectioned at 7 μm. Tumor vascularity and architecture were examined using either H&E staining or MLK2/MLK3/CD31/DAPI immunofluorescence staining.

To examine the expression of MLK2 and MLK3 (MAP3K10 and 11) in human lung adenocarcinomas, we analyzed previously published RNA-sequencing datasets from human patients with lung adenocarcinoma. Gene expression values for 87 lung adenocarcinomas and 77 adjacent normal tissues were obtained from the study by Seo JS et al. (2012), which was submitted to the NCBI Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE40419 [37].

Cell culture

Human aortic endothelial cells (HAECs; ATCC #PCS-100–011, Manassas, VA) were cultured in endothelial cell growth basal medium-2 supplemented with bullet kit growth factor (EBM-2 [endothelial cell growth basal medium-2], Lonza, Walkersville, MD), 5% fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin. All experiments were performed from cell passages from 3 to 6, and the cells were cultured at 37°C in 5% CO₂ balanced air.

To isolate mouse lung endothelial cells (MLECs), adult mice were sacrificed, and lung tissues were excised. The minced tissues were then transferred to a digestion solution [3 mg/ml Collagenase I (Worthington, 55A15492, Columbus, OH), 15 μg/ml DNase I (Sigma, D4527), endothelial mitogen (ECG, 1:10; Sigma B819-GS), and heparin (1:500; Sigma, H3149, St. Louis, MO in DMEM supplemented with 4.5 g/L D-glucose and L-glutamine] and incubated for 50 minutes at 37°C with gentle shaking every 10 minutes. The enzyme-digested tissue was filtered through a 70 μm cell strainer and rinsed by adding an additional 5 mL of ice-cold antibody selection solution (DMEM with 4.5 g/L D-glucose, L-glutamine, and 10% FBS), after which the cell suspension was centrifuged for 5 minutes at 1500 rpm at 4°C.

The negative selection was performed using 15 μL of anti-mouse CD45 antibody (BD Bioscience, 5520539, Franklin Lakes, NJ)-coated Dynabeads with sheep anti-rat IgG and 15 μL of anti-mouse CD326 (BD Bioscience, 552379) antibody-coated Dynabeads (Thermo Fisher, 11035, Waltham, MA) in 750 ul antibody selection solution. The cell-Dynabead suspension was incubated for 30 minutes at 4°C on a rotator, and the supernatant was harvested on a magnetic rack. For the primary selection of MLECs, 25 μL of anti-mouse CD31 (BD Bioscience, 553370) antibody-coated Dynabeads was added to the collected supernatant and incubated for 20 minutes at 4°C on a rotator, after which a magnetic separator was used to obtain the cell pellet. The cell pellets were subsequently resuspended in MLEC growth media supplemented with a 1:1 mixture of DMEM (Life Technology Thermo, 11885–084;) and Ham’s F-12 (Life Technology Thermo, 11765–054, Waltham, MA), 20% fetal bovine serum, 50 mg of endothelial mitogen, 50 mg/mL heparin, 100 units/mL penicillin, and 100 μg/mL streptomycin and plated on 0.2% gelatin (Sigma, G1392, St. Louis, MO)-coated cell flasks. The media was changed daily, and once the cells reached 80–90% confluence, the cells were collected in a conical tube with 20 μL of Dynabeads sheep anti-rat IgG coated with anti-mouse ICAM-2 (BD Biosciences, 553326) for the second selection. The cell-Dynabead suspension was rotated at 4°C for 20 minutes and harvested on a magnetic rack. The beads were resuspended in MLEC growth medium and plated into a new 0.2% gelatin-coated cell flask. The MLECs used for the experiments were between passages 2 and 4 [71].

Cell transfections

Transfection assays were performed using 100nM small interfering RNA oligonucleotides ON-TARGET plus SMART as a control (D-001810-10), Mlk2 (L-003576-00-0020), Mlk3 (L-003577-00-0020), and H19 (R-032256-00-005) (Horizon Discovery Ltd., Lafayette, CO) in DharmaFECT 3 reagent for 8–16 hours in Opti-MEM (Thermo, 31985070, Waltham, MA) [14]. For H19 overexpression in MLECs, a vector (pcDNa3.1(+)_A009) containing the synthetic H19 sequence was obtained from Dr. Alexander K. Kiemer laboratory at Saarland University, Germany [72]. WT and Mlk2−/− Mlk3−/− MLECs were transfected with 10 ng of either empty pcDNa3.1 vector (Addgene, 128047; Watertown, MA) or H19 vector overnight at 37°C in DharmaFECT 3 reagent in Opti-MEM. The media was then changed to the respective culture media, and after 48–72 hours (for RNA analysis) and 72–96 hours (for protein analysis) of siRNA/overexpression treatment, the cells were exposed to hypoxia.

In vitro Cre recombinase treatment to floxed MLECs

MLECs isolated from Mlk2^fl/fl mice were treated with either green fluorescent protein (GFP; control; 1×10⁵ pfu/mL) or expressing GFP-Cre recombinase (1×10⁵ pfu/mL) to delete the floxed alleles overnight at 37°C in MLECs media lacking FBS.

Hypoxia

The monolayers of WT and Mlk2−/−Mlk3−/− primary endothelial MLECs or siRNA-treated HAECs were exposed to normoxia (21% O₂) or hypoxia (1% O₂, Airgas, Radnor, PA) for 90 minutes for mRNA and either 30 minutes (for phosphor-protein) or 24 hours for protein isolation in a hypoxia chamber (Embrient, Inc., MIC-101, San Diego, CA). After exposure, the cells were lysed, and the expression of the genes was examined by RT-qPCR or immunoblotting.

Cell proliferation and migration

Primary mouse lung endothelial cells (MLECs) from wild-type (WT) control and Mlk2−/−Mlk3−/− mice were isolated, or human aortic endothelial cells (HAECs) were treated with control siRNA, siRNA against MLK2 and MLK3 or H19 for 72 hours. To determine proliferation, equal numbers of cells were plated, and an MTT assay (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide, Promega, G3580, Madison, WI) was performed according to the manufacturer’s protocol. In HAECs, proliferation was also determined using the BrdU Cell Proliferation Assay Kit (Cell Signaling, 6813, Danvers, MA) according to the manufacturer’s recommendation. Briefly, HAECs were treated with siRNA for 24 hours, after which they were plated at a density of 3,000 cells per well in 96-well plates and incubated overnight. The following day, cells were serum-starved for 6 hours before adding BrdU at a 1:500 dilution for 48 hours. After fixation, cells were incubated with 100 μL per well of anti-BrdU antibody working solution for 60 minutes at room temperature, followed by incubation with 100 μL per well of HRP-conjugated secondary antibody for 30 minutes at room temperature. After a 5-minute incubation with the substrate solution, proliferation was quantified by measuring absorbance at 450 nm (SpectraMax iD5 multimode microplate reader, Molecular Devices, San Jose, CA).

In MLECs, we performed an in vitro scratch wound closure assay to assess migration in both genotypes, as previously described [73]. Briefly, a vertical wound was created on the cell monolayers using a 200 μl pipette tip. After generation, the wound was inspected at 0 h, 4 h, 8 h, and 24 h for wound closure using phase contrast microscopy microscopy and ZEN microscopy software (Carl Zeiss Axiovert, White Plains, NY).

In HAECs, for migration assay, following siRNA treatment, endothelial cells were seeded in 0.2% gelatin (Sigma, G1393, St. Louis, MO)-coated 35 mm cell culture dishes containing a 4-well silicone insert (with 4 defined cell-free gaps; iBidi, 80466; Fitchburg, WI). After 24 hours, the silicon inserts were removed to create four 500 μm cell-free gaps and a 1 mm center gap in the middle, allowing us to observe the migration of four technical replicates. Images were captured by using a phase contrast microscope (Carl Zeiss Axiovert, White Plains, NY) to calculate the migration area. The migration area was calculated using the following mathematical equation: migration area (%) = (A0 - An)/A0 × 100, where A0 represents the area of the initial cell-free area (0 h) and An represents the remaining cell-free area at the measurement point. The cell-free area was measured at 5 different locations/areas, and an average of five measurements/area were then averaged together for four technical replicates to calculate the initial and final areas for each time point.

ERK and p38 inhibitor treatment

Human aortic endothelial cells (HAECs) were FBS-starved overnight before treatment with either vehicle (DMSO), 10 μM ERK pathway inhibitor (MEK inhibitor, U0126, Promega V1121, Madison, WI), or 20 μM p38 MAPK inhibitor (SB203580. Tocris Bioscience, #1202, Bio-Techne, Minneapolis, MN; Supplementary Table 4) for 30 minutes before being transferred to a hypoxia chamber for 90 minutes, making the total treatment duration 2 hours. After treatment, mRNA was isolated from HAECs, and RT-qPCR was performed to analyze H19 and angiogenesis-related gene expression.

VEGF-A treatment

Mouse lung endothelial cells (MLECs) were serum-starved overnight before treatment with either control PBS or VEGF-A (20–100 μg/ml, R&D Systems, 293-VE-050, Minneapolis, MN; Supplementary Table 4) for 30 minutes. After treatment, proteins were isolated, and immunoblots were performed to analyze phospho-protein levels.

RNA preparation and quantitative real-time polymerase chain reaction

Total RNA was extracted from tissues using TRIzol reagent (Thermo Fisher, 15596026, Waltham, MA) and from cells using the RNeasy Plus Mini Kit (Qiagen, 74136, distributed by Thermo Fisher, Waltham, MA). One microgram of total RNA was reverse-transcribed with oligo(dT) primers using iScript Reverse Transcription Supermix (Bio-Rad, 4453320, Hercules, CA). The resulting cDNA was diluted 1:5, and mRNA expression was examined by quantitative PCR on a QuantStudio^™ 6 Flex Real-Time PCR System (Thermo Fisher, 4485692, Waltham, MA). TaqMan probes were was used to quantify Vegfa (Mm01281449_m1, Hs00900055_m1), Pgf (Mm00435613_m1, Hs00182176_m1), Adam8 (Mm00545745_m1, Hs00923284_g1), Mmp9 (Mm0044299_m1, Hs00957555_m1), Angptl4 (Mm00480431_m1, Hs01101127_m1), Pecam1 (Mm01242584_m1), vWF (Mm00550376_m1), Tie1 (Mm00441786_m1), H19 (Mm01156721_g1, Hs00262142_g1) Hprt (Mm03024075_m1, Hs02800695_m1), Gapdh (4352339E-0904021, Hs99999905_m1), B2M (Mm00437762_m1, Hs00187842_m1), and Tbp (Mm00446973_m1, Hs00427620_m1).

The relative expression was calculated using the 2^−ΔΔCT method described previously [74,75], in which the expression was normalized to that of the reference genes Hprt, Gapdh, B2m, and Tbp.

Immunoblot analysis

Cell extracts were prepared using Triton lysis buffer [20 mM Tris (pH 7.4), 1% Triton X-100, 10% glycerol, 137 mM NaCl, 2 mM EDTA, 25 mM b-glycerophosphate, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, and protease inhibitor cocktail (complete ULTRA Tablets, Sigma, 5892791001, St. Louis, MO). The concentration of the protein lysate was measured using the Pierce BCA Protein Assay Kit (Thermo Fisher, 23225, Waltham, MA). Protein extracts (50 μg of protein) in DTT-containing SDS sample buffer were separated on 4–20% SDS-polyacrylamide gels and transferred to nitrocellulose membranes (Bio-Rad, Trans-Blot Turbo RTA, 1704271, Hercules, CA). The membranes were blocked using 5% nonfat dry milk, incubated overnight at 4°C with the appropriate antibodies (Supplementary Table 1), and incubated for 60 minutes with horseradish peroxidase-conjugated secondary antibodies at room temperature. Immunocomplexes were detected by a Chemidoc Imager (Bio-Rad, Hercules, CA) using Clarity Western ECL Substrate (Bio-Rad, 1705061, Hercules, CA). polyclonal antibody against MLK2 was purified from serum obtained from rabbits immunized with a peptide (EGQSQDNTVPLCGVY) corresponding to the amino acids 923–937 of mouse MLK2 (Boston Molecules, Inc., Waltham, MA).

Immunofluorescence staining

Human lung tumor tissues were fixed in 4% formalin and embedded in paraffin at Dr. Hassan Khalil’s laboratory, mouse tumor tissues were embedded in OCT (optimal cutting temperature compound, Leica, 14020108926, Teaneck, NY), or human/mouse endothelial cells were fixed in ice-cold methanol. The tissues were sectioned at a 7μm thickness. For tissue morphology, tissue slides were stained with Hematoxylin (Gill #2, Sigma, GSH216) and Eosin (Sigma, HT180, St. Louis, MO) according to the manufacturer’s instructions.

Tissues were permeabilized with 0.1% Triton X-100 in PBS (Cell Signaling Tech, #39487, Danvers, MA). Endogenous peroxidase and alkaline phosphatase (DAKO, S2003, Agilent, Santa Clara, CA), biotin activity (Vector Labs, SP2002, Newark, CA) and nonspecific binding sites were blocked in Envision Flex antibody diluent (DAKO, K800621–2; Agilent, Santa Clara, CA).

Sections or cells were incubated with the appropriate primary antibodies (Supplementary Table 2) in Envision Flex antibody diluent overnight at 4°C. Rabbit anti-MLK2 and rabbit anti-MLK3 were detected with a Tyramide Super Boost kit (Thermo, B40933, Waltham, MA) according to the manufacturer’s instructions. After the antibodies were removed in antibody stripping solution (VectaPlex Antibody Removal, Vector Labs, VRK-100, Newark, CA), multiplex immunofluorescence was carried out to co-localize with CD31. For mouse cryosections and HAEC cells, appropriate primary antibodies were incubated in Envision Flex antibody diluent overnight at 4°C, and secondary antibodies conjugated with Alexa 555 were applied for 1 hour at room temperature. Slides were mounted with an anti-fade mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI; Thermo Fisher, 8961; Waltham, MA) to stain the nuclei, and images were acquired with an inverted fluorescent wide filed microscope (Carl Zeiss, White Plains, NY) and ZEN microscopy software (Carl Zeiss, White Plains, NY).

RNA sequencing

RNA was extracted from WT and Mlk2−/−Mlk3−/− MLECs grown under hypoxia (1% oxygen for 90 minutes, Airgas, Hindham, MA) or normoxia (21% oxygen). An RNeasy plus kit (Qiagen, 74136, Germantown, MD) was used to isolate RNA, and three different methods were used for quality checking (QC) of the RNA samples: (1) a Nanodrop was used for preliminary quantitation, (2) agarose gel electrophoresis was used to test RNA degradation and potential contamination, and (3) an Agilent 2100 Bioanalyzer was used to check RNA integrity and quantitation and select samples with RNA quality > 9 or higher.

Library construction followed three steps of QC. After RNA QC, the rRNAs were removed with an Epicenter Ribo-Zero^™ Kit (Illumina, San Diego, CA). A total of 1 μg of RNA per sample was used for mRNA-Seq library construction using the NEBNext Ultra^™ RNA Library Prep Kit for Illumina (NEB, Ipswich, MA) according to the manufacturer’s recommendations. The purified RNAs were first fragmented randomly into short fragments of 150–200 bp in fragmentation buffer, followed by cDNA synthesis and the addition of random hexamers. After the first strand was synthesized, custom second-strand synthesis buffer (Illumina, San Diego, CA), dNTPs (dUTP, dATP, dGTP, and dCTP), and DNA polymerase I were added to synthesize the second strand. This was followed by purification by AMPure XP beads, terminal repair, polyadenylation, sequencing adapter ligation, size selection, and degradation of second-strand U-containing cDNA by the USER enzyme. The strand-specific cDNA library was generated after the final PCR enrichment. The concentration of the library was first quantified by a Qubit2.0 fluorochrome, after which the library was diluted to 1 ng/μL, after which the insert size was checked via an Agilent 2100, after which the library was further quantified via qPCR (library concentration > 2 nM).

Once the library had qualified, it was sent for sequencing on an Illumina HiSeq platform according to the effective concentration and data volume. After the raw sequenced reads came out with the available reference genome, bioinformatics analysis was performed at the Novogene Bioinformatics Institute in Beijing, China. Mapping to the reference sequences and transcript assembly were performed by aligning the clean reads to the reference genome with TopHat2 [76], and the Tophat2 algorithm involved three steps: (1) mapping the reads against the transcriptome (optional); (2) mapping the full-length reads to the exons; and (3) mapping the partial reads to two exons.

Statistical analysis

All data are expressed as means ± SEM. The experiments were performed multiple times, and the figures shown represent the results of 3–6 independent experiments. Our analysis strategy involved first testing for equal variance and normality to determine whether parametric or nonparametric tests were appropriate. Based on these results, we performed statistical comparisons between two groups using the parametric Student’s t test and among multiple groups using one- or two-way ANOVA with post hoc Tukey–Kramer multiple comparisons, as indicated in the figure legends. P values are reported to three decimal places in the figures. All statistical analyses were performed using StatView version 5.0 (SAS Institute, Cary, NC) or GraphPad Prism version 10 (GraphPad Software, La Jolla, CA).