The three members of the Vav family proteins form complexes that concur to foam cell formation and atherosclerosis[S]

During foam cell formation and atherosclerosis development, the scavenger receptor CD36 plays critical roles in lipid uptake and triggering of atherogenicity via the activation of Vav molecules. The Vav family includes three highly conserved members known as Vav1, Vav2, and Vav3. As Vav1 and Vav3 were found to exert function in atherosclerosis development, it remains thus to decipher whether Vav2 also plays a role in the development of atherosclerosis. In this study we found that Vav2 deficiency in RAW264.7 macrophages significantly diminished oxidized LDL uptake and CD36 signaling, demonstrating that each Vav protein family member was required for foam cell formation. Genetic disruption of Vav2 in ApoE-deficient C57BL/6 mice significantly inhibited the severity of atherosclerosis. Strikingly, we further found that the genetic deletion of each member of the Vav protein family by CRISPR/Cas9 resulted in a similar alteration of transcriptomic profiles of macrophages. The three members of the Vav proteins were found to form complexes, and genetic ablation of each single Vav molecule was sufficient to prevent endocytosis of CD36. The functional interdependence of the three Vav family members in foam cell formation was due to their indispensable roles in transcriptomic programing, lipid uptake, and activation of the JNK kinase in macrophages.

the signals resulting from CD36 ligand engagement, typically the modified LDL, such as oxidized (Ox)-LDL, remain to be fully characterized (4).
Vav1, a guanine nucleotide exchange factor for the Rho family guanosine triphosphatases, has been found to be critical for foam cell formation and activation of the JNK kinase (5,6). Vav1 shares highly conserved domains with Vav2 and Vav3, two other members of the Vav protein family. Interestingly, polymorphisms in Vav2 and Vav3 have been associated with human cardiovascular diseases as risk factors, and the loss of Vav2 or Vav3 resulted in a similar phenotype of disturbed cardiovascular homeostasis (7)(8)(9). A previous study showed that the treatment of macrophages with Ox-LDL induced phosphorylation of the three Vav family proteins (5). Although genetic models confirmed the contribution of Vav1 and Vav3 to foam cell formation and to atherosclerosis in ApoE-deficient mice, the role of Vav2 in the setting of atherosclerosis remains to be characterized (5,6).
In this study, we demonstrated that Vav2 contributed to atherosclerosis using both a macrophage cell line and animal models. We found that Vav2 deficiency significantly diminished Ox-LDL uptake and CD36 signaling in RAW264.7 macrophages as well as in primary macrophages. Genetic disruption of Vav2 in ApoE-deficient C57BL/6 mice dramatically inhibited the severity of atherosclerosis. Even though mice simultaneously deficient in the three Vav family proteins were resistant to foam cell formation by affecting actin polymerization, it remains to be deciphered how the deletion of each member of the three Vav family proteins contributes to atherosclerosis (10). Therefore, an important aim of our study was to compare the contribution of each Vav family protein to foam cell differentiation and provide a mechanistic explanation for their nonredundant role in foam cell formation.

Design of sgRNAs for macrophages and mice
sgRNAs for targeting different genomic loci were designed by the online bioinformatics tool CRISPOR (11). Candidate sgRNAs were selected according to the ranking of two critical scores. One was the Fusi/Doench score suggesting best performers of sgRNA expressed by the U6 promoter from transfected plasmids, and the other was the Moreno-Mateos score considering good candidates produced by the T7 promoter for knockout mice generation. In our study, different sets of sgRNAs were selected for CRISPR/ Cas9 editing in cell lines and mouse zygotes, as sgRNA expression was driven by U6 and T7 promoters, respectively. Animals C57BL/6 (B6) mice 6 to 8 weeks old from WT and ApoE-deficient backgrounds were purchased from Beijing Vital River Laboratory Animal Technology, and Vav1-OST knock-in mice were constructed at the Centre d'Immunophénomique. All animal procedures were performed according to guidelines approved by the committee on animal care at Xinxiang Medical University.

Microinjection of mouse zygotes
Eggs from ApoE / B6 mice were fertilized in vitro by isogenic sperm cells from males. Briefly, 127 oocytes were collected from 8-week-old ovulated female ApoE / mice and inseminated with spermatozoa obtained from the cauda epididymides of 12-weekold ApoE / males. The resultant 120 fertilized eggs that could be differentiated from nonfertilized cells under a stereomicroscope and such fertilized eggs were selected for microinjection. Cas9 mRNA (50 ng/l) and sgRNA (50 ng/l) were microinjected into the cytoplasm of fertilized eggs. Injected eggs were cultured in M2 medium at 37°C in 5% CO 2 overnight to the two-cell stage and then transferred into the oviductal ampullas of ICR foster maternal mice the next day.
Cell culture RAW264.7 macrophages were cultured in DMEM containing 10% FBS, 100 U/ml penicillin, and 100 µg/ml streptomycin. For the isolation of peritoneal macrophages, age-and sex-matched WT and Vav2 / mice were injected intraperitoneally with 1.5 ml 4% sterile thioglycollate brewer broth (BD). Four days after the injection, cells were harvested by intraperitoneal lavage with icecold PBS. Cells were then seeded in DMEM medium with 10% FBS.

Plasmids
CRISPR/Cas9-mediated gene deletion in RAW264.7 cells. To generate gene-specific deletion via the CRISPR/Cas9 system, two (or four) sgRNAs were designed for each member of the Vav family proteins and respectively cloned into CRISPR-expressing pX458 or its derivatives with various fluorescent reporters that enabled single-cell sorting as described previously (12).
Expression of OST-tagged Vav proteins via knock-in in the Rosa26 locus. To achieve CRIPSR/Cas9 cleavage that facilitated homology-directed repair (HDR) in the Rosa26 locus, two adjacent sgRNAs target sequences within the first intron of Rosa26 were selected and constructed into CRISPR-expressing pX458-DsRed2, respectively. To generate the template for HDR, pKR26-iBFP, a Rosa26 targeting backbone vector based on previous vector pR26 CAG/BFP Dest (Addgene), was synthesized (Bioligo) that contained 1 kb 5′ and 3′ homologous arms targeting into the Rosa26 locus, a CAG promoter and an AscI restriction site used for the insertion of a protein of interest, followed by a blue fluorescent protein reporter (BFP) linked with an internal ribosomal entry site (IRES). Mouse Vav1 and Vav3 cDNAs (ENSMUST00000005889.15; ENSMUST00000046864.13) were amplified by PCR using cDNA obtained from WT RAW264.7 total RNA, and Vav2 cDNA (ENSMUST00000056176.7) was amplified by PCR using the plasmid pCMV-mVav2-PGK-Puro (Genomeditech). Each cDNA and the synthesized OST tag (Bioligo) were assembled by PCR with overlapping primers and cloned into the pKR26-iBFP vector via the AscI restriction site using the NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs). All plasmids were confirmed by restriction enzyme digestion and Sanger sequencing. Before transfection, all targeting vectors were linearized with the unique restriction site XhoI or EcorRI and purified (Qiagen). 10 min and aspirated into the 10 l Neon ® Tip. RAW264.7 cells were treated using the electroporation condition with 1,400 V/20 ms/2 pulses. After 48-72 h of electroporation, cells were subjected to FACS sorting.

Generation of knockout and knock-in cell lines
For creating knock-in cell lines, a dual fluorescent reporter system was designed consisting of the DsRed2 reporter from the CRISPR/Cas9-expressing vector and the other BFP reporter from the linearized targeting vector. In bulk sorting 10 cells were sorted into each well of a 96-well microplate from a minor population by gating on BFP + DsRed2 + cells in the parental Vav1, Vav2, or Vav3 knockout RAW264.7 macrophages. The sorted cells were cultured in the growth medium for 7-14 days and further transferred into a 48-well plate for cell proliferation. All proliferated bulk cells were screened for BFP expression by flow cytometry as well as PCR genotyping to confirm successful recombination occurrence. A second sorting was applied for isolates of BFP + Vav-OST + cells.

Fluorescence PCR and capillary array electrophoresis
To genotype the knockout cell lines, DNA extracts of clonal cells were subjected to PCR using 5′-FAM-labeled primers (supplemental Table S1). The PCR amplicons were resolved using an ABI 3730 DNA analyzer. Data analysis was performed by GeneMapper software version 3.1. The positions of the peaks indicate the sizes or lengths of PCR products by using ROX-labeled standards as described previously (13).

Generation of Vav1-Halo, Vav2-SNAP, or Vav3-SNAP fusion protein vectors and multichannel fluorescent imaging
The coding sequences of Vav1 were cloned into EcoRI-linearized pSNAP-tag(m) vector (Addgene), and the coding sequences of Vav2 and Vav3 were cloned respectively into EcoRI-linearized pHalo-tag vector (Promega) using NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs). 293T cells were cotransfected with Vav1-Halo and Vav2-SNAP or Vav1-Halo and Vav3-SNAP using Lipofectamine 3000 (ThermoFisher Scientific). Cells were labeled 36 h after transfection with 5 mM of the SNAP-Cell 647-SiR substrate (New England BioLabs) and Janelia Fluor® 549 HaloTag® Ligand (Promega) for 30 min. After washing three times with DMEM, images were taken by a confocal microscope (Leica).

Detection of cell surface CD36 and DiI-OxLDL uptake by flow cytometry
Cells were collected and washed with cold PBS, and cell suspensions were incubated with Biotinylated anti-CD36 Antibody (Biolegend) for 30 min on ice, washed with cold PBS three times, and followed by fluorescent secondary antibody staining for another 30 min on ice. Unstained cells were used as a negative control. DiI-OxLDL (Yiyuan Biotechnology) was used to trace the Ox-LDL uptake. Cells were incubated with DiI-OxLDL for 1, 2, or 3 h at 37°C in 5% CO 2 . RAW264.7 macrophage cells were harvested, washed, and resuspended in FACS buffer. Samples were acquired on the FACS Canto flow cytometer (BD). Data were further analyzed by FlowJo10.1.

Foam cell formation assay
Macrophages were incubated with 50 g/ml Ox-LDL (Yiyuan Biotechnology) for 24 h. Cells were washed with PBS and fixed with 4% paraformaldehyde for 10 min. After incubation with isopropanol for 5 min, cells were stained with 0.5% Oil Red O in isopropanol for 30 min and washed with 85% isopropanol. After washing with 1× PBS, macrophages were photographed under a microscope at 200× magnifications.

Affymetrix microarray expression analysis and pathway enrichment analysis
Genome-wide gene expression analysis was performed using Affymetrix GeneChip Mouse Genome 430 2.0 Array (CapitalBio Technology). The data are deposited in the Gene Expression Omnibus database under the accession number GSE125746. Gene Ontology enrichment analysis to illustrate pathways affected by Vav deficiency were performed by using R package goseq version 1.16.2, and Gene Ontology terms with a corrected P value >0.05 were excluded. For Kyoto Encyclopedia of Genes and Genomes analysis, the genes were mapped directly to the Kyoto Encyclopedia of Genes and Genomes database. The enriched pathways were then obtained using a q-value cutoff of 0.05 with the R hypergeometric function and R q-value package (14).

Quantitative real-time PCR
RNA was extracted using RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. cDNA was synthesized using a RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). Individual quantitative real-time PCR was performed using gene-specific primers as shown in supplemental Table S2.

ApoE
/ and ApoE / mice were fed a high-fat diet (HFD; 40% fat, 1.25% cholesterol; Research Diets, Inc.). After 14 weeks the mice were euthanized by carbon dioxide. Hearts were perfused with PBS and 4% paraformaldehyde. For en face analysis, the entire aorta from the heart was removed, dissected, and stained with Oil Red O, and the lesion area was presented as a percentage of the total area of the aorta. For the aortic sinus analysis, serial cross-sections of the aortic root (10 m) were stained with H&E for lesion quantification. Images were acquired on a Pannoramic MIDI II (3D HISTECH).

Flow cytometry analysis and cell sorting
Flow cytometric analysis was performed by staining the blood or aorta cells of mice with monoclonal antibody mixes. The antibodies used in this study are listed in supplemental Tables S3 and S4. The antibody labeling experiments were done as documented in our previous study (15). In brief, for blood, 1 million cells were stained in 100 µl with antibody mixes and acquired on the FACS Canto flow cytometer (BD). For immunophenotyping of the aorta, mice were perfused with PBS from the left ventricle of the heart. Adipose tissues and para-aortic lymph nodes were removed before tissue dissociation and single-cell preparation. The whole aorta included ascending, aortic arch, descending thoracic aorta, and abdominal aorta in this study. The tissue was segmented by surgical scissor followed by digestion with 50 µg/ml Liberase DH (Roche) and 40 U/ml DNase I (New England Biolabs) for 30 min at 37°C (16). The single cells were stained with an antibody mix and acquired on the FACS Canto flow cytometer (BD).
Cell sorting was performed with a FACSAria™ Fusion flow cytometer (BD). Blood monocytes from WT B6 mice were gated on CD5  , Ly6G  , and CD19  cells and CD115 + and CD11b + cells, which were further divided by CD36 surface expression for RNA sequencing. FACS data were analyzed by FlowJo10.1.

Immunoprecipitation and immunoblot analysis
For immunoprecipitation (IP), protein lysates were incubated with prewashed Strep-Tactin Sepharose beads (IBA) for 1.5 h at 4°C on a rotary wheel. Beads were then washed five times with 1 ml lysis buffer in the absence of detergent and of protease and phosphatase inhibitors. Immunoblot analysis was performed as previously described (17). Anti-Vav1, phospho-JNK, and JNK antibodies were from Cell Signaling Technology. Anti-Vav2 and Vav3 antibodies were purchased from Abcam, and anti-phosphotyrosine antibody (clone 4G10) was from Millipore.

CD36 cross-linking and internalization assay
Cells were washed twice with ice-cold RPMI and cooled on ice. Anti-CD36 IgA was added to the cells for 10 min, and the cells were washed three times with cold RPMI. The cells were then incubated with anti-mouse IgA-FITC for another 10 min, the cells of the control group were left on ice, and the cross-linking group cells were transferred to the 37°C incubator for 30 min in RPMI. To wash out the surface-bound antibodies, the cells were incubated with cold acid wash buffer (0.5 M glacial acetic acid, 150 mM sodium chloride, pH 2.5) for 2 min, followed by recovery in ice-cold RPMI for 2 min. Cells were then fixed for 15 min in 4% paraformaldehyde on ice. Samples were imaged by confocal microscope (Leica) or analyzed by FACS Canto flow cytometer (BD) as described previously (18).

Statistical analysis
All data were compared between two groups and analyzed with GraphPad Prism software version 7.0. Statistical significance was assessed by unpaired, two-tailed Student's t-test. P < 0.05 was considered significant.

Genetic ablation of Vav2 in macrophages resulted in diminished foam cell formation
To determine the role of Vav2 in the formation of foam cells and compare it to that of Vav1 and Vav3, we first compared the expression level and intracellular distribution of Vav2 with that of Vav1 and Vav3. Accordingly, thioglycollate-elicited peritoneal macrophages from 8-week-old B6 mice were subjected to mRNA quantitation and confocal microscopic analysis (19). As shown in supplemental Fig. S1A, less expression of Vav2 mRNA was detected using quantitative real-time PCR expressed compared with that of Vav1 and Vav3. In further experiments, we compared the read counts of RNA sequencing from blood monocytes sorted from B6 mice and found that in both CD36 + and CD36  monocytes Vav2 read counts were also less than those of Vav1 and Vav3 (supplemental Fig. S1B). Interestingly, even though low levels of Vav2 mRNA were found expressed in both macrophages and blood monocytes, the Vav2 protein was readily detected in murine peritoneal macrophages and showed a cytosolic distribution. In comparison, Vav1 was distributed in both the cytosol and nucleus, while Vav3 was more abundant in the macrophage nucleus (supplemental Fig. S1C). We also performed confocal microscopic analysis in addition to immunoblotting (IB) of cytosolic or nuclear cell lysates for Vav1, Vav2, and Vav3 and found that these three proteins had both cytosolic and nuclear distribution. Confocal microscopic detection of Vav3 showed that it was more expressed in the nucleus, consistent with the IB results (supplemental Fig. S1D).
To unambiguously determine the function of Vav2 in foam cell formation, we used the CRISPR/Cas9 genome editing tool to delete the Vav2 gene in RAW264.7 macrophages.
As shown in Fig. 1A, guide RNAs targeting specifically Vav2 but not Vav1 and Vav3 were designed to disrupt exon 6 of Vav2 (ENSMUST00000056176.7). Plasmids expressing two independent sets of CRISPR/Cas9 sgRNAs targeting Vav2 were engineered with ECFP and DsRed2 fluorescent protein reporters and electroporated into RAW264.7 macrophages (12). In brief, cells were gated on ECFP + DsRed2 + macrophages and cloned by single-cell sorting into a 96well plate. Capillary array electrophoresis aided genotyping permitted to identify five independent clones with bi-allelic Vav2 DNA deletions. As shown in Fig. 1B, when tested by Western blot using a specific Vav2 rabbit monoclonal antibody that cross-reacts with human and mouse proteins (20), all five clones of mutant cells were found completely deprived of Vav2. Sanger sequencing depicted the exact alteration of DNA sequence as a result of CRISPR/Cas9 editing (Fig. 1C). Functional assays using regular Ox-LDL and DiI-OxLDL showed that Vav2 deficient macrophages were significantly diminished in lipid uptake (Fig. 1D, E). Strikingly, Ox-LDL-induced phosphorylation of JNK was also significantly inhibited in Vav2-deficient macrophages (Fig. 1F). Importantly, Vav2-deficient cells still expressed normal levels of Vav1, Vav3, and CD36 (supplemental Fig.  S2D; data not shown). Therefore, our results showed that regardless of the low Vav2 mRNA levels found in macrophages, Vav2 is indispensable for lipid uptake in macrophages and Ox-LDL-induced signaling in foam cells.

Reexpression of Vav2 completely rescues the foam cell phenotype in Vav2-deficient macrophages
To exclude off-target effects of CRISPR/Cas9 treatment during Vav2 deletion in RAW264.7 macrophages, we reexpressed Vav2 in the Vav2-deficient RAW264.7 cells using an expression cassette permitting the expression of Vav2 molecules under the control of the CAG promoter and tagged at their C terminus with OST. Knock-in of the expression cassette in the mouse Rosa26 locus was achieved via homologous-dependent recombination ( Fig. 2A). As shown below, the fused OST tag was used for affinity purification of Vav2 via Strep-Tactin Sepharose beads as described in our previous studies on the OST-tagged proteins such as Vav1 (17,21). As illustrated in Fig. 2A and B, the IRES-BFP reporter element linked to Vav2-OST enabled us to perform cell sorting without any drug resistance selection, facilitating isolation and identification of Vav2-OST knock-in cells. It is important to note that by transfecting another plasmid encoding the DsRed2 reporter together with the Cas9 protein and guiding RNA to target the Rosa26 locus, as described in the Materials and Methods section, we could find a minor BFP + DsRed2 + population in electroporated Vav2-deficient RAW264.7 cells (Fig. 2C). Out of such a population, 10 cells were bulk-sorted each into one well, which gave rise to BFP + Vav2-OST + cells when the successful recombination of the CAG-Vav2-OST-IRES-BFP cassette in the Rosa26 locus occurred (Fig. 2D). Vav2 and found that BFP + Vav2-OST + could be identified by anti-Vav2 monoclonal antibody after bead purification (Fig. 2F). In further experiments, we sought to confirm that reexpression could rescue the phenotype of Vav2deficient macrophages in foam cell formation. Indeed, as shown by FACS analysis, Vav2 knockout cells as well as BFP  cells that were still Vav2-deficient had significantly lower capacity in Ox-LDL uptake compared with the BFP + Vav2-OST + cells (Fig. 2G). Analysis of the reconstituted BFP + Vav2-OST + foam cells showed that they had completely restored the phosphorylation of key signaling molecules following Ox-LDL treatment (Fig. 2H). The knock-in experiments and functional analysis showed that the reexpression of Vav2 was sufficient to recover the capacity of macrophages in foam cell formation. Therefore, the phenotype observed in Vav2-deficient RAW264.7 macrophages was indeed due to the lack of Vav2 and not to adventitious effects resulting from CRISPR/Cas9 action.

Genetic ablation of Vav2 in C57BL/6 mice from an ApoEdeficient background
To confirm in vivo our observation made in Vav2-deficient RAW264.7 macrophages, we relied on a protocol that we have developed that permits the use of CRISPR/Cas9 to achieve Vav2 deletion in mice from an ApoE-deficient background (13). In brief, ApoE female mice were superovulated, followed by in vitro fertilization using the sperm of ApoE male mice. Four independent sgRNAs were designed to achieve the deletion of exon 1 of Vav2 in in vitro fertilization-derived F0 founder mice, which were further intercrossed to produce progeny for phenotyping (Fig. 3A, B). DNA sequencing showed that the founder mice harbored various mutated alleles with large DNA fragment deletions, resulting in frame shift (Fig. 3C). As shown in Fig. 3D, in the F1 mice derived from F0 intercrosses, we confirmed the lack of Vav2 protein expression using the livers of homozygous mutant mice because liver macrophages or Kupffer cells constitute a significant proportion of the organ. As expected, such mice lacked a detectable Vav2 protein but did not show an alteration of Vav1 and Vav3 expression (Fig. 3E). In parallel, we also developed Vav2-deficient mice from an ApoE WT B6 background by crossing such Vav2 / ApoE / mice to B6 WT mice.
Consistent with our results obtained with Vav2-deficient RAW264.7 cells, the Vav2-deficient peritoneal macrophages from the ApoE WT background showed a significant decrease in Ox-LDL uptake and phosphorylation of JNK (Fig. 3F, G). Therefore, using ex vivo peritoneal macrophages from mice deficient in Vav2, we were able to confirm a role for Vav2 in foam cell formation.

. Genetic deletion of Vav2 in ApoE
/ C57BL/6 mice. A: Workflow for generation of Vav2 knockout mice using CRISPR/Cas9. Experimental steps involved mice superovulation, collection of fertilized eggs, sgRNA and Cas9 mRNA synthesis, microinjection, and positive founder identification by PCR and sequencing. B: Schematic representation of the Vav2-targeting sgRNA sequences. The sgRNA-3, sgRNA-4, and sgRNA-5 targeting exon 1 and the sgRNA-6 located in the first intron adjacent to exon 1 were selected, and protospacer adjacent motifs (PAMs) are shown in red letters. C: DNA sequencing analysis showed the presence of the intended Vav2 knockout mutation in three F0 animals (denoted as 1, 2, and 3). The deletion or insertion size for each mutant is indicated below the WT sequence. Red letters correspond to the PAM sequences and purple letters to the sgRNA sequences; red dashes correspond to deleted nucleotides and green letters to nucleotide insertion. D, E: Western blot was performed using anti-Vav1, Vav2, Vav3, or GAPDH-specific antibodies to verify the loss of Vav2 and the presence of Vav1 and Vav3 using the cell lysate of Vav2 / liver. Results from our founder ApoE / Vav2 / mice are shown and compared with ApoE / mice. F: FACS analysis of DiI-Ox-LDL uptake and CD36 expression on peritoneal macrophages of B6 and Vav2 / B6 mice. Data are presented as the mean ± SEM of triplicated samples involving three different mice. Statistics by two-tailed, unpaired Student's t-test: ****P < 0.0001. G: Immunoblot analysis of phosphorylated JNK in lysates of peritoneal macrophages from B6 and Vav2 / B6 mice stimulated for 0-30 min with 50 g/ml Ox-LDL. Representative data are from three independent experiments.

Fig. 4. Assessment of atherosclerosis development in ApoE
/ Vav2 / mice. A: Body weight of ApoE / Vav2 / mice fed over 14 weeks with an HFD compared with ApoE / controls. At least eight animals were analyzed for each time point and each genotype. B: Comparisons of serum lipid composition between ApoE / Vav2 / and ApoE / mice. Total serum cholesterol, triglyceride, LDL, and HDL concentrations were measured in two independent experiments in mice fed an HFD for 14 weeks (n = 8-9). C: En face micrographs of mounted aortas stained with Oil Red O (red) of the specified animals fed an HFD for 14 weeks (Nikon; 0.335×). Quantitation of plaque areas relative to the area of the aorta from two independent experiments (n = 8-9). D: Two representative microscopy images of aortic root sections from ApoE / Vav2 / and ApoE / mice fed an HFD for 14 weeks (scale bar = 400 m) and quantitation of the plaque area relative to the area of the aortic lumen (n = 8-9). E: Gating strategy used to analyze aortic macrophages isolated from digested mouse aortas of ApoE / Vav2 throughout the experiment with the HFD (Fig. 4A). Vav2 deficiency in Vav2 / ApoE / mice did not change the levels of serum lipid fractions compared with ApoE / controls fed an HFD for 14 weeks (Fig. 4B). Strikingly, Vav2 / ApoE / mice had significantly decreased atherosclerosis compared with control mice in the en face analysis of aorta 14 weeks after HFD treatment. The lesion area of ApoE / controls was 14.36%, and it dropped to 10.24% in the Vav2 / ApoE / mice (Fig. 4C). The sections of aortic sinus showed comparable pathology by H&E staining between Vav2 / ApoE / mice and ApoE / controls (Fig.  4D). Due to the discrepancy in pathology found in whole aorta and the aortic root, we further analyzed Vav2-deficient mice from an ApoE-deficient background at an earlier stage, as differences in the pathology between the two groups of mice might be masked during more advanced stages. Vav2 / ApoE / mice and ApoE / controls that were fed an HFD for 8 weeks showed significant differences in pathological severity in the aortic root (supplemental Fig. S3A, B). Vav2 / ApoE / mice had significantly mitigated pathology in the aortic root by H&E staining (supplemental Fig. S3B). However, conversely, we did not observe significant differences in the entire aorta in mice fed an HFD for a shorter period of time (supplemental Fig. S3A). We further compared the macrophage phenotype in the aorta between Vav2 / ApoE / mice and ApoE / controls fed an HFD for 20 weeks, and no significant differences were observed between aortic Ly6C high and Ly6C low macrophages (Fig. 4E). In both of these populations of aortic macrophages, Vav2 / ApoE / mice and ApoE / mice had no significant difference in the expression of CD115, CD11b, and F4/80 (Fig. 4F). It is interesting to note that we did observe decreased CD36 expression in blood monocytes from Vav2 / ApoE / mice placed on an HFD for 4 weeks or on a normal diet compared with ApoE / controls (supplemental Fig. S3C, D). However, such a difference in circulating monocytes was not observed in Vav2 / ApoE / mice placed on an HFD for 8 weeks and 14 weeks, as shown in supplemental Fig. S3E, suggesting that the decrease in CD36 expression in doubleknockout mice may not explain the mitigated atherosclerosis. Therefore, Vav2 deficiency ameliorates atherosclerosis, as observed in mice deficient in either Vav1 or Vav3, and the loss-of-function Vav2 could not be compensated by the presence of Vav1 and Vav3. The possibility to directly edit Vav2 from an ApoE-deficient B6 background allowed us to analyze in a fast-track mode the effect of Vav2 deficiency on atherosclerosis by placing the Vav2 / ApoE / mice on an HFD for 12-20 weeks, and experiments with Vav2 / ApoE / mice fed an HFD confirmed the contribution of Vav2 to atherosclerosis in vivo.

Deficiency of each member of the Vav family proteins resulted in highly conserved transcriptomic alteration in macrophages
To analyze whether the loss of individual members of the Vav family proteins resulted in similar transcriptomic alterations, we performed a transcriptomic analysis of macrophages that were deficient in Vav1, Vav2, or Vav3. The RAW264.7 cells deficient in Vav1 or Vav3 were prepared by CRISPR/Cas9 targeting in the same manner as described above for Vav2 knockout. sgRNAs targeting Vav1 and Vav3 were designed on the basis of the exon conservation between different transcripts for each gene (supplemental Fig. S2A). After transfection of the plasmids engineered with ECFP and DsRed2 reporters, fluorescent protein-positive cells were sorted to obtain individual clones. Three representative clones for Vav1 or Vav3 knockout based on genotyping results were further subjected to Sanger sequencing, all of which confirmed CRISPR editing. Among the mutant clones, Vav3 knockout cell clone C26 harbors two mutant alleles caused by sgRNA-3 and sgRNA-4 targeting, while the rest of CRIPSR/Cas9 targeting results in only one detectable allele (supplemental Fig. S2B). Furthermore, the absence of Vav1 or Vav3 was observed by IB (supplemental Fig. S2C). It is of note that such Vav3 knockout cells clones as C17, C24, and C26 were validated as mutants by DNA sequencing; however, in C17 and C24 but not C26 cells Vav3 was still detected by IB, suggesting that the anti-Vav3 antibody was not as specific as the anti-Vav1 and anti-Vav2 antibodies. To exclude the possibilities of off-targeting among Vav genes by CRISPR/Cas9, we validated macrophages deficient in a given Vav gene for the expression of the other two Vav molecules (supplemental Fig. S2D). The results showed that the sgRNAs we designed were of the intended specificity. As expected, we found that the mutant RAW264.7 cells that were deficient in Vav1 or Vav3 had a significant decrease in Ox-LDL uptake (Fig. 5A, B). The attenuated activation of foam cell signaling as measured by JNK phosphorylation following Ox-LDL treatment was also observed with a magnitude comparable to Vav2-deficient RAW264.7 cells (Fig.  5C, D). It is important to note that we further generated RAW264.7 cells deprived of the three Vav family members. Such Vav1 / Vav2 / cells were validated by fluorescent PCR and capillary array electrophoresis and IB (supplemental Fig. S4A, B). In three independent clones of triple-knockout cells, Vav1 and Vav2 proteins were found to be absent, whereas the Vav3 showed decreased protein band intensity, which is likely due to adventitious cross-reactivity of the anti-Vav3 antibody. In these triple-knockout RAW264.7 cells, we found that their lipid uptake was diminished in an extent comparable to Vav1 single-gene-knockout RAW264.7 cells at three different time points (1, 2, and 3 h), which supported that each individual Vav member was essential for foam cell formation (supplemental Fig. S4C). We next extracted total RNA from Vav1-, Vav2-, and Vav3-deficient macrophages as well as from RAW264.7 WT cells and subjected them to genome-wide transcriptomic analysis. The number of upregulated and downregulated genes between Vav1, Vav2, and Vav3 knockout macrophages was comparable (Fig. 5E). The magnitude of transcriptomic alteration in knockout cells involving three different Vav genes was also similar, as shown in Fig.  5F. More importantly, we used the web-based GeneVenn (http://genevenn.sourceforge.net) gene overlap analysis to assess the transcriptional consequence of each Vav gene deletion in macrophages (22). We found that 72.4% (176 of 243) of upregulated genes in Vav1 knockout were present in Vav2 knockout macrophages, and more than 79.9% (135 of 169) of upregulated genes in Vav3 knockout were found in Vav2 knockout cells. In the downregulated genes, a similar trend was observed: 81.3% (390 of 480) of the downregulated genes in Vav1 knockout macrophages being present in Vav2 knockout cells and 77.1% (326 of 423) of downregulated genes in Vav3 knockout cells being found in Vav2 knockout cells (Fig. 5G). To specify the genes that are not commonly modulated by three Vav members, information of exclusively upregulated or downregulated genes in each Vav-deficient macrophage are listed in supplemental Table S5. Pathway classification and enrichment analysis of the conserved transcriptomic alteration suggested that each of the three Vav proteins could be participating in cell cycle, cell adhesion, metabolism, and phagocytosis function (supplemental Fig. S5). Therefore, transcriptomic analysis showed that each of the three Vav genes is necessary to maintain conserved transcriptional programs in macrophages.

Vav1, Vav2, and Vav3 proteins form ternary complexes and are simultaneously required for CD36 internalization
Vav proteins have both guanine nucleotide exchange factor activity and scaffolding functions that lead them to participate in a broad range of protein-protein interactions (21,23). We sought to determine whether the three Vav family members could form complexes in macrophages. For that purpose, we generated Vav1-OST and Vav3-OST knock-in macrophages using the same strategy described above for Vav2-OST from Vav1-and Vav3-deficient backgrounds. Following the procedure described in Vav2-OST knock-in experiments, protein levels of Vav1 or Vav3 were detected by Western blot in knock-in cells and pull-down with Strep-Tactin Sepharose beads to validate them (Fig.  6A, B). Akin to Vav2-OST molecules, Vav1-OST and Vav3-OST molecules were also capable of rescuing Vav1deficient and Vav3-deficient RAW264.7 cells, respectively, for lipid uptake and activation of foam cell signaling (supplemental Fig. S6A-C). Vav1-OST, Vav2-OST, and Vav3-OST proteins were isolated by pull-down and immunoblotted with an anti-phosphotyrosine antibody, demonstrating that Vav1-OST, Vav2-OST, and Vav3-OST proteins could be phosphorylated following Ox-LDL treatment of RAW264.7 cells (Fig. 6C).
Considering that the Vav1, Vav2, and Vav3 fused to an OST tag were functional, we isolated each of them by affinity purification and analyzed whether they were capable of forming complexes with the two remaining members of the Vav family proteins. Following pull-down of Vav1, we were able to detect Vav2 and Vav3 as the Vav1 interactors, WT macrophages constituting negative control. Similarly, after pull-down of Vav2, we could detect Vav1 and Vav3 as Vav2 interactors (Fig. 6D). We also identified that Vav2 and Vav3 could be detected as Vav1 interactors using peritoneal macrophages from the Vav1-OST knock-in mice (Fig. 6E). To further validate the interaction between Vav members, we applied high-resolution imaging using the Halo and SNAP tag technology (24,25), which allowed us to label the Vav molecules in live cells. Confocal microscopy images showed that Vav1 colocalized with both Vav2 and Vav3   . 6F). Therefore, the three members of the Vav family proteins could form interactive complexes in macrophages, which could explain their nonredundant contribution in foam cell formation and atherosclerosis development. CD36 internalization is a critical process for diverse signaling processes, including the uptake of Ox-LDL (26,27). As Vav1, Vav2, and Vav3 form interactive complexes, and the absence of a single Vav gene resulted in highly conserved transcriptomic alteration and identical deficiency in foam cell formation, we therefore further tested the dependence of CD36 internalization on such interactive complexes. The cells from Vav1, Vav2, and Vav3 knockout clones with WT controls were incubated with anti-CD36 IgA that was then cross-linked using FITC-labeled goat anti-mouse IgA  chain. In Vav1, Vav2, and Vav3 knockout RAW264.7 macrophages, CD36 staining on the cell surface showed comparable fluorescence intensity under a confocal microscope; however, the intercellular staining was dramatically reduced in knockout cells (Fig. 6G). In further experiments we analyzed eight replicates for each genotype involved in the imaging study by flow cytometry. The FACS data of the CD36 mean fluorescence intensity (MFI) showed that deficiency in each member of the Vav family proteins resulted in a significant decrease in CD36 internalization (Fig. 6H). Taken together, our results showed that Vav1, Vav2, and Vav3 form interactive complexes and all three Vav family members are mandatory for CD36 internalization and therefore interdependent in promoting foam cell formation and atherosclerosis.

DISCUSSION
Atherosclerosis is a metabolic disorder accompanied by inflammation involving both innate and adaptive immune cells (28,29). A prerequisite pathological change is that foam cells are formed when macrophages engulf an excessive amount of lipids, and foam cell formation itself could be target for therapeutic intervention (30). Compelling evidence shows that the scavenger receptor CD36 and its downstream signaling molecules in macrophages play crucial roles in foam cell formation and atherosclerosis (3,(31)(32)(33). Even though numerous studies confirmed the role of the CD36-JNK signaling axis in foam cell formation, more detailed dissection of signaling molecules that participate in this critical process of foam cell formation and progression of atherosclerosis is still necessary (3,4). Vav1 has been implicated in CD36-mediated foam cell formation (6). As the Vav family proteins share highly conserved functional domains, one could anticipate that Vav family molecules could redundantly transduce CD36 signaling and that a deleterious mutation in one member of the Vav protein family can be compensated by the two remaining members of the Vav protein family. Unexpectedly, the results from the current study showed that three Vav proteins are required simultaneously for proper foam cell formation. We first delineated that three Vav genes are expressed in different abundance and that their intracellular distribution differs in primary mouse macrophages. The expression and distribution might partially explain a paralleled requirement of three Vav proteins in foam cell formation because Vav3 was found to be more abundant in the nucleus of macrophages. Prior to the current study, the contribution of Vav2 to atherosclerosis has not been characterized, even though its phosphorylation has been detected in murine macrophages upon Ox-LDL treatment (5). Using CRISPR/Cas9 editing in the RAW264.7 macrophage cell line, we determined that Vav2 deficiency in the presence of normal levels of Vav1 and Vav3 was sufficient to inhibit foam cell signaling. The knock-in experiment proved that Vav2 was indispensable for foam cell formation and activation of JNK. We performed genetic deletion of Vav2 in ApoE / mice and confirmed in vivo the contribution of Vav2 in atherosclerosis development.
To understand the basis of the indispensable role of each Vav protein in atherosclerosis, we performed unbiased whole genome RNA sequencing of macrophages deficient in each individual Vav family gene. Strikingly, we found that the deletion of each individual Vav family gene resulted in an extremely conserved transcriptomic alteration in macrophages. We found that 72.4% of upregulated genes in Vav1 knockout were present in Vav2 knockout macrophages, and more strikingly 79.9% of upregulated genes in Vav3 knockout were found in Vav2 knockout cells. In the downregulated genes, 81.3% in Vav1 knockout macrophages were present in Vav2 knockout cells, and 77.1% in Vav3 knockout cells were found in Vav2 knockout cells. Among the conserved genes found to be altered in IP using Strep-Tactin Sepharose beads followed by IB with anti-phosphotyrosine antibody (4G10). D: Following pull-down of Vav1, Vav2 and Vav3 were detected as the Vav1 interactors (left). Pull-down of Vav2 and IB of Vav1 and Vav3 showed their interaction with Vav2 (right). WT macrophages were used as a negative control. E: Peritoneal macrophages from Vav1-OST knock-in mice were subjected to IP with Strep-Tactin Sepharose beads followed by IB with anti-Vav1, anti-Vav2, and anti-Vav3 antibodies. Peritoneal macrophages from WT B6 mice were used as the control. F: Cotransfection of Vav1-Halo and Vav2-SNAP or Vav1-Halo and Vav3-SNAP. 293T cells were cotransfected with Vav1-Halo and Vav2-SNAP or Vav1-Halo and Vav3-SNAP. After 36 h of transfection, cells were labeled with a specific ligand and imaged by confocal microscope (Leica) (scale bars = 10 m). G: CD36 Internalization in RAW264.7 cells deficient for Vav1, Vav2, or Vav3. WT, Vav1 / , Vav2 / , or Vav3 / RAW264.7 cells were incubated with anti-CD36 IgA that was then cross-linked using FITC-labeled goat anti-mouse IgA  chain. After incubation at 37°C for 30 min, the noninternalized ligand was removed by acid washing, and the cells were imaged by confocal microscope (Leica) (scale bar = 2 m). H: Flow cytometry analysis of CD36 internalization in RAW264.7 cells deficient for Vav1, Vav2, or Vav3. Histograms on the left showed each replicate involving at least 4,500 cells. The MFI of CD36 was quantified and represented as mean ± SEM (n = 8). Statistics by two-tailed, unpaired Student's t-test: ****P < 0.0001. macrophages deficient in Vav1, Vav2, or Vav3, a broad range of functions was affected by deficiency in each individual Vav protein, suggesting that each Vav family member could be important for cell cycle, cell adhesion, metabolism, and phagocytosis function. Our experiments of analyzing transcriptomic alteration in each Vav-deficient protein were set to assess in a genomic view whether functional compensation could occur between different members. Interestingly, our results showed that the deletion of each single Vav gene resulted in conserved transcriptomic alteration in macrophages, indicating that each member of Vav in macrophages is functionally nonredundant. In the current study we were not able to explore the function of such differentially expressed genes in Vavdeficient macrophages.
Using the tagged Vav proteins, we further showed that Vav1, Vav2, and Vav3 formed ternary complexes in macrophages. Such experiments avoided bias in IP by antibodies; instead, identical Strep-Tactin Sepharose beads were used to pull down each Vav protein in RAW264.7 macrophages on the basis of VAV-OST knock-in cellular models. We also used primary macrophages from Vav1-OST knock-in mice and confirmed the same results that three Vav proteins form complexes. To our knowledge, our experiments are the first to present evidence that three Vav proteins form interactive complexes, which also explains why each member of the Vav protein family is indispensable in foam cell formation. In further experiments, we found that absence of each Vav protein could cause diminished CD36 internalization that could affect lipid uptake and foam cell signaling. Our findings revealed the critical role of Vav2 and further identified mechanisms underlying the parallel requirement of three Vav proteins in atherosclerosis. We did notice that a previous study showed Vav1 single-gene knockout from an ApoE-deficient background was sufficient to attenuate atherosclerosis (5), but additive effects of Vav1 and Vav3 were also reported (6). Our functional study of Vav2 was performed both in primary macrophages and RAW264.7 macrophages. We validated that Vav2 was essential for both primary murine macrophages and RAW264.7 cells. However, we did not observe additive roles of Vav proteins in RAW264.7 macrophages in terms of foam cell formation. We postulate that such differences in results might be caused by the different systems, as previous studies concerning Vav1 and Vav3 were done in primary macrophages (5). Strikingly, we found that each Vav protein was mandatory for CD36 endocytosis. It is important to note that Vav deficiencies broadly affect hematopoiesis, and Vav1/Vav3 deficiency could affect multiple other cell types in addition to macrophages, which could also explain why their double knockout resulted in less severe atherosclerosis in mice. Our current study aimed at elucidating the three Vav members in macrophages; therefore, we performed experiments involving knockout and knockin models mainly in the macrophages. We confirmed that Vav2 deficiency inhibited foam cell formation in both RAW264.7 cells and primary murine macrophages. It is of note that Vav2 deficiency in mice could result in hypertension and defects in the kidneys and sympathetic nervous system (8,9). Further studies elucidating the impact of Vav2 deficiency on other cell types in hematopoietic and nonhematopoietic systems are still valuable.
Taken together, our study confirmed the essential role of Vav2 in foam cell formation and atherosclerosis, which led us to uncover the novel mechanisms involving the pivotal CD36-JNK signaling pathway for foam cell differentiation. Such mechanisms suggest that the interdependence of Vav members could be targeted for intervention of the CD36-JNK signaling pathway and CD36 internalization. Because CD36 has been found to be a target for atherosclerosis and more recently for cancer (34), it could be invaluable to perform further studies at structural and pharmaceutical levels to disrupt the interdependence of Vav family proteins as a therapeutic intervention.