Glycoprotein A repetitions predominant (GARP) positively regulates transforming growth factor (TGF) β3 and is essential for mouse palatogenesis

Glycoprotein A repetitions predominant (GARP) (encoded by the Lrrc32 gene) plays important roles in cell-surface docking and activation of TGFβ. However, GARP's role in organ development in mammalian systems is unclear. To determine the function of GARP in vivo, we generated a GARP KO mouse model. Unexpectedly, the GARP KO mice died within 24 h after birth and exhibited defective palatogenesis without apparent abnormalities in other major organs. Furthermore, we observed decreased apoptosis and SMAD2 phosphorylation in the medial edge epithelial cells of the palatal shelf of GARP KO embryos at embryonic day 14.5 (E14.5), indicating a defect in the TGFβ signaling pathway in the GARP-null developing palates. Of note, the failure to develop the secondary palate and concurrent reduction of SMAD phosphorylation without other defects in GARP KO mice phenocopied TGFβ3 KO mice, although GARP has not been suggested previously to interact with TGFβ3. We found that GARP and TGFβ3 co-localize in medial edge epithelial cells at E14.5. In vitro studies confirmed that GARP and TGFβ3 directly interact and that GARP is indispensable for the surface expression of membrane-associated latent TGFβ3. Our findings indicate that GARP is essential for normal morphogenesis of the palate and demonstrate that GARP plays a crucial role in regulating TGFβ3 signaling during embryogenesis. In conclusion, we have uncovered a novel function of GARP in positively regulating TGFβ3 activation and function.

GARP 2 is a type 1 transmembrane protein with 20 leucinerich repeats in the extracellular domain, a single transmembrane domain, and a short 14-amino acid cytoplasmic tail (1)(2)(3)(4)(5). GARP is highly expressed by platelets, activated regulatory T cells (Tregs) (6 -9), mesenchymal stromal cells (10), hepatic stellate cells (11), and transformed tumor cells (12,13). Multiple studies have established that GARP plays a role in activating TGF␤ by mediating the surface expression and integrin-mediated activation of latent TGF␤ (1, 2, 8, 14 -18). GARP associates with latency-associated peptide to form an alternative cell-surface platform for latent TGF␤ presentation (1,2). GARP participates in the regulatory function of activated Tregs, demonstrated by impaired suppressive activity upon GARP silencing (4,6) and inefficient support by GARP KO T cells for the generation of inducible Tregs (19). Aberrant overexpression of GARP in the tumor microenvironment also promotes Treg generation and immune escape (12,13). Recently, we further identified that gp96 plays an essential role in surface GARP-TGF␤ complex expression by chaperoning GARP in the endoplasmic reticulum (5). Further, GARP has been shown to function in promoting immune tolerance and tumor progression in cancer (5,12,15,17). Thus, GARP is a key posttranslational regulator of TGF␤ biogenesis and immune tolerance.
The TGF␤ superfamily is a crucial cytokine family for both development and immunity, performing various functions in cell proliferation, differentiation, and cancer (20 -22). There are three TGF␤ isoforms. GARP reportedly binds only to TGF␤1 and TGF␤2 but not TGF␤3 (1). Genetically modified mice lacking Tgfb1 are born with autoimmune, endocrine, reproductive, vascular, and developmental abnormalities and die at 3-4 weeks of age (23). Tgfb2-null mice show perinatal mortality and multiple developmental defects, including cardiac, lung, craniofacial, limb, spinal column, eye, inner ear, and urogenital defects (24). Tgfb3 KO mice have a cleft palate caused by defects in medial edge epithelial (MEE) seam degeneration and palate fusion, leading to death within the first day after birth (25)(26)(27). Although all three TGF␤s are expressed in the palate during mouse palate development, only inhibition of TGF␤3, but not TGF␤1 or TGF␤2, results in the palate fusion defect (28,29), demonstrating the isoform-specific role of TGF␤ family members in vivo. However, the regulation of TGF␤3 biogenesis and signaling remains largely unknown.
To determine the GARP function in vivo, we have developed a reversible GARP KO mouse model using the flexible accelerated STOP TetO (FAST) knockin system (30), the GARP FAST model. The homozygous GARP KO mice died within 24 h after birth. Unexpectedly, the only developmental defect observed in KO mice was a cleft palate, which is identical to Tgfb3 KO mice (26,31). We further performed studies to establish the critical roles of GARP in binding and activating TGF␤3 and identified a novel role of GARP in TGF␤3 biogenesis and function.

Knockout of Lrrc32 causes postnatal lethality and a cleft palate
Successful disruption of GARP was accomplished by inserting a stop cassette and TetO element into the endogenous promoter region of the Lrrc32 locus. Genomic PCR and flow cytometry analysis on platelets of day 0 live pups confirmed the absence of GARP expression in the homozygous GARP FAST neonates (supplemental Fig. S1).
Immediately postpartum, the ratios of wild-type, heterozygous, and homozygous offspring (total n ϭ 94) from intercrossing between heterozygotes followed the expected Mendelian pattern (24.5% ϩ/ϩ, 23.4% Ϫ/Ϫ, 52.1% ϩ/Ϫ; supplemental Table S1), demonstrating that the loss of GARP does not cause prenatal mortality. However, no offspring with homozygous mutated alleles survived 24 h after birth. The dead GARP-null offspring showed no visible milk in their stomachs. Further gross anatomical analysis and IHC of the KO mice showed no apparent abnormalities in the major organs, including the heart, lung, spleen, and liver (data not shown).
Importantly, cleft palates were observed in all homozygous mutated pups (Fig. 1). No developmental defects were identified in heterozygous mice. Thus, we subsequently focused on analyzing palate development in GARP mutant embryos and wild-type littermate controls to specify the cause of the pathology. At embryonic day 13.5 (E13.5), when the palatal shelves form from the maxillary processes, proliferate, and undergo elevation to reach the horizontal position, the palates in mutants and controls were indistinguishable (data not shown). The differences between GARP KO and wild-type embryos became evident at E14.5, when the palatal shelves begin to fuse ( Fig. 1). At E14.5, the formation and elevation of the palatal shelf was comparable between wild-type and GARP mutant mice at the rostral region ( Fig. 1, A and E). However, at the caudal region, where palatal fusion has begun, a midline epithelial seam is seen in WT palate (Fig. 1B, arrowhead) but is absent from GARP mutants (Fig. 1F, asterisk). From E15.5 to E16.5, palatal fusion is completed throughout the entire palate in the wild type ( Fig. 1, C and D) but not observed in the mutants (Fig. 1, G and H). Instead, a large gap can be seen in GARP mutants (Fig. 1, G and H, asterisk), demonstrating a cleft palate phenotype. Together, these observations suggest that GARP is required for epithelial fusion during palatogenesis.

Palatal edge epithelial cell proliferation, apoptosis, and TGF␤ signaling in the absence of GARP
We next analyzed the possible mechanism underlying the defective palatogenesis caused by GARP deletion. First, by staining for and quantifying phospho-histone H3, we found no difference in palatal edge epithelial cell proliferation between mutant and WT embryos (supplemental Fig. S2), suggesting that the proliferation abnormality is not the cause for the cleft palate.
MEE cell apoptosis has been shown to be important for palate fusion (32). We therefore used TUNEL assays to determine whether apoptotic signaling is compromised in KO palates. Indeed, we observed a significant decrease of apoptosis at the MEE seam in the mutant palate (Fig. 2). Because GARP is known to enhance TGF␤ activation, and the TGF␤3-SMAD2/3 signaling pathway plays an essential role in palatogenesis (32)(33)(34), we next focused on discerning the possible TGF␤ signaling defect in the developing palates of KO mice. Indeed, decreased pSMAD2 was observed in E14.5 mutant palates, revealing clear defective canonical TGF␤ signaling in the absence of GARP (Fig. 3).

Expression and co-localization of GARP and TGF␤3 in palatal edge epithelial cells
The isolated phenotype observed in GARP KO mice is conspicuously similar to Tgfb3 KO mice (25,26). This suggests the possibility that GARP and TGF␤3 function together in vivo to regulate palatogenesis. If so, then GARP and TGF␤3 are expected to co-express at the MEE region. To address this pre- GARP is essential for mouse palatogenesis diction, we performed GARP IHC using a sheep anti-mouse GARP antibody and discovered that GARP was expressed at the MEE region in WT E14.5 palates but not in the corresponding KO palates (Fig. 4A). Consistently, TGF␤3 was also detected in WT MEE cells (Fig. 4B). However, as IHC cannot differentiate various forms of TGF␤3, we did not observe a clear difference in TGF␤3 levels between WT and KO palates, proving that GARP does not affect total TGF␤3 expression. Altogether, these results suggest that strategically expressed GARP in MME cells is required for TGF␤3 activation and palatogenesis.

GARP interacts with proTGF␤3 and is required for its cellsurface expression
Up to this point, the literature has suggested that GARP regulates both TGF␤1 and TGF␤2 but not TGF␤3 (1). We next revisited this question by performing extensive in vitro biochemical studies. We co-overexpressed proTGF␤3 and HAtagged mouse GARP in HEK293 cells, followed by immunoprecipitation analysis. As expected, TGF␤3 was co-pulled down with GARP (Fig. 5, A and B). In addition, we overexpressed proTGF␤3 in a mouse mammary carcinoma cell line, 67NR, which was made to express either mouse full-length GARP or the soluble extracellular domain of GARP fused with the Fc domain of IgG (GARP-Fc), which can interact with protein A/G beads (supplemental Fig. S3). Thus, using protein A/G beads, we were able to immunoprecipitate GARP-Fc with TGF␤3 from the cell lysates and the supernatants of cells that express both GARP-Fc and TGF␤3. These results demonstrate a direct interaction between the two proteins.
Because GARP is a cell-surface TGF␤ docking receptor (1, 2, 4 -6), we next investigated whether GARP is important for the surface expression of TGF␤3. Similar to what was previously done with TGF␤1 (35), we tagged human TGF␤3 with a His tag right after the furin cleavage site. The tagged TGF␤3 (His-TGF␤3) was expressed in HEK293 cells with or without GARP. We demonstrated that surface His-TGF␤3 is only detected in cells that co-express GARP (Fig. 5, C and E), despite similar levels of secreted TGF␤3 from GARP ϩ and GARP Ϫ cells (Fig.  5D). Finally, by confocal microscopic analysis, co-localization of GARP and His-TGF␤3 on the cell surface was demonstrated (Fig. 5E). Altogether, we identified TGF␤3 as a novel GARP ligand and that GARP is required for cell surface expression of TGF␤3.

Discussion
TGF␤ is a pleotropic cytokine whose activity is subject to extensive regulation to control its expression level, activation status, spatial and temporal availability, and downstream signaling. GARP is a non-signaling receptor that is primarily responsible for cell-surface docking of TGF␤ onto Tregs and platelets. Not surprisingly, the research focus on GARP has been in the area of immune regulation. However, there is evidence for the notion that GARP also functions in cell differentiation and embryonic development. As early as in 1996, Roubin et al. (36) reported that GARP transcripts are expressed in various tissues in mouse embryos. Importantly, the highest levels were detected in E13.5 and E15.5 embryos (36), with unclear significance. This study is the first to establish a role for GARP in mouse embryonic development.
Although 100% lethality was observed with GARP KO neonates, analysis of embryos and day 1 neonates did not reveal other malformations in any other organs except for a cleft palate. The isolated defect in palatogenesis in GARP KO mice is identical to the one observed in TGF␤3 mice. Thus, although GARP is broadly expressed in embryos, its most important function during embryogenesis appears to be selectively facilitating the function of TGF␤3. Indeed, we found that both GARP and TGF␤3 are co-expressed in MEE cells at E14.5. More importantly, we showed that GARP directly interacts with TGF␤3 and is responsible for cell-surface expression of TGF␤3, challenging a previous study claiming that GARP binds to the latent forms of TGF␤1 and TGF␤2 but not TGF␤3 (1). Sequencing alignment of all three TGF␤s (supplemental Fig.  S4) showed that the Cys-4 of proTGF␤1, which is important for interaction with GARP (2), is conserved among proTGF␤1, proTGF␤2, and proTGF␤3. In addition, the RGD motif is present in proTGF␤1 and proTGF␤3 but not in proTGF␤2. This suggests that TGF␤1 and TGF␤3 share similar mechanisms of activation by GARP and ␣V integrins (␣V␤6 or ␣V␤8).
It should be noted that GARP-null embryos show no clear phenotype in TGF␤1or TGF␤2-null embryos. Does this mean that GARP is more important for the function of TGF␤3 than that of TGF␤1 and TGF␤2 in vivo? The answer is not immediately clear, but it is plausible that the loss of GARP in control- ling TGF␤1 and TGF␤2 activity is compensated by alternative or GARP-independent mechanisms of TGF␤ regulation. In addition to GARP, the large TGF␤-binding protein can also be involved in binding, transporting, and activating proTGF␤s. In addition, as the concentrations of TGF␤1 and TGF␤2 in mouse blood are much higher than that of TGF␤3 (below the detection limit by conventional assays), it is possible that active TGF␤1 and TGF␤2 can be delivered into the embryo through the placenta from the heterozygote mother, compensating for the loss of GARP-dependent TGF␤1 and TGF␤2 activation. Our discovery that GARP is expressed at a high level in MEE cells also further underscores the unique function of TGF␤3 in palatogenesis (28,29). In TGF␤3 KO mice, the expression of GARP in MEEs most likely ensures abundant TGF␤1 and TGF␤2 in this location. However, for reasons that are unclear, neither of these two molecules is able to compensate for the loss of TGF␤3 (28,29,37,38).
Interestingly, a recent study observed no pups for CMV-cre/ GARP fl/fl mice and suggested an "embryonic lethal phenotype" (39). As neither embryos nor dead pups have been analyzed for defects of CMV-cre/GARP fl/fl mice/embryos, whether the two kinds of GARP KO mice have the same lethal defects need to be further determined by analyzing CMV-cre/GARP fl/fl mice. Nevertheless, the study further confirms the essential role of GARP in embryonic development.
Cleft palate in humans occurs in about one in 700 live births worldwide. Although mutations in a few genes have been suggested to contribute to cleft lip and cleft palate (40), our knowledge of genetic factors that contribute to the more common isolated cases of cleft palate is still incomplete. In addition to family genetics, a cleft palate is present in many different chromosome disorders, such as Patau syndrome. A study has reported the de novo deletion of chromosome 11q13.4-q14.3 in a boy with microcephaly, including a cleft palate, ptosis, and developmental delay (41). A 2014 review further summarized that seven cases among 32 patients with deletion of 11q13-q23.2 have a cleft palate (42). Based on this study and the knowledge that Lrrc32 locates in chromosome 11q13.5, we suggest that GARP mutation in patients may constitute a novel mechanism of cleft palate clinically.
In summary, we have uncovered, genetically, an important developmental function of GARP in regulating palatogenesis. Mechanistically, it is the first time that GARP is linked to regulate TGF␤3 function in vivo. Contrary to a previous report, we demonstrated that GARP binds directly to TGF␤3 and that it is important for TGF␤3 expression on the cell surface. Our study points to another mechanism of human craniofacial deformities. The GARP-targeted strategy will also be useful for probing other unique aspects of TGF␤3 function, such as in immune regulation, in the future.

The GARP-FAST mouse model
All animal procedures in this study were approved by the Medical University of South Carolina Institutional Animal Care and Use Committee. We developed a reversible GARP KO mouse model using the flexible accelerated STOP TetO knockin system (inGenious Targeting Laboratory) (30). The GARP-FAST mouse model was generated by inserting the promoter region of Lrrc32 with a stop cassette and a TetO-responsive element, which collectively deactivate the endogenous Lrrc32 promoter and enable the use of the tetracycline transcription system to induce/silence GARP expression.

Sample preparation and H&E staining
All samples for histological analysis were fixed in 4% formaldehyde and processed into optimal cutting temperature

GARP is essential for mouse palatogenesis
compound-embedded sections. Sections (6 m) were mounted on poly-L-lysinecoated slides. For general morphology, sections were stained with hematoxylin and eosin using standard procedures.

Apoptosis assay
Following treatment with proteinase K for 30 min at room temperature, apoptotic cells on histological slides were assayed by the TUNEL procedure using a terminal deoxynucleotidyltransferase in situ apoptosis detection kit (R&D Systems) according to the manufacturer's protocol.

Immunoblot analysis and immunoprecipitation
A human TGF␤3 expression vector (pLVE-hTGFB3-IRES-Red) was purchased from Addgene and transiently overexpressed in HEK293 cells or HEK293 cells stably expressing HAtagged mouse GARP. Two days after transfection, the cell pellets were collected. The cell pellets were lysed in radioimmune precipitation assay buffer and subjected to immunoblot analysis using antibodies against mouse GARP (Roche) and TGF␤3 (R&D Systems). Pulldown of GARP-HA was performed using an anti-HA antibody (Proteintech) and protein A/G beads (Bio-Rad) following the the manufacturer's protocol. Briefly, the cell lysates were incubated with the beads and antibody overnight at 4°C. The immunoprecipitated protein was released by treating the beads with 2ϫ SDS Laemmli buffer for immunoblot analysis.

Surface detection of TGF␤3 expression
A nine-histidine (His) repeat was inserted between Arg-300 and Ala-301 in human TGF␤3. The tagged TGF␤3 (His-TGF␤3) was cloned into a MigR1 retroviral vector with GFP as a reporter. His-TGF␤3 was expressed in HEK293 cells or 293-GARP cells, followed by surface staining with an anti-His antibody (1:100, Proteintech) and a secondary antimouse-allophycocyanin antibody. TGF␤3 surface expression was then detected by a flow cytometer (BD FACSVerse). For microscopy analysis, the above cells were fixed with 4% formaldehyde for 10 min and permeabilized with cold methanol for 5 min, followed by anti-His and GARP antibody staining. The images were taken by an Olympus FV10i confocal microscope.

ELISA
The TGF␤3 ELISA was performed using an ELISA kit (R&D Systems, DY243) according to the manufacturer's instruction.

Statistical analysis
Results are expressed as mean Ϯ S.D. Comparisons among groups were made by Student's t test. A p value of 0.05 or less was considered statistically significant.