A vascular smooth muscle cell X-box binding protein 1 and transglutaminase 2 regulatory circuit limits neointimal hyperplasia

Neointimal hyperplasia, stimulated by injury and certain vascular diseases, promotes artery obstruction and tissue ischemia. In vascular smooth muscle cell (VSMCs), multiple modulators of protein handling machinery regulate intimal hyperplasia. These include elements of the VSMC unfolded protein response to endoplasmic reticulum stress (UPRER), and transglutaminase 2 (TG2), which catalyzes post-translational protein modification. Previous results for deficiency of UPRER-specific mediator XBP1, and of TG2, have been significant, but in multiple instances contradictory, for effects on cultured VSMC function, and, using multiple models, for neointimal hyperplasia in vivo. Here, we engineered VSMC-specific deficiency of XBP1, and studied cultured VSMCs, and neointimal hyperplasia in response to carotid artery ligation in vivo. Intimal area almost doubled in Xbp1fl/fl SM22α-CRE+ mice 21 days post-ligation. Cultured murine Xbp1 deficient VSMCs migrated more in response to platelet derived growth factor (PDGF) than control VSMCs, and had an increased level of inositol-requiring enzyme 1α (Ire1α), a PDGF receptor-binding UPRER transmembrane endonuclease whose substrates include XBP1. Cultured XBP1-deficient VSMCs demonstrated decreased levels of TG2 protein, in association with increased TG2 polyubiquitination, but with increased TG transamidation catalytic activity. Moreover, IRE1α, and TG2-specific transamidation cross-links were increased in carotid artery neointima in Xbp1fl/fl SM22α-CRE+ mice. Cultured TG2-deficient VSMCs had decreased XBP1 associated with increased IRE1α, and increased migration in response to PDGF. Neointimal hyperplasia also was significantly increased in Tgm2fl/fl SM22α-CRE+ mice at 21 days after carotid ligation. In conclusion, a VSMC regulatory circuit between XBP1 and TG2 limits neointimal hyperplasia in response to carotid ligation.

Several processes pertinent to intimal hyperplasia are promoted by activation of the transmembrane endonuclease inositol requiring enzyme-1α (IRE1α) pathway of the UPR ER [10,14,15]. These include activation of inflammatory signaling via transmembrane IRE1αinduced phosphorylation of IκB kinase (IKK) and c-Jun N-terminal kinase (JNK) [10,14,15], and direct interaction between IRE1 and PDGFRβ that promotes downstream responses to PDGF [1]. IRE1 plays a mediating role in cultured VSMC proliferation in response to PDGF, as shown by IRE1 siRNA knockdown [1,2]. Injured arterial intima demonstrate the UPR ERspecific process of alternative splicing and activation of mRNA of the transcription factor XBP1 by generating the transcriptionally active form of XBP1, termed XBP1s [1]. This process is executed via upstream endonuclease activity of IRE1 [10,14,15]. However, results of 2 prior studies have been contradictory for effects on cultured VSMCs and neointimal hyperplasia of XBP1 [1,2].
The ubiquitously expressed enzyme TG2 has been implicated in intimal hyperplasia [3,4]. TG2 not only cross-links proteins by transamidation, but also modulates the autophagy arm of the proteostasis network, and functions as a protein disulfide isomerase and as a GTPase and GTP-modulated signal transducer [18,19]. Effects of TG2 to promote arterial diseases include mediation of PDGF and β-catenin signaling in VSMCs, and promotion of VSMC proliferation, migration, and phenotypic switching [3,4,20]. Conversely, the capacity of TG2 to inhibit certain cytokines and MMPs, and to interact with fibronectin and transforming growth factorβ, can promotes stabilization of the arterial extracellular matrix [21]. Global knockout of TG2 markedly limited inhibited neointimal hyperplasia in response to carotid artery loop injury (3), but either inhibited or failed to inhibit neointimal hyperplasia in response to carotid artery ligation in two different studies [3,4].
Since both XBP1 and TG2 regulate a variety of cell homeostasis and injury response processes, this study investigated the relationship between XBP1 and TG2 in VSMCs. To simplify our approach to study VSMC responses, we focused on effects of VSMC-specific deficiency of XBP1 and TG2 in neointimal hyperplasia in response to carotid artery ligation, a model that does not involve endothelial denudation [22]. In doing so, we assessed for cross-talk between XBP1 and TG2 in VSMCs. Our results reveal a VSMC regulatory circuit between XBP1 and TG2 that limits neointimal hyperplasia in response to carotid artery ligation.

Carotid artery ligation
Eight weeks old mice were anesthetized by intraperitoneal injection of a solution of xylazine (5 mg/kg body weight, AnaSed; Lloyd Laboratories), ketamine (80 mg/kg body weight) and acepromazine maleate (1.2 mg/kg (Boehringer Ingelheim, St. Joseph, MO). We followed a protocol previously described in detail [7]. In brief, the left common carotid artery was ligated near its bifurcation with the use of 6-0 silk, with right carotid artery as uninjured control. At 14 and 21 days after injury, animals were anesthetized and perfused with 5 ml PBS, followed by 5 ml of 4% paraformaldehyde for 3 min. Left and right carotid arteries were excised, further fixed for 16 hours and embedded transversely in paraffin. The entire length of and lesion apex identified. For morphometric analyses, images of Verhoeff-van Gieson elastin-stained cross-sections of left and right carotid arteries (6 micron thickness) were analyzed in a blinded manner using AxionVision microscope (Carl Zeiss, AG Oberkochen, DE), and intimal and medial area calculated [7].

VSMC isolation and culture
To isolate primary mouse VSMCs, aortae from 8 weeks old mice were isolated (from aortic root to bifurcation of the renal arteries) and media was digested, as described (7). VSMC differentiation was verified by staining for smooth muscle -actin (mouse monoclonal antibody; cat#ab7817, Abcam) by immunofluorescence (>95% of cells stained positive for smooth muscle -actin). VSMCs were cultured in complete DMEM medium at 37˚C, in 5% CO2. Mouse VSMC line Movas-1 cells (ATCC CRL-2797, Manassas, VA), were transfected with XBP1 siRNA (Santa Cruz Biotechnology, sc-38628) where indicated, and collected at 24 h and 48 h post transfection and lysed in RIPA buffer and cell proteins immunoprecipitated using Dynabeads (Thermo Fisher).

In Situ Tissue Transglutaminase (tTG) activity assay of TG2
For in situ TG2 activity measurements, XBP1Δ SMCs were preincubated with 5-(biotinamido) pentylamine (Thermo Fisher), a biotinylated polyamine TG2 substrate that the enzyme can use to post-translationally modify proteins, followed by determination of its incorporation into proteins. [26] We measured absorbance at 492 nm on microplate spectrophotometer. As a negative control, we used Tgm2Δ VSMCs treated under identical conditions, and Xbp1WT VSMCs served as the positive controls for normal levels of TG2 activity.

Transwell migration assay
To determine cell proliferation, we used the Invitrogen non-radioactive Cell Proliferation Assay kit, per manufacturer instructions. VSMC migration was assayed, as previously described [7]. In brief, we used a Transwell plate (Cat #3422, 8 μm pore size, Corning, Sunnyvale, CA), with VSMCs plated at 1.0 x 10 6 cells/ml in DMEM supplemented with 10% heatinactivated fetal calf serum [7]. Numbers of cells migrated per well, were determined after 48 h treatment with or without PDGF, as described [7]. Images were analyzed using a microscope with digital camera (EVOS).

Morphometry
Morphometric analysis was carried out on carotid arteries harvested at 14 and 21 days after ligation. All animals were perfused and fixed under physiological pressure. Images of arteries were analyzed with Axion Vision Software (Carl Zeiss AG, Oberkochen, DE). Additionally, under the assumption that the structures were circular, measurements were used to calculate intimal and medial area.

Statistical analyses
All values are expressed as mean ± SD. Statistical differences between WT and mutant mice or cells were analyzed by a Student's t-test. Morphometric analysis of carotid artery was done by 1-way ANOVA. Differences were considered statistically significant at values of p <0.05. Data for cell culture experiments represent three experiments in triplicates using separate cultures.

Increased intimal thickening after carotid artery ligation in mice with VSMC-specific XBP1 deficiency
We generated Xbp1Δ mice (Xbp1 fl/fl SM22α-Cre), in which Xbp1 exon 2 was efficiently deleted specifically in aortic VSMCs. In parallel, we similarly generated Tgm2Δ mice (Tgm2 fl/fl SM22α-Cre). Unlike the case for isolated VSMCs, in the Xbp1Δ and Tgm2Δ mouse whole tissues studied, there was no gross attenuation of XBP1 or TG2, respectively (S1 Fig).
Neointimal hyperplasia developed in carotid arteries, in response to ligation, in both Xbp1WT and Xbp1Δ mice. This phenotypic response was significantly increased in Xbp1Δ mice ( Fig 1A). The lumen of the ligated carotid arteries was larger in Xbp1WT mice than in Xbp1Δ mice (Fig 1A, whereas intima-media ratio was significantly increased in Xbp1Δ mice compared to Xbp1WT animals (Fig 1B)). In isolated, cultured aortic VSMCs from Xbp1Δ mice treated with PDGF, there was significantly increased migration but not proliferation, compared to control cells (Fig 1C).

XBP1 modulates TG2 and IRE1α in VSMCs
Isolated VSMCs from cultured XBP1 deficient mouse aortae demonstrated decreased TG2 (Fig 2A). Conversely, TG2-deficient VSMCs demonstrated decrease in XBP1u (Fig 2A). Transfection of XBP1s into XBP1-deficient VSMCs significantly increased TG2 protein (Fig 2C and  inset). Silencing of either XBP1 or TGM2 expression by siRNA in MOVAS-1 cells affected the protein expression level of the other protein ( Fig 2D). Paradoxically, XBP1 deficiency in vitro was associated with increased activity of the TG2 present, as assessed by tTG assay [29] (Fig  2E). In response to carotid ligation, the carotid artery sections, analyzed by IHC using an antibody to Nε-(γ-L-glutamyl)-lysine-isopeptide bond, showed increase of the selective TG2 crosslinking marker in Xbp1Δ carotid arteries compared to wild type (Fig 2F).
IRE1α has multiple binding partners, and many effects on cell differentiation and function independent of action on XBP1 [10,14,15,30,31]. Hence, we assessed IRE1α levels in vitro and in situ. In VSMCs extracted from aortae of the conditional knockout mice, we observed increased IRE1α levels in Xbp1Δ, and also, though to a lesser degree, more IRE1α in Tgm2Δ cells (Fig 3B). Similarly, IRE1α was increased in MOVAS-1 cells in which Xbp1 was knocked down by siRNA (Fig 3C). Moreover, we observed markedly increased IRE1α immune staining in neointima in situ in carotid artery sections of XBP1Δ mice (Fig 3D). The mRNA levels for Tgm2 and Xbp1 did not significantly decrease with deficiency of the other protein in cultured VSMCs (S2 Fig). Thus, we next tested for effects of XBP1 deficiency on the stability of TG2. demonstrated substantial plaque development after 14 and 21 days post ligation. (B) Intima-media ratio was compared for Xbp1WT vs Xbp1Δ at 14 days post ligation (p<0.0001 n = 8 Xbp1WT; n = 12 Xbp1Δ), and at 21 days, where intimal cell densities were increased in both Xbp1WT and Xbp1Δ (p<0.0001, unpaired t-test, n = 10 Xbp1WT; n = 11 Xbp1Δ;). (C) Xbp1Δ (Xbp1 f/f ;SM22α-Cre) and Xbp1WT (Xbp1 f/f ) aortic VSMCs were treated with PDGF (10 ng/ml) for 18h. The numbers of cells migrated, and proliferated, were studied as described in the Methods. XBP1 deficiency increased VSMC migration (assessed as the increase in migration in response to PDGF relative to no PDGF) in vitro (p<0.0001; n = 4). XBP1 deficiency did not significantly increase proliferation in response to PDGF. Data represent mean ± SD of 4 independent experiments. Results were analyzed by 1-way ANOVA, followed by Bonferroni correction).

XBP1 affects TG2 stability in VSMCs
TG2 levels significantly increased in Xbp1Δ VSMCs treated over time with the proteasome inhibitor MG132 (Fig 4A). Since XBP1 deficiency appeared to promote increased proteasomal TG2 degradation, we next studied TG2 SUMOylation. We also assessed if TG2 cellular stability was potentially subject to dynamics between TG2 SUMOylation and ubiquitylation [32]. To do so, we generated stable HeLa cell lines expressing distinct SUMO molecules (SUMO1, SUMO2 and SUMO3) [33,34], and we transfected these cells with TG2-HA expression plasmid. We first confirmed [32] that TG2 was SUMOylated by SUMO1, (Fig 4B), and predominantly modified by SUMO2 and to lesser extent by SUMO3. The SUMOylation was compared to control RanGAP SUMOylation (S3 Fig). To test whether TG2 ubiquitylation/SUMOylation dynamics can be affected by XBP1 deficiency, MOVAS-1 cells were transfected with XBP1 siRNA, followed by immunoprecipitation at different post transfection times. TG2 demonstrated reduced SUMOylation (SUMO1) in response to Xbp1 knockdown, and, conversely, there was increased TG2 K48-linked ubiquitylation (Fig 4C), which is known to promote proteasomal degradation [35].

Increased intimal thickening after carotid artery ligation in TGM2Δ mice
Neointimal hyperplasia and associated lumen occlusion developed in the ligated carotid arteries of both Tgm2WT and Tgm2Δ mice, and these responses were significantly increased in Tgm2Δ mice (Fig 5A and 5B). Last, isolated Tgm2Δ aortic VSMCs demonstrated both increased migration and proliferation in response to PDGF compared to Tgm2WT (Fig 5C).

Discussion
VSMC responses to stress are modulated by the UPR, other proteostasis pathways, and by the multifunctional molecule TG2, a component of the protein handling machinery [1][2][3][4]8]. This study identified a regulatory circuit in VSMCs via cross-talk between TG2 and XBP1, a transcription factor whose activation by IRE1 endonuclease action is a UPR-specific event [11][12][13][14]. Specifically, deficiency of XBP1 or TG2 promoted decrease in the level of the other respective protein in VSMCs. These changes were not associated with corresponding alteration of XBP1 or TGM2 mRNA levels due to deficiency of TG2 or XBP1, respectively. Instead, we demonstrated an increased capacity of XBP1-deficient cells to promote TG2 proteasomal degradation by increase in K48 polyubiquitylation relative to SUMOylation. Protein SUMOylation has the capacity to promote protein stabilization, in part by blocking ubiquitination at lysine residues, and partly because SUMOylated proteins have a distinct overall surface charge distribution compared to ubiquitin [33,34]. XBP1 was previously recognized to regulate ER-associated protein degradation and to have major effects on the transcriptional profile of the proteostasis network, and on successful UPR resolution [12,13]. These new findings on regulation of TG2 by  suggest additional means by which XBP1 can modulate cell fate and inflammatory responses following activation of the IRE1 arm of the UPR [10][11][12][36][37][38].
VSMCs, as well as endothelial, vascular precursor, and inflammatory cells, and platelets have been implicated in neointimal hyperplasia [1][2][3][4][5][6][7][8]. Arterial wire injury, employed in several of these studies, is designed to resemble post-revascularization injury in stenotic human arteries [22]. However, endothelial injury, and substantial roles of inflammatory cell influx and activation in such murine models of neointimal hyperplasia, render arterial wire injury mechanistically and cellularly more complex than carotid artery ligation [22]. The carotid ligation model is triggered by marked reduction of vessel diameter and attenuation of blood flow, with consequent effects on arterial wall shear stress stimulating rapid VSMC-enriched neointima formation [22]. Our study focused squarely on VSMC biology via unique use of newly generated SMC-specific knockout mice for Xbp1 and Tgm2, and we examined only neointimal hyperplasia following carotid artery ligation.
We demonstrated that both SMC-specific XBP1 and TG2 deficiency increased neointimal hyperplasia in response to carotid artery ligation in vivo. In complementary in vitro studies here, VSMCs isolated from aortae of both these conditional knockout animals were observed to migrate more in response to PDGF. Furthermore, TG2-deficient VSMCs proliferated more in response to PDGF than did VSMCs from control animals. Here, as in prior studies involving XBP1 and TG2 [1][2][3][4], results for cultured VSMC responses to PDGF, and for neointimal hyperplasia were largely aligned. However, multiple differences existed between the conclusions of prior studies VSMC XBP1 function in vitro and in neointimal hyperplasia in vivo [1,2], now magnified by results of our study. In one prior report, global Xbp1 haploinsufficiency increased neointimal hyperplasia in response to femoral artery wire injury, and XBP1 siRNA augmented cultured human coronary VSMC migration and proliferation in response to PDGF [2]. In contrast, SM22α-targeted SMC-specific Xbp1 deficiency increased neointimal hyperplasia in response to femoral artery wire injury in another prior report [1]. Overexpression XBP1s increased human VSMC migration, and conversely, cell migration out of aortic ring explants was decreased by Xbp1 knockout in one study [2]. However, XBP1 deficiency in cultured VSMCs failed to significantly alter PDGF-induced migration in the same report [2].
Differences in results from small animal neointimal hyperplasia models partly reflect distinctions in experimental model behavior between species, but also expose differences between pathogeneses of models [22]. It is possible that factors in discrepancies between findings on XBP1 in VSMC behavior and neointimal hyperplasia include differences in degree of experimental depression of XBP1 levels, and impact of distinct protocols and reagents and agonists, such as using VEGF or PDGF as agonist. Also involved may be beneficial effects downstream of XBP1 in stressed cells, such as modulation of apoptosis and autophagy [11][12][13], successful resolution of proteotoxic cell stress, and prevention of oxidative stress by modulating reactive oxygen species generation and catalase expression [39]. Conversely, XBP1s promotes activation of VSMCs via PI3K/Akt signaling and the capacity to down-regulate calponin h1 through mir-1274B-mediated mRNA degradation [1]. Increases in VSMC PDGF and PDGF receptor levels in XBP1-deficient VSMCs [2], and in IRE1α levels as observed here, also merit consideration. IRE1α endonuclease activity is critical to induce XBP1 activation by generating alternatively spliced XBP1s transcript [12,13], but IRE1α also can enhance downstream effects of PDGF receptor signaling [10,14,15]. Moreover, IRE1α endonuclease and kinase activities have multiple additional consequences on cell differentiation and function independent of action on XBP1 [10,14,15]. These include activation of inflammatory signaling and cytokine expression via transmembrane IRE1α -induced phosphorylation of IKK and JNK [12][13][14][15], and direct interaction between IRE1 and PDGFRβ that promotes downstream responses to PDGF [1].
TG2 not only has been observed to increase VSMC migration and proliferation in vitro [3,4], but also to promote inward remodeling of injured arteries in vivo [40]. Moreover, previous work observed significant decreases, in global TG2 knockout mice, of neointima formation in both PDGF-treated heart slices and in carotid arteries following ligation [3,4]. Significantly, prior study reported attenuated neointimal hyperplasia following carotid artery loop injury in global TG2 knockout mice, though there no significant difference between global TG2 knockout and wild type mice in neointimal hyperplasia following carotid artery ligation [3]. As such, our finding that SMC-specific TG2 deficiency increased neointimal hyperplasia, in response to carotid artery ligation, could be viewed as paradoxical. However, we did observe that XBP1 deficiency was associated with increased VSMC TG activity in vitro, and neointimal TG2 protein cross-links in response to carotid artery ligation in vivo. Hence, our results suggest a model in which increased transamidation activity of TG2, and potentially reciprocal decrease in GTPase-related TG2 activity, as well as decrease in TG2 levels and nonenzymatic activities, could be among the mechanisms whereby stressed XBP1-deficient VSMCs can promote neointimal hyperplasia.
Limitations of this study include that it did not specifically examine effects of VSMC XBP1 and TG2 deficiencies on the extracellular matrix synthesis and composition, or on cell viability, and on the global proteostasis network. Tgm2 deficient VSMCs had a proliferative response to PDGF-BB (10 ng/ml). However, fleshing out individual factors mediating the migration response in TG2-deficient VSMCs was beyond the scope of this study. It will be of interest to test for potential contributory effects of mitochondrial function regulation by the IRE-XBP1 signaling axis [41] in arterial VSMC responses to vascular injury. It remains to be determined if the VSMC XBP1 and TG2 regulatory loop defined here has generalized impact across intimal hyperplasia disease models, including restenosis after direct endothelial denudation injury.
We conclude that cross-talk between XBP1 and TG2 in VSMCs is substantially involved in limiting the extent of neointimal hyperplasia in carotid arteries injured by ligation. Our findings identify how dysregulated UPR and protein handling machinery responses can not only be intertwined but also amplify adverse adaptive responses to tissue injury. The results provide further support for the potential of systemically and locally controlling the UPR and TG2 to productively intervene in the phenotype of tissue responses to injury. For affinity purification of SUMO proteins, we subjected eluates from Nickel-NTA beads to SDS/PAGE-Western blotting, using TG2 and RanGAP as controls for protein SUMOylation. The figure shows anti-TG2, RanGAP1 and input immunoblots of nickel affinity purifications from HeLa cells. The high molecular weight species were detected by TG2 antibody, with strong signal above~75kD representing TG2 eluted from Ni-NTA columns. Endogenous TG2 was covalently conjugated to 6His-SUMO1, 2 and 3, with eluates compared from stable lines expressing SUMO1,2, and 3. RanGAP antibody was able to detect endogenous protein which is modified by SUMO1, and to lesser extent by SUMO3, and was used as control for the procedure. Non-specific bands are indicated by asterisks. (TIF) S1 Dataset. Sminimaldataset.xlsx: The minimal data set, for the manuscript results presented here, is attached. Funding acquisition: Robert Terkeltaub. hyperplasia at 14 and 21 days after vessel ligation. (B) Intima-media ratio was compared for Tgm2WT vs Tgm2Δ after 14 days after ligation surgery (p<0.0001 n = 9 WT; n = 9 Tgm2 f/f ). At 21 days after ligation, cell densities in the intimal lesions in Tgm2WT and Tgm2Δ were compared (p<0.0001, unpaired t-test, n = 9 TGM2WT; n = 12 TGM2Δ;). (C) Tgm2Δ and Tgm2 WT VSMCs were treated with PDGF (10 ng/ml) for 18 to assess migration and proliferation, as described above. VSMC migration and proliferation were increased in VSMCs in vitro (p<0.0001; n = 9). Data represent mean ± SD of 4 independent experiments. Results were analyzed by 1-way ANOVA, followed by Bonferroni correction).