Increased TGFβ1 and SMAD3 Contribute to Age-Related Aortic Valve Calcification

Aims Calcific aortic valve disease (CAVD) is a progressive heart disease that is particularly prevalent in elderly patients. The current treatment of CAVD is surgical valve replacement, but this is not a permanent solution, and it is very challenging for elderly patients. Thus, a pharmacological intervention for CAVD may be beneficial. In this study, we intended to rescue aortic valve (AV) calcification through inhibition of TGFβ1 and SMAD3 signaling pathways. Methods and Results The klotho gene, which was discovered as an aging-suppressor gene, has been observed to play a crucial role in AV calcification. The klotho knockout (Kl–/–) mice have shorter life span (8–12 weeks) and develop severe AV calcification. Here, we showed that increased TGFβ1 and TGFβ-dependent SMAD3 signaling were associated with AV calcification in Kl–/– mice. Next, we generated Tgfb1- and Smad3-haploinsufficient Kl–/– mice to determine the contribution of TGFβ1 and SMAD3 to the AV calcification in Kl–/– mice. The histological and morphometric evaluation suggested a significant reduction of AV calcification in Kl–/–; Tgfb1± mice compared to Kl–/– mice. Smad3 heterozygous deletion was observed to be more potent in reducing AV calcification in Kl–/– mice compared to the Kl–/–; Tgfb1± mice. We observed significant inhibition of Tgfb1, Pai1, Bmp2, Alk2, Spp1, and Runx2 mRNA expression in Kl–/–; Tgfb1± and Kl–/–; Smad3± mice compared to Kl–/– mice. Western blot analysis confirmed that the inhibition of TGFβ canonical and non-canonical signaling pathways were associated with the rescue of AV calcification of both Kl–/–; Tgfb1± and Kl–/–; Smad3± mice. Conclusion Overall, inhibition of the TGFβ1-dependent SMAD3 signaling pathway significantly blocks the development of AV calcification in Kl–/– mice. This information is useful in understanding the signaling mechanisms involved in CAVD.


INTRODUCTION
Calcific aortic valve disease (CAVD) is a progressive heart disease in which aortic valve sclerosis progresses to aortic valve stenosis with severe calcification and impaired leaflet function (1)(2)(3)(4)(5). The aortic valve (AV) calcification affects 25% of the population over 65 years of age and about 50% of those over 85 years (2). Surgical valve replacement is the most effective treatment for valvular heart disease (6). Old age and chronic kidney disease (CKD) are important risk factors for CAVD (1). The serum levels of klotho (KL) decrease in CKD patients with heart valve calcification and therefore klotho is considered an independent risk factor for CAVD in CKD (7). The klotho gene (Kl) is identified as an anti-aging gene in mice that prolongs the life span (8). Since its serendipitous discovery, klotho has drawn significant attention owing to its potential role in aging, chronic kidney disease (CKD), and several cardiovascular diseases (9,10). Earlier investigations indicated that klotho expresses predominantly in the distal tubular epithelial cells of the kidney (11,12). Another study has confirmed klotho expression in human artery and cardiac myocytes (10). Functions of klotho include regulation of energy metabolism, anti-inflammatory and anti-oxidative effects, modulation of ion transport, and regulation of mineral metabolism (13). Several reports also endorsed the protective role of klotho in several organs as well as reversal of disease mechanisms including cardiovascular disease (14). The Kl encodes a single-pass transmembrane protein of 135 kDa. The klotho protein is clipped on the cell surface by membrane-anchored proteases and the entire extracellular domain (∼130 kDa) is released into the systemic circulation (8,15,16). Thus, klotho protein exists in two forms: membranebound klotho and secreted klotho (10,17). More probably, membrane klotho solely serves as a co-receptor for the binding to fibroblast growth factor 23 (FGF23) (18). FGF23 is a bonederived hormone that acts on kidney to stimulate phosphate elimination into urine and inhibit vitamin D synthesis, thereby modulating negative phosphate balance (19,20). One significant characteristic of FGF23 is that it has very poor affinity to FGF receptors (21,22). Binding of membrane klotho generates a constitutive binary complex with FGF receptors (FGFRs) which creates a de novo high-affinity binding site for FGF23 (22). Thus, FGF23 requires membrane klotho to bind to its cognate FGF receptors and exert its biological activity (22). The klotho (Kl −/− ) and/or Fgf23 (Fgf23 −/− ) knockout mice exhibit many key aspects associated with human CAVD including, premature aging, kidney disease, increased serum phosphate levels (i.e., hyperphosphatemia), and increased osteogenic gene expression (7,23,24). The Kl −/− mice die by 12 weeks of age due to multiple age-related complications. The Kl −/− mice develop calcific nodules in the AV hinge and aortic annulus, but they do not show significant AV leaflet thickening, extracellular matrix (ECM) disorganization, or inflammation (25). In addition, the Kl −/− mice also develop ectopic calcification of aorta and kidneys (26). It has been reported that valve interstitial cells (VIC) in the calcified AV hinge in Kl −/− mice become activated and express several pro-calcific markers, including COX2, RUNX2, osteopontin (OPN1/SPP1), and alkaline phosphatase (ALP). Recently, Kl −/− mice have been successfully used in preclinical testing of potential new drugs for AV calcification (27). Since CKD is associated with klotho and human CAVD (28) and the AV calcification in Kl −/− mice resembles human CAVD (29), the Kl −/− mice are useful mouse models for investigating the mechanisms involved in pathogenesis of CAVD (29,30).
Surgical specimens of the aortic valve (AV) obtained from older patients with CAVD have increased levels of transforming growth factor beta1 (TGFβ1) (31)(32)(33)(34)(35)(36)(37). Elevated TGFβ1 levels are frequently observed during vascular calcification and CKD (38)(39)(40) and contribute to the progression of calcification of cardiovascular tissues (41,42). Both AV calcification and vascular calcification are regarded as the crucial risk factors in CKD and are associated with cardiovascular and all-cause morbidity and mortality (43) (44). The AV calcification is an active process with some parallels to physiological bone formation (3,45). A critical role is attributed to vascular smooth muscle cells (VSMC) and VIC, which can differentiate and convert into myofibroblasts, osteoblast, and chondroblast-like phenotypes (46,47). Osteoblasts or chondrogenic VSMCs/VICs actively promote valvular or vascular tissue mineralization (25,48). Although the intracellular signaling pathways that control this trans-differentiation into myofibroblasts, osteoblasts, or chondrocytes are yet not understood completely (49). Human myxomatous valve disease (MMVD) is characterized by increased VIC proliferation, leaflet thickening, increased proteoglycan expression, and ECM remodeling (50). Cell proliferation is unaffected and αSMA expression is reduced in calcified AV of Kl −/− mice (25), suggesting that VIC activation or MMVD is not present in Kl −/− mice. TGFβ1 plays a crucial role in both myofibroblasts and osteogenic trans-differentiation of the VSMCs/VICs and induces cellular senescence through the upregulation of plasminogen activator inhibitor (Pai1) (38,47). Moreover, TGFβ1-dependent osteoinductive signaling involves altered expression of the chondrogenic transcription factor SRY-Box 9 (SOX9), which is involved in AV calcification (51)(52)(53)(54).
Increased levels of TGFβ ligands can contribute to cell degeneration, inflammation, metabolic malfunction, tissue fibrosis, and calcification. While the roles of TGFβ signaling depend on cellular contexts, age-related changes are also a potential context, and therefore the relationship between TGFβ signaling and klotho, which is involved in cellular senescence and aging-related diseases such as CAVD, warrants serious attention. There are many investigations that showed that Kl −/− mice can induce calcification in aortic valves (25,27,55,56). The AV of Kl ± mice do not show any calcification, although a high-fat diet resulted in collagen-I deposition and fibrosis of the aortic valve cusps on the aortic side (57). The signaling mechanisms of AV calcification in klotho deficient mice are still unclear and require further investigations. In this work, we investigated the impact of reduced TGFβ signaling on the AV calcification in Kl −/− mice. We observed that partial inhibition of TGFβ signaling through haploinsufficiency of Tgfb1 and Smad3 improves the pathological condition of the AV calcification in klotho-deficient mice.

Ethics Statement
All animal procedures were performed according to the Guidelines for the Care and Use of Laboratory Animals published by the National Institutes of Health and were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of South Carolina. Mice were euthanized by an overdose of isoflurane in a sealed container as approved by the IACUC.

Tissue Collection and Processing for Histology
Wild-type controls Kl −/− , Kl −/− ; Tgfb1 ± , and Kl −/− ; Smad3 ± mice were sacrificed, and aortic valve tissues were collected for histological, immunohistochemical, morphometric, and molecular analyses. The whole heart of each mouse was perfused with 1XPBS and later fixed in 4% paraformaldehyde in PBS

Histology
All histological staining procedures were performed according to the protocol provided by the manufacturers. Before various histological staining procedures, the tissues were de-paraffinized and rehydrated. Hematoxylin and eosin staining (H&E staining) was carried out using Harris' hematoxylin (Cat # HS-400) and Eosin (Sigma, Cat # HT110380). Tissue sections were kept in alizarin red stain (American MasterTech Scientific, Lodi, CA, United States) for 5 min for detection of calcium deposits. For alizarin red staining, tissue sections were fixed and dehydrated with -20 • C cold acetone (Thermo Fisher Scientific, Waltham, MA, United States) and cleared with an acetone/xylene solution (50% acetone and 50% Safe Clear II Xylene substitute, both purchased from Thermo Fisher Scientific, Waltham, MA, United States). Alcian blue stain kit (catalog # KTABP2.5) was used to detect proteoglycans. The Verhoeff 's Elastin staining kit (Cat# KTVEL) was used for elastic fiber (black color) and collagens (red color) staining. Collagens were also detected by using the Manson's Trichrome 2,000 stain kit (cat # KTMTR2PT). All kits for detecting proteoglycans, collagens, and elastin fibers were purchased from the American MasterTech Scientific (Lodi, CA, United States). At the end of each staining, the tissue slides were mounted with a permanent mounting medium (Vector Laboratories, Burlingame, CA, United States). All tissue sections were subsequently visualized and photographed in low and high magnifications under bright field optics on the Nikon Optiphot-2 (equipped with AxioCam MRc Camera) and EVOS TM FL Auto Imaging System (Thermo Fisher Scientific, Inc., Grand Island, NY, United States). Morphometric quantification of signal intensity was done on multiple serial sections from each animal per group by using the NIH-Image J (Fiji) software.

Immunohistochemistry
Paraformaldehyde-fixed paraffin-embedded serial sections representing annulus, aortic leaflets and hinge, and a portion of the sinus of Valsalva were used for immunohistochemistry. Sections were de-paraffined and hydrated with two changes in 1X PBS and one change in deionized water (5 min each). Heat-mediated antigen retrieval was performed by dipping the slides in a mildly boiling 1X citric acid buffer (catalog no. S1700; Agilent Dako, Santa Clara, CA, United States) for 10 min in a microwave, cooled to room temperature, and rinsed in PBS. Endogenous peroxidases were blocked with freshly prepared 0.5% H 2 O 2 /methanol for 30 min, followed by non-specific epitope blocking with 5% goat serum/0.1% Tween/0.02% sodium azide in PBS for 20 min. RNA Isolation, cDNA Synthesis, and Quantitative PCR The AV tissue was dissected manually from wild-type, Kl −/− , Kl −/− ; Tgfb1 ± , and Kl −/− ; Smad3 ± mice. The AV tissue contained aortic roots including annulus, aortic leaflets and hinge, and a portion of the sinus of Valsalva. Five individual biological samples from each group were analyzed for gene expression study. Total RNA was isolated from the AV tissue of wild-type and Kl −/− , Kl −/− ; Tgfb1 ± , and Kl −/− ; Smad3 ± mice using Trizol (Invitrogen/ThermoFisher, Grand Island, NY, United States) and miRNeasy micro kit (Qiagen, Germantown, MD, United States) according to the manufacturer's protocols. cDNA was generated from 500 ng total RNA using an RT-PCR kit according to the instructions provided by the manufacturer (Bio-Rad Laboratories, Inc.). cDNA was then diluted 10 times and later subjected to quantitative PCR amplification (Bio-Rad CFX) using pre-validated gene-specific primers procured from the vendor (Bio-Rad Laboratories, Inc., Hercules, CA, United States) ( Table 2). Approximately, 10 ng of cDNA was used for each 20 µl qPCR reaction. Following qPCR analyses, the cycle count threshold (Ct) was normalized to species-specific housekeeping genes (B2m; purchased from Bio-Rad, Inc.) and the 2ˆ− Ct values were determined and graphically presented. Statistically significant differences in gene expression levels were determined using Student's t-test, indicated in the figure legends, on at least three or more independent experiments with p < 0.05 considered significant.

Western Blot Analysis
Three independent "pooled" samples of the AV tissue from each group of mice were used for western blotting. The AV tissues in each sample were "pooled" from two individual mice for each genotype (wild-type, Kl −/− , Kl −/− ; Tgfb1 ± , and Kl −/− ; Smad3 ± ). Thus, each independent pooled sample consists of two biological replicates (n = 6 per genotype). The AV tissue contained aortic roots including annulus, aortic leaflets and hinge, and a portion of the sinus of Valsalva. The AV tissue was cut into small pieces and homogenized using Wheaton tapered tissue grinders (Thermo Fisher Scientific, Rockford,   Values indicate mean ± SD, and significant "p-values" between wild-type and Kl −/− groups were given on the top of histograms. Frontiers in Cardiovascular Medicine | www.frontiersin.org area. Our alizarin red staining confirmed the presence of calcific nodules in the AV hinge and aortic annulus of the Kl −/− mice (n = 12) (Figures 1A-D). There was no AV calcification seen in the control mice, which included Kl +/+ and Kl ± mice. Immunohistochemistry analysis indicated that increased phosphorylated SMAD2 (pSMAD2) was associated with AV calcification in the hinge and aortic annulus (n = 6) (Figures 1E-I). We tested if the Tgfb1 mRNA level was upregulated in the AV tissue of Kl −/− mice by qPCR analysis. The data indicated significant upregulation of Tgfb1 mRNA expression (n = 5, p = 0.031) in Kl −/− mice compared to the wildtype control ( Figure 1J). Next, we determined if the increased Tgfb1 mRNA expression level is consistent with the induction of the downstream TGFβ signaling molecules at the protein level. We analyzed the downstream TGFβ signaling pathway molecules in tissue samples collected from the AV tissue of wildtype and Kl −/− mice. Since phosphorylation of SMAD2 and SMAD3 is typically used as a surrogate for TGFβ signaling, western blot analysis was used to quantify both phosphorylated and total (no-phosphorylated) SMAD and SMAD3. The data showed statistically significant induction of activated forms of the SMAD2 (i.e., pSMAD2) and SMAD3 (i.e., pSMAD3) (TGFβspecific SMADs), in Kl −/− mice when compared to wild-type mice (n = 6) (Figures 1K,L).

Genetic Inactivation of TGFβ Signaling Pathway Components Improved Structural Features and Extracellular Matrix Organization of Aortic Valve in Klotho-Deficient Mice
Histological staining was carried out to examine the morphological, cellular, and structural changes in the AV tissue of 8-10-week-old Kl −/− mice following the partial genetic deletion of TGFβ1 and SMAD3 genes (Figures 2A-H).
Histological examination of H&E-stained serial tissue sections confirmed calcified nodules in the AV hinge and annulus in the Kl −/− mice (n = 12) (Figures 2A,B). Morphometric quantitative analysis indicated no significant changes in the overall thickening of AV leaflets in Kl −/− mice compared to wild-type mice (n = 6) ( Figure 2I). Histological and morphometric evaluation of Kl −/− ;Tgfb1 ± mice showed partial improvement in the overall tissue structure of the AV hinge and annulus compared to Kl −/− mice (n = 6) (Figures 2A-C,I). Similar analysis of Kl −/− ;Smad3 ± mice revealed a complete rescue of the AV hinge and aortic annulus (Figures 2A-D,I). Alcian blue staining was used to observe chondrogenic differentiation and proteoglycans or glycosaminoglycans (GAGs) distribution in the AV hinge and aortic annulus in all four groups of mice (n = 6-12). The data indicated abnormal proteoglycans in the AV hinge and aortic annulus in both Kl −/− (n = 12) and Kl −/− ;Tgfb1 ± (n = 6) mice compared to wild-type or Kl −/− ;Smad3 ± mice (n = 12), but morphometric quantification did not reveal any significant changes in the overall proteoglycans content (n = 6) (Figures 2E-H,J). Finally, we determined collagen and elastic fibers' organization through the VVG staining of serial sections (n = 6-12) (Figures 3A-J). The quantitative analysis did not reveal any significant changes in overall collagen content (Figure 3I). Histological assessment of high-powered images and morphometric evaluation showed dysregulated elastic fiber organization in the AV hinge of Kl −/− mice (n = 12), which was significantly rescued by partial genetic deletion of Smad3 (n = 12) but not Tgfb1 (n = 6) (Figures 3A-H,J).
There was no AV calcification in wild-type mice (Figures 4A,E).

DISCUSSION
The klotho deficiency has been shown to cause age-related calcification of AV hinge and aortic annulus in mice (25). The Kl −/− mice also exhibit increased expression of many key osteogenic genes involved in human CAVD (25,63). We demonstrated that upregulation of TGFβ1 and SMAD3 are involved in AV calcification in klotho-deficient mice. The data also indicate that calcific nodule formation in the aortic valve of klotho-deficient mice is associated with increased Tgfb1 expression and elevated levels of activated forms of SMAD2, SMAD3, SMAD1/5, p38, and ERK1/2 MAPK. Importantly, partial genetic ablation of Tgfb1 or Smad3 significantly decreases the expression of Tgfb1 and both canonical (SMAD-dependent) and non-canonical (MAPKmediated) TGFβ signaling pathways and blocks the pathological progression of the AV calcification in Kl −/− mice. Collectively, while Tgfb1 haploinsufficiency significantly improved the structural and ECM features, Smad3 haploinsufficiency almost fully reversed pathological structural changes and disrupted elastin fiber organization in the aortic valve of klotho-deficient mice. Dysregulated elastin fibers in the AV leaflet hinge in Kl −/− mice are consistent with a recent report by Gomez-Stallons et al. (64) showing increased elastin fragmentation and loss of elastin integrity in postmortem leaflet patients with CAVD.
Haploinsufficiency of Smad3 has been observed to be more potent in inhibiting aortic valve calcification compared to Tgfb1 ligand inhibition. This is consistent the notion that signaling from TGFβ1 and/or other TGFβ ligands could be effectively blocked by partial deletion of Smad3. Thus, the effect of lowering TGFβ1 on SMAD2/3 activation and AV calcification rescue may be somewhat less pronounced in Kl −/− ;Tgfb1 ± mice as compared to Kl −/− ;Smad3 ± mice. This may be due to functional compensation of the loss of TGFβ1 by other TGFβ ligands (i.e., TGFβ2, 3) and SMAD2. All three TGFβ ligands (TGFβ1, TGFβ2, and TGFβ3) interact with TGFβ Type I and II receptors and lead to activation of SMAD3 and SMAD2. Activated SMAD2/3 with SMAD4 translocate to the nucleus and regulate TGFβ target genes (e.g., Pai1) (36,65). Increased levels of TGFβ1 (31) and pSMAD2 and PAI1 (66) are associated with CAVD. PAI1 is a known TGFβ target, and therefore, its expression was reduced along with the AV calcification rescue in response to partial inactivation of Tgfb1 (Kl −/− ;Tgfb1 ± mice) or Smad3 (Kl −/− ;Smad3 ± mice) in klotho-deficient mice. There are several experimental limitations in this study. The effect of sex is not specifically addressed. The levels of active and latent TGFβ1 protein are not determined. Since most widely used anti-TGFβ1 antibodies (2G7, 1D11) crossreact to TGFβ3 and not TGFβ2 ligands (67), this study only demonstrated upregulation of Tgfb1 transcript levels in the aortic valve of Kl −/− mice. Regardless of these limitations, the molecular data indicate that both TGFβ1 and SMAD3 contribute to the expression of critical genes involved in the pathogenesis of CAVD. Collectively, our findings represent novel genetic approaches for blocking the progression of CAVD.
Another important piece of information that emerged from this study is the effect of lowering TGFβ1 or SMAD3 on BMP signaling pathway in CAVD. TGFβs and the BMPs are members of the TGFβ family (68). It is well established that increased BMP signaling contributes to AV calcification in Kl −/− mice. It has been reported that both Bmp2 and pSMAD1/5 levels are increased in AV tissue of Kl −/− mice and that genetic inactivation of BMP receptor Bmpr1a in the aortic valve interstitial cells can prevent AV calcification (56). It is remarkable to find that lowering of TGFβ1 as well as SMAD3 has significantly reduced the expression of BMP ligand (Bmp2) and BMP Type I receptor (Alk2) and activation of BMPdependent SMAD (pSMAD1/5). Constitutively active mutants of ALK2 have been identified as causative of Fibrodysplasia Ossificans Progressiva (FOP), which is an extremely rare heritable disorder of connective tissues characterized by progressive heterotopic ossification in various skeletal sites (69). It is known that TGFβ1 via TGFβ Type I receptor can interact with ALK2 resulting in the simultaneous activation of SMAD2/3 and SMAD1/5 (70,71). Also, Alk2 deletion can reduce both SMAD2/3 and SMAD1/5 in cushion mesenchymal cells during heart development (72). Importantly, lowering pSMAD3 (this study) or pSMAD1/5 (56) has a similar effect in blocking AV calcification in Kl −/− mice, suggesting crosstalk between TGFβ and BMP pathways in maintaining AV homeostasis (Figure 7). It is noteworthy that SMAD7, an inhibitory SMAD and TGFβ target gene, binds to and blocks all type I receptors in the TGFβ family, whereas SMAD6 shows preferential binding and inhibitory activity toward the BMP type I receptors ALK3 and ALK6 (73). Both SMAD 6 and 7 cooperate in suppressing the physiological BMP signaling during the differentiation of mesenchymal progenitor cells to osteoblasts (74). This is due to rapid and direct induction of Smad6 expression by the primary BMP stimulus, followed by activation of autocrine TGFβ signaling, which then induces a second wave of Smad7 expression that, together with the preexisting SMAD6, shuts down BMP receptor activity in a more sustained manner and limits the rate of differentiation to osteoblasts (74). Whether the negative regulation of TGFβ and BMP pathway via SMAD7 is failed in Kl −/− mice remains a potential area of future investigation.
This study indicates that the haploinsufficiency of Smad3 is more effective than Tgfb1 heterozygous deletion in decreasing the activation of p38 MAPK in AV tissue of klotho-deficient mice. Improper activation of p38 or ERK MAPKs is a precursor of constitutive SMAD2/3 signaling associated with aortopathy (75,76). TGFβ-activation of the SMAD2/3 requires the kinase activity of the TGFβRI, while the TGFβRI kinase activity is not required for the TGFβ-induced oligomerization of the TGFβRII/TGFβRI complex, where the TRAF6 binding to the TGFβRI results in the activation of TAK1-p38 MAPK pathway. Thus, TGFβ activation of TGFβRII/TGFβRI-complex leads to (a) the TGFβRI-kinase-dependent activation of SMAD2/3 and (b) the TGFβRII/TGFβRI-hetero-oligomerization which will cause a rapid activation of TRAF6 resulting in p38 MAPK pathway activation. It is possible that changes in the non-canonical TGFβ pathway may be a primary determinant of AV calcification in Kl −/− mice. Thus, the exact contribution of the non-canonical (non-SMAD) TGFβ pathway in the AV calcification requires further investigation.
Overall, our findings suggest that klotho is an important mediator of TGFβ signaling in vivo in adult aortic valves (Figure 7). Both TGFβ1 and TGFβ-dependent canonical (pSMAD2/3) and non-canonical (p38 MAPK, ERK1/2 MAPK) pathways are activated in the absence of klotho. This suggests that klotho restrains TGFβ ligand-receptordependent TGFβ signaling and that increased TGFβ-induced activation of SMAD2/3 and/or SMAD1/5 somehow drives the AV calcification in klotho-deficient mice. This is consistent with the observation that partial genetic inactivation of TGFβ1 and SMAD3 (this study), and BMPR1A (56) can block the development and progression of AV calcification in klotho-deficient animals. Our observations are further supported by in vitro evidence that klotho can bind to the TGFβ Type II receptor complex, thereby contributing to the inhibition of TGFβ1 signaling (22,77). In conclusion, our results demonstrated that inhibition of TGFβ signaling by selectively targeting TGFβ1 or SMAD3 could inhibit AV calcification. This information will be useful in designing safer treatments for CAVD.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by the IACUC University of South Carolina.