Functional investigation of the coronary artery disease gene SVEP1

A missense variant of the sushi, von Willebrand factor type A, EGF and pentraxin domain containing protein 1 (SVEP1) is genome-wide significantly associated with coronary artery disease. The mechanisms how SVEP1 impacts atherosclerosis are not known. We found endothelial cells (EC) and vascular smooth muscle cells to represent the major cellular source of SVEP1 in plaques. Plaques were larger in atherosclerosis-prone Svep1 haploinsufficient (ApoE−/−Svep1+/−) compared to Svep1 wild-type mice (ApoE−/−Svep1+/+) and ApoE−/−Svep1+/− mice displayed elevated plaque neutrophil, Ly6Chigh monocyte, and macrophage numbers. We assessed how leukocytes accumulated more inside plaques in ApoE−/−Svep1+/− mice and found enhanced leukocyte recruitment from blood into plaques. In vitro, we examined how SVEP1 deficiency promotes leukocyte recruitment and found elevated expression of the leukocyte attractant chemokine (C-X-C motif) ligand 1 (CXCL1) in EC after incubation with missense compared to wild-type SVEP1. Increasing wild-type SVEP1 levels silenced endothelial CXCL1 release. In line, plasma Cxcl1 levels were elevated in ApoE−/−Svep1+/− mice. Our studies reveal an atheroprotective role of SVEP1. Deficiency of wild-type Svep1 increased endothelial CXCL1 expression leading to enhanced recruitment of proinflammatory leukocytes from blood to plaque. Consequently, elevated vascular inflammation resulted in enhanced plaque progression in Svep1 deficiency. Electronic supplementary material The online version of this article (10.1007/s00395-020-00828-6) contains supplementary material, which is available to authorized users.


Introduction
Coronary artery disease (CAD) and myocardial infarction (MI) are the leading causes of death in industrialized countries [4]. Genome-wide association studies have identified more than 160 variants, mainly located in the non-coding genome, that are associated with CAD/MI [13]. In an exome-wide association study, a coding missense variant in SVEP1 (rs111245230, p.D2702G) is genome-wide significantly associated with CAD/MI (Odds Ratio 1.14 per risk allele) [24]. SVEP1 encodes sushi, von Willebrand factor type A, EGF and pentraxin domain containing protein 1 (SVEP1), an extracellular matrix (ECM) protein that binds to integrin α9β1 [32], which is expressed in lymphatic system endothelial cells (EC) [3]. Integrin α9β1 plays a role in adhesion and transendothelial migration of peripheral neutrophils, hence promoting inflammatory processes. In context of stroke, impairing the function of integrin α9β1 led to reduced thrombosis and inflammation and limited short-and long-term brain damage [10]. Svep1 and integrin β1-pathways are also involved in maintaining vascular integrity. This is achieved by induction of Tie1 expression via binding of Svep1 to angiopoietin-2 [23]. Under acute inflammatory conditions, Tie1 is cleaved, resulting in reduced Tie1 quantity on the cell surface and subsequently weakening cell-cell junctions. Thus, reduced Tie1 expression leads to vascular destabilization and Svep1 was found to be crucial to maintain vascular integrity under baseline conditions [19]. Integrin β1 pathways are also involved in cell migration to the lymphatic system. In line, absence of Svep1 has been found to impair lymphatic vessel formation in zebrafish [17] and mice [23].
Despite its strong association with CAD, the role of SVEP1 in atherosclerosis has yet to be determined. In this project, we sought to investigate the functional involvement of SVEP1 in mice and patients with atherosclerosis.

SVEP1 expression in the STARNET study
The Stockholm-Tartu Atherosclerosis Reverse Networks Engineering Task study (STARNET) subjects recruitment and tissue collection were described previously [14]. Briefly, patients with coronary artery disease (CAD) who were eligible for open-thorax surgery at the Department of Cardiac Surgery, Tartu University Hospital in Estonia as well as control subjects without CAD were enrolled after informed consent. Tissue biopsies were obtained to study tissue-specific gene expression and the disease. Tissues were rinsed and RNA was extracted as described previously [14]. In the case-control matched study, cases and controls with matched age, gender, and BMI were selected and sequenced for whole transcriptome. Samples were sequenced with poly(A)+selection on Illumina HiSeq with single-end at read lengths of 100 base pairs. Quality control was performed using FASTQC [1] checking raw sequence data for per-base quality, per-sequence quality, number of duplicate reads, number of reads with an adaptor, sequence length distribution, per-base GC content, per-sequence GC content and Kmer content. GENCODE was used as reference annotation to quantify gene and isoform expression. Sequencing reads (fastq files) were mapped with STAR [12] onto the human genome. Raw reads were summarized by feature counts [20]. Differential gene expression between cases and controls was analyzed using R package limma [22].

Human carotid artery plaque specimen, immunohistochemistry, and single-cell RNA sequencing
Human carotid arterial plaque material of the Munich Vascular Biobank [29] was sampled during carotid endarterectomy (CEA), fixed for 48 h in 2% zinc-paraformaldehyde at room temperature, paraffin-embedded, and finally cut into 5-μm thick slides. Per carotid plaque specimen, four slides were stained with hematoxylin and eosin (HE) as well as Elastica van Gieson's staining. For immunohistochemical analysis, consecutively cut tissue sections of 3 µm thickness were deparaffinized, permeabilized with H 2 O 2 and blocked in milk powder. Sections were then incubated for 1 h at room temperature with primary antibodies against SVEP1 (HPA020610, Sigma-Aldrich, St. Louis, MO, USA), CD31 (M0814, DAKO, Carpinteria, CA, USA), CD68 (M0823, DAKO, Carpinteria, CA, USA), smooth muscle a-actin/ SMA (M0635, DAKO, Carpinteria, CA, USA). SVEP1 staining was established and compared to isotype control IgG antibody/no primary control (Suppl. Fig. S1). Secondary antibodies were provided with the DAKO REAL Detection Kit Rabbit/Mouse (DAKO, Carpinteria, CA, USA). Detection was mediated by a 3,3′-diaminobenzidine-coupled reaction according to the manufacturer's instructions. Nuclear staining with hematoxylin was performed before dehydrating the slides in an increasing ethanol row followed by xylene. Slides were finally mounted with EUKITT (Kindler, Bobingen, Germany) and imaged.
Single-cell RNA sequencing data from 18 (14 male and 4 female) individuals were obtained from Depuydt et al. [9] within the Athero-Express Biobank Study (www.ather oexpr ess.nl) [36], an ongoing biobank study at the University Medical Centre Utrecht (UMCU). Details are available in the Supplemental Material.

Primary cells
Human coronary artery smooth muscle cells (CASMC), human aortic adventitial fibroblasts (AAF), human aortic endothelial cells (AEC), human aortic smooth muscle cells (ASMC), human coronary artery endothelial cells (CAEC), human coronary artery smooth muscle cells (CASMC), and human umbilical vein endothelial cells (HUVEC) were purchased from commercial suppliers (CASMSC: Cell Applications, San Diego, CA, USA; AAF, AEC, CAEC, CASMC: from ScienCell, Carlsbad, CA, USA; HUVEC: PromoCell, Heidelberg, Germany). Monocytes were isolated from whole blood from one healthy individual using a combined Ficoll-Paque density gradient and a CD14 magnetic bead separation approach. Details on isolating blood monocytes as well as the investigated primary cells are available in the Supplemental Material.
Isolation of nucleic acids, cDNA synthesis, and (quantitative) polymerase chain reaction RNA from isolated blood monocytes was extracted using the RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's recommendations. Otherwise, cells were washed with Dulbecco's PBS (Biochrom, Berlin, Germany), lysed by addition of 500 µl TRIzol (Thermo Fisher Scientific, Waltham, MA, USA) and stored at − 80 °C. After thawing, 120 µl DEPC treated H 2 O (Invitrogen, Carlsbad, CA, USA) and 100 µl chloroform were added. After 10 min incubation at room temperature, the mixture was transferred to pre-chilled phase lock tubes (Quantabio, Beverly, MA, USA) and centrifuged at 13,000 rpm and 4 °C for 10 min. 280 µl of isopropanol and 1.5 µl of GlycoBlue (Invitrogen, Carlsbad, CA, USA) were added to the aqueous phase. The isolated RNA was incubated at − 20 °C for 30 min and pelleted by centrifugation at 14,000 rpm for 30 min at 4 °C and washed twice with 75% ethanol. After evaporation of ethanol, RNA pellet was resuspended in 14 µl of DEPC treated water. After DNA digestion with DNase I ( Real-time quantitative PCR was performed using TaqMan Universal Master Mix II (Thermo Fisher Scientific, Waltham, MA, USA) in a total volume of 20 µl using MicroAmp Fast 96-Well 0.1 ml reaction plates (Thermo Fisher Scientific, Waltham, MA, USA). All reactions were conducted over 40 cycles and performed in duplicates on a ViiA 7 system using Taqman probes (both Thermo Fisher Scientific, Waltham, MA, USA). The Taqman probes which were used to detect genes of interest are listed in Suppl. Table S2. GAPDH/Gapdh or RPLP0/Rplp0 were used as housekeeping genes. Expression of genes of interest comparing SVEP1 and SVEP1_p.D2702G was analyzed using the 2 −ΔΔCt method. Absolute mRNA levels were compared using 2 −ΔCt values.

Histology
The aortic root was embedded in Tissue-Tek O.C.T. Compound (Sakura Finetek, Alphen aan den Rijn, Netherlands) and snap-frozen in a methylbutane bath cooled with dry ice.

Flow cytometry
The aorta of each mouse was used to obtain single-cell suspensions. Aortae were extensively flushed with phosphate buffered saline to remove blood leukocytes and then excised from aortic root to iliac bifurcation. However, a marginal contamination with blood cells cannot be ruled out.
Perivascular tissue was carefully removed using a microscope (microdissection). Subsequently, aortae were minced with scissors and digested in collagenase I (450 U/ml), collagenase XI (125 U/ml), DNaseI (60 U/ml), and hyaluronidase (60 U/ml) (Sigma-Aldrich, St. Louis, MO, USA) at 37 °C and 750 rpm for 1 h for staining of myeloid cells as it has been described previously [31]. Unlike histology, flow cytometry does not allow to distinguish between intimal and adventitial leukocytes. Consequently, changes in plaque leukocyte numbers-assessed by flow cytometry-may not solely be driven by intimal leukocytes, but also by adventitial leukocytes.
To obtain purified neutrophils and monocytes for adoptive transfer experiments, cells were isolated from the bone marrow of Ubc-GFP donor animals using Ly6G-PE (127608, clone 1A8) and CD115-biotin (135508, clone AFS98, both BioLegend, San Diego, CA, USA) that allowed coupling to magnetic beads (anti-PE and streptavidin microbeads, 130-048-801 and 130-048-101, Miltenyi Biotec, Bergisch Gladbach, Germany) and separation of cells via magnetic-activated cell separation columns (130-042-401, Miltenyi Biotec, Bergisch Gladbach, Germany). Equal amounts of purified neutrophils and monocytes were injected i.v. into ApoE −/− Svep1 +/− or ApoE −/− mice which were fed a high-cholesterol diet for 6 weeks. The aortae were harvested and analyzed as described above. The number of CD11b high GFP high cells within the aortae was quantified using flow cytometry.

Cloning of SVEP1 constructs
A vector containing the SVEP1 open reading frame followed by a Myc-DDK-tag (NM_153366, RC214271; pCMV_SVEP1) was purchased from OriGene Technologies (Rockville, MD, USA). The p.D2702G amino acid change (pCMV_SVEP1_p.D2702G) was introduced using in vitro site-directed mutagenesis (QuikChange II XL; Agilent Technologies, Santa Clara, CA, USA) according to the supplier's protocol and using the following mutagenesis primers: SVEP1var_for: 5′-GCC ATT CCA AGT TCC ACC TTC CTG GCA GAT CAG-3′, SVEP1var_rev: 5′-CTG ATC TGC CAG GAA GGT GGA ACT TGG AAT GGC-3′. Coding sequence of human SVEP1 was then cloned into a pcDNA-pDEST40 expression vector C-terminally tagged with a HA-tag (Life Technologies, Carlsbad, CA, USA). The expression plasmids were sequenced on both strands prior to transfection of eukaryotic cells.

CXCL1 enzyme-linked immunosorbent assays
CXCL1 enzyme-linked immunosorbent assays (ELISA) were used to determine human (ab100530) and murine (ab216951; both Abcam, Cambridge, UK) CXCL1 protein levels in cell culture supernatant and plasma samples, respectively, according to the suppliers' recommendations.

Stimulating HUVEC with SVEP1 or SVEP1 and interleukin 1β
HUVEC (PromoCell, Heidelberg, Germany) were seeded at a density of 200,000 cells in 12-well cell culture plates. 24 h after seeding the cells were stimulated with concentrated conditioned media containing either SVEP1-HA, SVEP1-D2702G-HA or mock control. After 24 h, cells were harvested for RNA isolation. To investigate interleukin 1β-induced CXCL1 expression, HUVEC were seeded at a density of 160,000 cells in 24-well cell culture plates. Two h after seeding the cells were stimulated with a final concentration of 12.5 ng/µl recombinant human interleukin 1β (R&D systems, Minneapolis, MN, USA) and concentrated conditioned media either containing SVEP1 or mock control. After 24 h, supernatants were collected for subsequent analyses. Samples were frozen in liquid nitrogen and stored at − 80 °C.

Phenome-wide association study of SVEP1 variants in UK Biobank
Phenome-wide association analysis for rs111245230 was conducted using the UK Biobank dataset. We integrated the International Classification of Disease ICD9, 10, OPCS-4 (Office of Population, Censuses and Surveys: Classification of interventions and Procedures, version 4) and self-reported information to define individual condition with different phenotypes. Disease classification was used to better navigate through the data, and we grouped phenotypes into cardiovascular, endocrine, neurological, digestive, genito-urinary, musculoskeletal, respiratory, eye, cancer and others [11,26]. Finally, we identified 64 phenotypes and grouped them into nine classes. We extracted the genotype for rs111245230 from the full UK Biobank imputed dataset on 487,406 participants and did a quality control to filter samples with high kinship coefficient (> 0.088) and more than ten putative third-degree relatives in the kinship table. Samples without clear disease definition were also removed during analysis. We used PLINK [7] to test the association between rs111245230 and each phenotype independently based on logistic regression with the additive genetic model. The adjustment of population stratification includes age, gender and the top two principal components. A Bonferroni corrected threshold (p < 0.0008) was used to determine significantly related phenotypes.

Statistical analysis
Data distribution was assessed using the Kolmogorov-Smirnov test. Data were analyzed using Student's/ Welch's unpaired/paired/one-column t test (in case of normally distributed data) or Mann-Whitney/Wilcoxon test (in case of not normally distributed data), as appropriate and indicated in the figure legends. p values < 0.05 were regarded as significant. GraphPad Prism version 8 for Mac OS X (GraphPad Software, La Jolla, CA, USA) was used.

SVEP1 is expressed in vascular and metabolic tissues
Currently, SVEP1's role in CAD has not yet been demonstrated. To test whether SVEP1 is localized in vascular tissue, we explored relevant SVEP1 expression in human tissues. Here, we evaluated SVEP1 mRNA levels in RNA sequencing (RNAseq) data from the STARNET study which examined gene expression in seven metabolic and vascular tissues [14]. We found the highest SVEP1 expression levels in adipose tissue (visceral adipose tissue and subcutaneous fat) and vessels (aorta and mammary artery) (Fig. 1a). Of note, these tissues also show highest SVEP1 expression levels in the GTEx database [16] (Suppl. Fig. S2). The strongest difference in SVEP1 expression was detected in aortae of patients with CAD as compared to those from controls with lower levels in cases (Fig. 1b). In addition to mRNA, we next examined whether also SVEP1 protein is detectable in vascular tissues. Using immunohistochemistry in carotid artery plaques from patients undergoing carotid artery endarterectomy, we detected SVEP1 in proximity to smooth muscle cells (SMC), endothelial cells (EC), and CD68-positive cells as monocytes/macrophages/secretory VSMC (using alphasmooth muscle actin, CD31, and CD68 as markers respectively) (Fig. 1c; Suppl. Figs. S3, 4). These data indicate that SVEP1 is present in the vasculature with reduced levels in patients with atherosclerosis, i.e. CAD cases.

Endothelial cells and smooth muscle cells represent a major cellular source of SVEP1 in the vasculature
As SVEP1 is an ECM protein and hence released from cells, immunohistochemistry is less suitable to precisely determine the cellular sources of SVEP1. We thus examined SVEP1 expression in single cell RNAseq (scRNAseq) data from human carotid artery plaques to determine the cellular source of SVEP1 in the vessel wall. We found EC and vascular SMC had the highest expression levels ( Fig. 2a; Suppl. Fig. S5). In addition, we analyzed SVEP1 mRNA in primary fibroblasts, SMC, and EC in cell culture, as well as in classical monocytes isolated from human whole blood. Strong expression of SVEP1 was detected in primary fibroblasts, SMC, and EC, whereas blood monocytes showed only weak expression (Fig. 2b). Taken together, these data indicate that SVEP1 is mainly produced by EC and vascular SMC within vascular and atherosclerotic tissue.
In addition to the chemokine CXCL1, we also detected lower SELE expression in EC when exposed to wild-type SVEP1 compared to control (Suppl. Fig. S11). SELE encodes E-selectin, a cell adhesion molecule that together with chemokines is pivotal in mediating leukocyte recruitment [15]. Our data indicate that wild-type SVEP1 controls EC leukocyte recruitment capacities and that the CAD risk variant SVEP1_p.D2702G mimics reduced intact wild-type SVEP1 levels.

Discussion
Knowledge of the molecular and cellular functions of SVEP1 is limited. Besides its role in lymphatic vessel formation [17,23], SVEP1 has been investigated in the context of septic shock and endotoxinemia [25,34]. The fact that homozygous loss of Svep1 is lethal [17,23] suggests an important role in developmental processes. Our data provide a first insight into the role of SVEP1 in atherosclerosis. In a series of in vitro and in vivo experiments, Fig. 6 Phenotype-wide association study of SVEP1_p. D2702G. The CAD risk variant was associated with an increased risk of hypertension and inguinal hernia but a reduced risk of uterine fibroid after correction for multiple testing (red). Furthermore, there was a nominally significant trend toward higher risk of leukemia, but reduced risk for uterine cancer (orange). Odds ratios and unadjusted p values are depicted in Suppl. Table S2 we explored the role of the recently identified CAD gene. We found that SVEP1 was detectable in human atherosclerotic plaques and produced by various vascular wall cell types, including vascular SMC but particularly EC. Under proatherogenic conditions, ApoE −/− mice with Svep1 haploinsufficiency developed larger atherosclerotic plaques compared to ApoE −/− mice with normal Svep1 levels. In addition, plaque leukocytes were more numerous in ApoE −/− Svep1 +/− compared to ApoE −/− Svep1 +/+ mice. In adoptive transfer experiments, we demonstrated that plaque leukocyte expansion is a consequence of enhanced leukocyte recruitment in the presence of reduced Svep1 levels. EC are critically involved in leukocyte recruitment under proatherogenic conditions [15,35]. This and the fact that SVEP1 is located extracellularly led to the hypothesis that it might influence EC phenotypes. To get a first insight into whether the CAD risk variant SVEP1_p.D2702G (like Svep1 +/− mice, carriers of this missense variant also lack wild-type SVEP1) leads to decreased SVEP1 function, we analyzed EC phenotypes after incubation with either SVEP1 or SVEP1_p.D2702G. Adding SVEP1_p. D2702G was associated with increased expression of the chemokine CXCL1 and the adhesion molecule SELE suggesting wild-type SVEP1 plays a rather atheroprotective role that was altered in the presence of the SVEP1_p. D2702G variant. In CAD cases, we further found reduced SVEP1 mRNA levels compared to controls. In addition, our data show that SVEP1 exerts regulating effects on EC resulting in decreased CXCL1 and SELE expression in the presence of higher SVEP1 levels in vitro. Both CXCL1 and E-selectin mediate leukocyte recruitment from blood to plaques and hence play a crucial role in the early stages of atherosclerosis when dysfunctional EC secrete chemokines/upregulate adhesion molecules and thereby induce monocyte and neutrophil adhesion [15,37]. In particular, EC-derived CXCL1 has been shown to promote leukocyte recruitment under proatherogenic conditions [37]. SVEP1 led to a dose-dependent decrease in CXCL1 protein levels, demonstrating that SVEP1 likely helps regulate EC phenotypes. This hypothesis is further supported by the finding that mice with a heterozygous loss of Svep1 also displayed elevated Cxcl1 plasma levels under proatherogenic conditions. Also, the observed increased atherosclerotic plaque formation and more numerous neutrophils, Ly6C high monocytes, and macrophages inside plaques of ApoE −/− Svep1 +/− compared to ApoE −/− Svep1 +/+ mice accord with the downstream effects of CXCL1 and E-selectin in atherosclerosis [15,37]. Of note, we previously identified SVEP1 as a target of the protease ADAMTS-7 [18]. The ADAMTS7 gene has been linked to atherosclerotic plaque formation in humans [8,30,33] and mice [2]. While the molecular and cellular mechanisms involving ADAMTS-7 in atherosclerosis have not yet been elucidated, SVEP1 degradation may be involved.
Drug targets identified by genetic analyses reportedly have higher success rates in the drug development pipeline [27]. Our findings raise the question of whether increasing wild-type SVEP1 protein levels, e.g. via inhibiting its degradation, might also have beneficial effects regarding other disorders and phenotypes. To explore this, we conducted a PheWAS in UK Biobank. We replicated the association between the CAD risk variant and hypertension; this risk variant was also associated with a higher risk of inguinal hernia and trend for leukemia. When evaluating this concept's therapeutic potential, it is important to also consider possible detrimental effects. The CAD risk allele was associated with a reduced risk of uterine fibroids and trend for uterine cancer. While uterine fibroids are considered benign, familial prevalence of fibroids has been linked with a higher risk of uterine cancer [28]. However, reliable data are scarce. While modulating SVEP1 may be considered as a novel treatment strategy its consequences on other phenotypes, especially malignancies, nevertheless warrant further investigation.
In summary, we hypothesize that SVEP1 which is produced by vascular cell types and secreted to the ECM of the vascular wall acts as a silencing factor on EC activity by reducing the expression of chemokines and cell adhesion molecules, e.g., CXCL1 and E-selectin. If wild-type SVEP1 protein levels are reduced, e.g. via degradation by MMPs, or in the presence of mutant SVEP1 (SVEP1_p.D2702G missense variant), CXCL1 and E-selectin expression rises leading to enhanced inflammatory cell influx. Figure 7 displays this hypothesized functional cascade.
Whereas in this study, we postulate an atheroprotective effect of SVEP1 on EC, we cannot exclude further functions of SVEP1, particularly in other cell types. Specifically, the CAD risk variant has been shown to be associated with Fig. 7 Schematic illustration of the role of SVEP1 and EC in atherosclerosis. In the presence of SVEP1 (blue), CXCL1 expression is silenced. After SVEP1 degradation (light blue) or in the presence of SVEP1_p.D2702G (red), CXCL1 expression and secretion are increased, resulting in higher recruitment of inflammatory cells slight effects on blood pressure [24]. As SVEP1 is expressed in SMC, it might also influence SMC phenotypes. However, the association with blood pressure could not explain the risk variant's association with CAD [24]. Furthermore, the role of SVEP1 in silencing CXCL1 and E-selectin expression by EC is independent of blood pressure since experiments were performed under no-flow, static conditions. Nevertheless, further effects, e.g. lower blood pressure in the presence of wild-type SVEP1, might provide additional atheroprotection. We also observed SVEP1 expression in plaque areas that also stained positive for the macrophage marker CD68. Although isolated blood monocytes only weakly express SVEP1, other cell types might also produce SVEP1 and thus influence plaque monocyte/macrophage phenotypes.

Study limitations
Our study has several limitations. First, we do not know how SVEP1 mediates its effects on EC on a molecular level. ECM proteins have diverse effects on adjacent cell types that may surpass the known interaction with integrin α9β1. Second, we investigated a heterozygous knockout of Svep1 which might not fully replicate the effects of the human risk variant since loss of intact wild-type SVEP1 may vary. However, we detected reduced SVEP1 mRNA levels in vascular tissues from patients with CAD, as compared to controls, an event reflected by the mouse model we used. Third, and most importantly, findings from in vitro experiments do not necessarily reflect in vivo situations. In most in vitro approaches, the environment that naturally surrounds cells-in our case the ECM-is missing. Consequently, in vitro assays are somewhat artificial. We found that loss of SVEP1 leads to increased CXCL1 levels, both in vitro and in vivo. This indicates that at least with respect CXCL1 release, our in vitro data resemble our in vivo findings. It is of course possible that soluble SVEP1 exerts further in vivo effects that are not addressed in this study.