Refinement of the HIVAN1 Susceptibility Locus on Chr. 3A1-A3 via Generation of Sub-Congenic Strains

HIV-1 transgenic mice on the FVB/NJ background (TgFVB) represent a validated model of HIV-associated nephropathy (HIVAN). A major susceptibility locus, HIVAN1, was previously mapped to chromosome 3A1-A3 in a cross between TgFVB and CAST/EiJ (CAST) strains, and introgression of a 51.9 Mb segment encompassing HIVAN1 from CAST into TgFVB resulted in accelerated development of nephropathy. We generated three sub-congenic strains carrying CAST alleles in the proximal or distal regions of the HIVAN1 locus (Sub-II, 3.02–38.93 Mb; Sub-III, 38.45–55.1 Mb and Sub-IV, 47.7–55.1 Mb, build 38). At 5–10 weeks of age, histologic injury and proteinuria did not differ between HIV-1 transgenic Sub-II and TgFVB mice. In contrast, HIV-1 transgenic Sub-III and Sub-IV mice displayed up to 4.4 fold more histopathologic injury and 6-fold more albuminuria compared to TgFVB mice, similar in severity to the full-length congenic mice. The Sub-IV segment defines a maximal 7.4 Mb interval for HIVAN1, and encodes 31 protein coding genes: 15 genes have missense variants differentiating CAST from FVB, and 14 genes show differential renal expression. Of these, Frem1, Foxo1, and Setd7 have been implicated in the pathogenesis of nephropathy. HIVAN1 congenic kidneys are histologically normal without the HIV-1 transgene, yet their global transcriptome is enriched for molecular signatures of apoptosis, adenoviral infection, as well as genes repressed by histone H3 lysine 27 trimethylation, a histone modification associated with HIV-1 life cycle. These data refine HIVAN1to 7.4 Mb and identify latent molecular derangements that may predispose to nephropathy upon exposure to HIV-1.


Introduction
HIV-1 associated nephropathy (HIVAN) is a major complication of HIV-1 infection, and results in end-stage renal disease without antiviral treatment [1,2]. Clinically HIVAN manifests with proteinuria, and histologically it is characterized by collapsing focal and segmental glomerulosclerosis, microcystic tubular dilatation, and interstitial inflammation [3]. HIVAN arises due to HIV-1 induced dysregulation of podocytes, the glomerular epithelial cells that maintain the kidney filtration barrier [4][5][6][7]. HIVAN development has a strong genetic component both in humans and mouse models. In humans, HIVAN predominantly develops in individuals of African descent, and relatives of HIVAN patients have a higher incidence of endstage renal disease [8]. Recently, studies have shown that susceptibility in humans is attributable to coding variants in APOL1 that confer resistance to trypanosomiasis but increase susceptibility to kidney failure [9]. The mechanisms through which APOL1 variants produce kidney injury are under active investigation [10,11]. Although mice do not have an APOL1 ortholog, transgenic expression of a replication deficient HIV-1plasmid that contains all the structural viral proteins except Gag and Pol reproduces characteristic lesions of HIVAN in the FVB/NJ genetic background (TgFVB strain) [4][5][6]. This finding indicates that perturbations in alternative biological pathways, in the absence of APOL1, can produce HIVAN in the mammalian kidney, and hence analysis of mouse models of HIVAN may inform the pathogenesis of human disease. The development of murine HIVAN is strain dependent, with the FVB/NJ as the most susceptible strain, while F1 hybrids of TgFVB with other inbred strains show variable susceptibility to disease [12][13][14]. We have used crosses between TgFVB and other inbred strains to map four nephropathy susceptibility loci (named HIVAN1-4) [12][13][14]. The HIVAN1 susceptibility locus was previously mapped to chromosome 3A1-A3 in a cross between TgFVB and CAST/EiJ (CAST) strain. [12]. To confirm this locus, we previously generated a congenic strain, TgFVB-HIVAN1 CAST , by introgressing a 51.9 Mb CAST interval encompassing the HIVAN1 locus into the FVB genome [15]. While wild-type FVB-HIVAN1 CAST mice were phenotypically normal, HIV-1 transgenic counterparts developed early onset and more severe kidney disease by 6-8 weeks of age compared to TgFVB. This initial congenic interval contained over 300 protein coding genes, leaving open the possibility that multiple genes contribute to increased susceptibility to nephropathy. Here, we report generation and characterization of three sub-congenic strains that carry sub-regions of the original HIVAN1 locus. These new HIVAN1 sub-congenic strains allowed us to refine the HIVAN1 locus to a maximum 7.4Mb interval, enabling detailed annotation of positional candidates and analysis of molecular pathways producing susceptibility to nephropathy.

Mouse strains and their genotypes
This study was carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the IACUC committee at the Columbia University Medical Center. The mice were housed in a pathogen-free facility with 12 hour light cycle and were fed with a regular chow ad libitum.
The FVB/NJ mice were purchased from Jackson Laboratories. The HIV-1 transgenic mouse line TgN(pNL43d14)Lom 26 (TgFVB) on the inbred FVB/NJ genetic background and the TgFVB-HIVAN1 CAST congenic strain were previously described [15][16][17]. We backcrossed TgFVB-HIVAN1 CAST mice to FVB/NJ strain and identified recombinant mice to generate three sub-congenic lines containing smaller regions of the original congenic locus (Fig 1). The Tg-Sub-II-HIVAN1 CAST (Sub-II) strain carried CAST alleles between rs6372626 (4.25 Mb) and rs46441005 (38.55  The marker positions are indicated by genome build 38p.3. SNP IDs and annotation across the HIVAN1 interval were obtained from the Mouse Phenome Database (http://phenome.jax.org/). The SNP annotations are presented in S1 Table. Animals were euthanized (by CO 2 asphyxiation followed by cervical dislocation) at 5-10 weeks of age and urine and kidneys were collected for phenotypic studies. Proteinuria and renal histology were compared between mice of differing genotypes at the HIVAN1 locus.

Evaluation of renal histopathology and albuminuria, and statistical analysis
Kidneys were formalin fixed and paraffin embedded, and 3 um sections were cut and stained with periodic-acid Schiff (PAS). Renal histology was scored independently by an investigator (VDD) blinded to genetic background, using a semi-quantitative scale. We scored the severity of glomerular injury (segmental and global glomerulosclerosis), tubulo-interstitial disease (tubular proteinaceous casts/ tubular cystic dilatation, tubular atrophy /interstitial fibrosis), and interstitial inflammation. The histology phenotypes were quantified according to the percent of glomeruli or percent cortical parenchyma affected in whole kidney cross-sections after visualization of at least 200 glomeruli. Representative images of characteristic HIVAN kidney histopathology features are shown in Fig 2 and S1 Fig. Albuminuria was quantitated in the urine of random subsets of mice of each genotype (4-10 mice in each group) and presented as albumin-to-creatinine ratio (ug/mg). Albumin and creatinine were measured with Albuwell M and Creatinine ELISA kits (Exocell, Philadelphia, PA). Statistical analyses of proteinuria and histologic injury between TgFVB and TgFVB-HIVAN1-CAST strains were performed using Kruskal-Wallis Anova and two-sided Mann-Whitney nonparametric tests, using GraphPad Prism 7.01 software. P-values < 0.05 were considered significant.

RNA isolation and microarray analysis
Total kidney RNA was isolated using trizol reagent (Invitrogen, Grand Island, NY), followed by treatment with DNaseI and clean-up using the RNeasy kit (QIAGEN) according to the protocols recommended by the manufacturers.
We performed microarray analysis with the Affymetrix ST 1.0 gene arrays (Santa Clara, CA). Total kidney RNA was extracted from 20 HIVAN1 congenic mice (11 females / 9 males) and 19 FVBN/J littermates (10 females / 9 males). Sample preparation, labeling and hybridization were performed as per Affymetrix recommended protocol. Signal intensities were normalized using the RMA method. Differential gene expression was analyzed with two sided t-tests and corresponding False Discovery Rates (FDR) q-values were calculated. Pathway analysis was performed by computing overlap with two curated gene sets from the Molecular Signature Database (Canonical Pathways and Chemical and Genetic Perturbations, http://www. broadinstitute.org/gsea/msigdb/).We also cross-annotated the congenic kidney transcriptome with a recently described RNAseq transcriptome from murine FACS-sorted podocytes [18]. The transcriptome datasets are presented in S2 and S3 Tables. The images depict features of the histopathology scores that were close to the average for each group. Focal segmental and global glomerulosclerosis, tubular casts and focal interstitial inflammation were more severe in Sub-IV mice than in Tg-FVB mice (Periodic acid-Schiff, x200). B& C. Severity of renal histopathology in mice heterozygous (B) and homozygous (C) for CAST alleles across congenic segments. HIV-1 transgenic congenic, Sub-III and Sub-IV mice, but not Sub-II mice have statistically significant increased renal injury compared to TgFVB mice (* -p<0.05, ** -p<0.01, *** -p< 1x10 -3 , **** -p< 1x10 -4 ., E&F. Additive effect of CAST alleles on renal injury in HIV-1 transgenic Sub-III (E) and Sub-IV (F) mice. The traits are shown as mean ± standard errors of mean. Panels E and F show genotypic comparisons for each trait by nonparametric test (Kruskal-Wallis one way ANOVA) with the associated p-values shown to the right of each curve.
HIV interactions were queried from the NCBI HIV-1 Human Interaction Database.

Results
Characterization of HIVAN1 sub-congenic Sub-II, Sub-III and Sub-IV strains, carrying distal or proximal regions of the HIVAN1 locus The TgFVB-HIVAN1 CAST congenic strain carries a 51.9 Mb segment on Chr. 3 containing a susceptibility allele(s) for nephropathy from CAST, introgressed into the TgFVB genome [15]. To dissect the HIVAN1 locus, we generated three sub-congenic strains carrying proximal or distal regions of the HIVAN1 locus (Sub-II-HIVAN1 CAST , Sub-III-HIVAN1 CAST and Sub-III-HIVAN1 CAST , abbreviated as Sub-II, Sub-III and Sub-IV, respectively). The boundaries of the sub-congenic intervals are depicted in Fig 1. In the absence of the HIV-1 transgene, all congenic mice were phenotypically normal for up to 9 months of age and showed no histopathologic or biochemical evidence of nephropathy.
We first characterized HIV-1 transgenic mice heterozygous for each sub-congenic segment. The renal injury parameters did not differ between Tg-Sub-II CAST/FVB and TgFVB (Fig 2B). In contrast, Tg-Sub-III-CAST/FVB and Tg-Sub-IV CAST/FVB mice showed a 1.5-2.4 fold increase in glomerulosclerosis, tubule-interstitial casts/cysts, tubular atrophy/interstitial fibrosis and inflammation ( Table 1 and Fig 2), and were comparable to the TgFVB-HIVAN1 CAST strain mice carrying the full congenic segment. This suggested that the HIVAN susceptibility gene(s) is encoded within the smaller Sub-IV interval.
Next, we generated HIV-1 transgenic mice that were homozygous for each sub-congenic segment. Consistent with the phenotype of heterozygous congenic mice, Tg-Sub-II CAST/CAST mice were indistinguishable from TgFVB. However, Tg-Sub-III CAST/CAST and Tg-Sub-IV CAST/CAST mice displayed advanced kidney disease, with 2.5-4.4 fold increase in severity across all histological parameters (p = 6x10 -7 -0.003 compared to TgFVB, Table 1 and Fig 2C). Thus the severity of kidney disease increased with the number of CAST alleles in Tg-Sub-III and Tg-Sub-IV mice, demonstrating an additive effect (Fig 2E and 2F, nonparametric p-value = 9x10 -6 5x10 -3 ). Similar to the histopathology traits, albuminuria levels were not statistically different between Tg-Sub- II-CAST/CAST and Tg-FVB mice, but were up to 6-fold higher in Tg-Sub-III and Tg-Sub-IV congenic mice, with an additive effect of CAST alleles (Fig 3).
In summary, only the Tg-Sub-III and Tg-Sub-IV mice showed increased severity of disease, capturing the phenotypic effect observed in the original congenic mice carrying the full HIVAN1 congenic segment. Taken together, these data further confirm the HIVAN1 locus and refine the susceptibility gene(s) to the 7.4Mb delimited by rs30102504 and rs30553284 within the Sub-IV region.

Annotation and prioritization of genes within the Sub-IV locus
The Sub-IV locus spans 7.4Mb and encodes 31 RefSeq/UCSC annotated protein coding genes, 2 tRNAs, and 13 pseudogenes ( Table 2 and S1 Table). To identify putative sequence variants that may account for differential susceptibility to nephropathy, we compared the Sub-IV locus sequence between CAST/EiJ and FVB/NJ strains (Genome Build 38.p3). Consistent with the known genetic diversity between CAST/EiJ and laboratory-derived strains, all 31 RefSeq/UCSC annotated genes contained at least one coding and/or non-coding SNP differentiating the two strains. Among these, 15 genes had coding non-synonymous variants, but none harbored loss of function variants between the two strains (S1 Table). Missense variants in Frem2, Mgarp and Rfxap were predicted to be damaging by at least 1 program. There were multiple non- coding structural variants within this interval, including eight located within intronic regions of seven genes (S1 Table).
Concurrently with generation of sub-congenic strain, we performed genome-wide expression profiling of whole kidneys from wild-type full congenic FVB-HIVAN1 CAST and FVB strains. We only profiled healthy wild-type kidneys because the profound histopathological lesions of HIV-1 transgenic mice introduce many secondary gene expression changes that can confound interpretation of transcriptomic data. We identified 327 differentially expressed transcripts between the two strains at a FDR q-value < 0. 1 (corresponding to a nominal Pvalue < 0.0014, Table 2 and S1 Table). As expected, the majority of the genes that were most differentially expressed were encoded within the congenic interval, indicating cis-eQTL effects. Within the Sub-IV locus,14 genes were differentially expressed between the two strains at a nominal p-value < 0.05. Three genes (Frem2, Foxo1 and Setd7) have been implicated in nephropathy [19][20][21]. In addition, 6 genes have an interaction with HIV-1 documented in the NCBI database ( Table 2). Finally we cross-annotated our data with a recently published podocyte RNA-seq transcriptome dataset (S2 Table) [18]. Five genes located within the Sub-IV locus (Ndufc1, Setd7, Ufm1, Alg5, Cog6) are in the top 50 th percentile for podocyte-expressed transcripts and are also differentially expressed in congenic mice ( Table 2). Of note, Ndufc1, encoding a subunit of the NADH dehydrogenase in mitochondria, is in the top 10 th percentile of podocyte expressed transcripts [18] and is also highly expressed in human kidney according to Genotype-Tissue Expression database (GTEx: http://www.gtexportal.org/). Complex I enables electron transfer from NADH to Coenzyme-Q 10 and mutations affecting CoQ 10 biosynthesis can cause nephrotic syndrome [22,23]. Hence annotation of the Sub-IV locus identified a number of plausible candidates that require further investigation.

Molecular perturbations in the renal transcriptome encoded outside the HIVAN1 interval
To gain insight into pathways that are regulated downstream of the HIVAN1 locus, we examined transcripts encoded outside the 51.9 Mb HIVAN1congenic interval. Although FVB and FVB-HIVAN1 CAST mice are genetically identical outside the HIVAN1 interval, we identified 287 differentially expressed transcripts at FDR q-value <0.1 (corresponding to a nominal pvalue of 0.0013), whereas only 34 transcripts would be expected to reach this significance level by chance. Because the two strains are genetically identical outside the HIVAN1 interval, these expression differences are attributable to a primary genetic perturbation within the HIVAN1 locus. Pathway analysis of all differentially expressed transcripts identified significant enrichment for multiple molecular signatures, including apoptosis induced by TRAIL, doxorubicin and serum deprivation ( Table 2 and S1 Table). In addition, we detect signatures of histone methylation, extracellular matrix components, and adenoviral infections. A number of these molecular signatures may be attributable to genetic perturbations within the HIVAN1 locus. For example, the molecular signature for TRAIL-induced apoptosis is likely a consequence of a strong cis-eQTL for Tnfsf10, encoded within the HIVAN1 locus, with the CAST allele associated with a nearly two-fold increased expression. Tnfsf10 encodes TRAIL, a cytokine involved in induction of apoptosis in transformed and tumor cells. The molecular signature for extracellular matrix components is also noteworthy, because the HIVAN1 Sub-IV interval contains Frem2, encoding a component of the extracellular matrix within the glomerular filtration barrier. The HIVAN1 congenic mice have reduced expression of Frem2, which may account for reduced expression of multiple extracellular matrix components, such as Col4a3 and Col4a4, encoded outside the HIVAN1 locus. In addition, HIVAN1 congenic kidneys harbor the signature of Polycomb target gene sets (histone H3 lysine 27 trimethylation), which marks repressed gene transcriptional programs observed in embryonic stem cells and poorly differentiated tumors [24]. Consistent with these data, the majority of these Polycomb targets show reduced expression in HIVAN1 congenic kidneys. In addition, this histone modification is associated with HIV-1 latency and reactivation [25,26]. Finally, analysis of whole kidney and the podocyte-enriched transcripts in HIVAN1 mice revealed significant overlap with molecular signatures of viral infection, particularly adenovirus ( Table 3, Table B in S2 Table) and Table B in S3 Table)). These data further suggest the presence of baseline molecular perturbations that may be magnified in the setting of HIV-1 infection.

Discussion
HIVAN and other forms of collapsing glomerulopathy have a complex determination and result from environmental insults (e.g. viruses or drugs) [27] as well as host genetic lesions [9,22,23]. Although mice do not have an APOL1 ortholog, the TgFVB mice recapitulates all of the clinical and molecular features of HIVAN [4][5][6], providing a model enabling for studying molecular mechanisms of glomerulosclerosis independent or downstream of APOL1. Murine susceptibility loci may also explain pathways leading to nephropathy in patients who do not harbor APOL1 risk alleles.
We had previously generated a HIVAN1 congenic mouse strain which carries a~52 Mb segment from CAST in the FVB genome [15]. This strain did not show any spontaneous signs of renal disease, but in the presence of the HIV-1 transgene, showed increased severity of nephropathy under an additive genetic model. Because large congenic intervals may contain multiple linked genes that may together contribute to the association with disease severity, we further dissected the HIVAN1 locus by generation of three new HIVAN1 sub-congenic strains. The two congenic strains carrying the distal portion of the HIVAN1 locus captured all the phenotypic severity of the original congenic strain, and refine HIVAN1 to a 7.4 Mb interval within the Sub-IV region. These data indicated that that the HIVAN1 QTL signal is not attributable to widely distant genes within the original interval. Annotation of the remaining positional candidates identified several genes that may contribute to disease. Three positional candidates have been implicated in kidney disease. Frem2 is expressed in adult glomeruli, collecting ducts and transiently expressed in nascent nephrons (tubule and podocyte epithelia) [28]. FREM2 is required for maintenance of the integrity of the skin epithelium in utero, for renal development and for the maintenance of renal epithelial structure in adult mice [19]. Mutations in the human ortholog cause Fraser syndrome, which features renal agenesis and cystic, dysplastic or hypoplastic kidneys. Although Frem2 haploinsufficiency does not overtly affect nephrogenesis in mice, expression of Frem2 in adult kidneys correlated with cyst formation in homozygous mutant mice, indicating that the gene is required for maintaining the differentiated state of renal epithelia [28]. The CAST strain harbors multiple linked non-synonymous variants in Frem2, and this gene is also differentially expressed in the congenic mouse kidney. This variation in Frem2 sequence and expression likely accounts for perturbed expression of multiple matrix components, such as Col4a3 orCol4a4, which are encoded outside the HIVAN1 locus (S1 Table).
The transcription factor Foxo1 has been implicated in progression of nephropathies of different etiology, including hypertensive and diabetic nephropathy [29]. A recent study showed that upregulation of Foxo1 expression in the kidney by transduction with recombinant lentivirus ameliorated podocyte injury and reduced severity of the symptoms in diabetic rats [20]. Foxo1may participate in the pathogenesis of HIVAN via multiple biological mechanisms including in cell cycle regulation [30], oxidative stress response [31,32] and inflammation pathways [33,34].
The Sub-IV interval also encodes SETD7, which plays a prominent role in lysine methylation of histone and non-histone proteins and is an important regulator of different transcription factors, including p53 [35], E2 promoter-binding factor 1 (E2F1) [36], the islet β cell factor PDX1 [37], NF-kB and others [38]. SETD7 can affect cell proliferation and apoptosis via co-activation of E2F1, modification of Wnt signaling, or regulation of β-catenin stability [39]. SETD7 is also a co-activator of HIV-1 transcription, which could contribute to the development of HIVAN: binding of SETD7 to HIV-1 TAR RNA and monomethylation of the viral transactivator Tat enhances HIV transcription [40]. A recent study also reported that SETD7 expression is associated with the degree of fibrosis in patients with IgA and membranous nephropathy and inhibition of SETD7 suppressed renal fibrosis in unilateral ureteral obstruction mice [21].
Ndufc1, encoding a subunit of the NADH dehydrogenase (complex I) in mitochondria, is highly enriched in podocytes and was also overexpressed in HIVAN1 congenic kidneys. Complex I enables electron transfer from NADH to Coenzyme-Q 10 and mutations in the biosynthetic pathway for CoQ 10 cause syndromic as well as isolated forms of nephrotic syndrome [22,23]. However, Ndufc1 is overexpressed in the congenic kidneys and together with the absence of perturbations of oxidative phosphorylation pathways, this reduces the likelihood that Ndufc1is the causal gene in the HIVAN1 interval.
We had previously hypothesized that HIVAN susceptibility loci introduce moderate genetic lesions that are tolerated, but are unmasked in the presence of the HIV-1 gene product [13]. Consistent with this hypothesis, analysis of apparently healthy HIVAN1 congenic kidneys, in the absence of HIV-1, demonstrated perturbations in many transcripts encoded outside the locus. Analysis of differentially expressed transcripts indicated overlap with multiple gene sets for apoptotic pathways and tissue matrix components. We also a signature for targets of Krüppel-like factor1, which belongs to a class of transcription factors that have been implicated in HIVAN and other forms of nephropathy [41][42][43]. Furthermore, we detected significant overlap with genes silenced Polycomb-group protein-mediated histone H3 lysine 27 trimethylation (H3K27me3). This chromatin modification is observed in embryonic stem cells and in poorly differentiated tumors [24]. Moreover, H3K27me3 is implicated in epigenetic silencing of HIV-1 long terminal repeats and regulation of viral latency [25,26]. Recent data also indicate that the HIV-1 Tat protein, which activates host programs that augment HIV-1 transcription, preferentially binds to host transcription start sites enriched for H3K27me3 marks [44]. Interestingly, we also detected a molecular signature of adenoviral infection, potentially indicating latent perturbations that may enhance susceptibility to viral injury. These data suggest a complex interplay between viral and host histone modification, and susceptibility to nephropathy.
In summary, analysis of congenic lines identified a number of plausible candidates that can single-handedly or cooperatively contribute to increased susceptibility to nephropathy. Transcriptomic analyses also suggested that HIVAN1 congenic kidneys may be poised for dysfunction, and exposure to appropriate triggers such as HIV-1 gene products may produce molecular decompensations that lead to overt kidney disease. The standard follow-up of these findings would involve generation of additional sub-congenic strains harboring smaller HIVAN1 segments to pinpoint the causal allele(s). In addition, newer mouse strains such as the Collaborative Cross or the Diversity Outbred strains offer a high resolution map of mouse haplotypes and may aid in refinement of QTL intervals [45][46][47]. Most importantly, the availability of CRISPR/Cas technology now allows rapid introduction of CAST alleles into the FVB germline, enabling assessment of phenotypic consequences of candidate sequence variants [48]. The combination of these approaches is expected to accelerate the identification of causal alleles contributing to kidney disease in mouse models.