The vector VHL-wild-type (WT)-Venus-Retro (26) and the Quikchange II XL Site-Directed Mutagenesis Kit were used following manufacturer’s protocols to perform the specific mutations P138R (CCA➔CGA) and L163R (CTC➔CGC). Mutations were verified by DNA sequencing in ABI PRISM 310 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA).

HEK293T cells were used as a helper cell line in order to obtain retrovirus with the desired vectors as previously described by Ding et al. (27). Briefly, HEK293T cells were transfected with 3 different vectors: 1) pcGp, 2) pVSVG, and 3) either one of the following: GFP-Retro/VHL-WT-Venus-Retro/VHL-P138R-Venus-Retro/VHL-L163R-Venus-Retro using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA). Upon assembly, supernatant was used to infect RCC 786-0 cells (ATCC^® CRL-1932™, American Type Culture Collection, Manassas, VA, USA), and after 20h, selection was performed with 1 mg/ml of G418 antibiotic (Sigma Aldrich, St. Louis, MO, USA). Four different cell lines were obtained expressing green fluorescent protein (GFP), VHL-WT-Venus, VHL-P138R-Venus, and VHL-L163R-Venus. All cell lines were cultured in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and maintained at 37°C in a humidified 5% CO₂ environment.

Proteins were obtained as previously described (28) and resolved on a 12.5% sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE). After transferring to polyvinylidene fluoride (PVDF) membranes, blots were blocked and probed with different primary antibodies: VHL (BD Biosciences, # 556347, diluted 1/5,000), GFP (Santa Cruz, sc-8334, diluted 1/1,000), HIF-2α (Novus Biologicals, NB100-122, diluted 1/1,000), β-actin (Cell Signaling, #4970, diluted 1/1,000), Elongin B (Santa Cruz, sc-133090, diluted 1/500), and Elongin C (Santa Cruz, sc-1559, diluted 1/500). The following secondary antibodies were used accordingly: anti-rabbit (Cell Signaling, #7074, diluted 1/5,000), anti-goat (Santa Cruz, sc 2020, diluted 1/2,000), and anti-mouse (Cell Signaling, #7076, diluted 1/2,000).

Cell lines were seeded on 6-well plates and incubated with 50 µg/ml cycloheximide to interfere with protein synthesis, or 5 µg/ml MG132 to inhibit the proteasome, or 100 µM CoCl₂ (29) to simulate hypoxia. After treatment, proteins or RNA was extracted.

The amount of protein coming from GFP, WT VHL-Venus, P138R VHL-Venus, and L163R VHL-Venus cell lines was determined by Bradford assay, and 1 mg of protein was immunoprecipitated using GFP-Trap^®_A kit (Chromotek GmbH, Germany). The immunocomplexes were detected by Western blot using the antibodies described above. Protein from WT VHL-Venus cell line was used as positive control and that from the cell line expressing GFP as a negative one.

Total RNA from the different cell lines was extracted with Direct-Zol RNA Kit (Zymo Research, Irvine, CA, USA) following manufacturer’s protocol. To perform RT-qPCR, 1 µg of RNA from each sample was used together with random hexamers and Super Script II (Invitrogen, Carlsbad, CA, USA). Resulting cDNA was diluted by 1:10, and 3 μl from each dilution was subject to qPCR in triplicate using Kapa Syber Fast qPCR master mix (Kapa Biosystems, Boston, MA, USA) in Step One Plus Real-Time PCR System (Life Technologies, Carlsbad, CA, USA). mRNA values were calculated using relative quantitation method and are presented as fold change compared to control conditions. Specific primers were designed to assess fibronectin, vascular endothelial growth factor A (VEGF-A), and glucose transporter 1 (GLUT1) normalized to TATA box-binding protein (TBP) or VHL and α subunit α of HIF-2 (HIF-2α) normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH).

Using anti-fibronectin antibody combined with a secondary antibody conjugated with Cy5, matrix deposition by all cell lines was analyzed according to the protocol of Debnath et al. (30). Briefly, cells were plated on coverslips, fixed, and permeabilized after 6 days of culture. Nuclei were dyed with Hoechst (5 µg/ml), and pictures were taken on a Carl-Zeiss AxioScope A1 microscope.

Immunodeficient mice [N:NIH (S)-Fox 1^nu] were housed in standard conditions of 12-h light/12-h dark cycle with water and food ad libitum, in accordance with National Institutes of Health guide for the care and use of laboratory animals (31).

A solution of 1 × 10⁷ viable cells was injected subcutaneously on 6–8-week-old male mice and monitored weekly for tumor development. At 16 weeks post cell injection or when tumor reached 2-cm diameter, mice were sacrificed, and tumor histology was evaluated by hematoxylin and eosin (H&E) staining.

All animals were treated and cared for in accordance with standard international animal care protocols. All procedures were approved by the Animal Care and Use Committee of the Hospital de Niños Dr. Ricardo Gutiérrez.

We searched for these variants in the Genome Aggregation Database (gnomAD) (32), dbSNP (33), and ClinVar (34) databases to look at allele frequency, and if they had been reported by other groups. We also used online tools that predict the effect of protein variants: SIFT (35), Polyphen (36), Mutation Taster (37), and Human Splicing Finder (38). To classify these variants according to the American College of Medical Genetics Guidelines (39), we used VarSome (40).

The crystal structure of a human VBC: HIF-1α complex PDB 4AJY (X-Ray diffraction, 1.73 Å resolution) was used as starting structure (41). Missing residues of EloC (amino acids 106–118) were added using the SWISS-MODEL workspace (42, 43). The following six macromolecular systems were considered: WT and P138R and L163R variants of pVHL inserted in VBC: HIF-1α complexes, both under normoxia or hypoxia (the latter simulated replacing Hyp564 by Pro564 in HIF-1α). Lacking experimental structures of the two variants considered, in silico mutations were introduced by replacing the residue of interest at the native structure using the SWISS-PDB Viewer software (42). Protonation states of titratable residues were determined with PROPKA 3.0 (44), then all missing hydrogen atoms were added with the ProToss utility of the Proteins Plus server. All the systems were solvated with a truncated-octahedral box of TIP3P water 12 Å around the solute and neutralized with K⁺ ions using the leap module of AmberTools17 (45). Each of the systems was minimized (2,000 steps applying a 500 kcal mol⁻¹ Å⁻² harmonic potential over solute atoms, followed by 20,000 steps without restraints), then heated to 310 K [500 ps molecular dynamics (MD) simulation in NVT ensemble] and equilibrated at 1 atm (1 ns MD simulation at 310 K in NPT ensemble), prior to run 400 ns of productive MD simulations (NPT, 310 K and 1 atm). Minimizations and MD simulations were carried out with the pmemd.cuda module of AMBER16 (45). Protein residues were treated using the AMBER ff14SB force field. An integration step of 2 fs was used, constraining bonds involving hydrogen with SHAKE algorithm (46). Temperature and pressure were controlled applying the Langevin thermostat (47) and the Monte Carlo barostat (48), respectively. An 8.0-Å cutoff was used for direct non-bonded interactions, and the Particle Mesh Ewald (PME) method (49) was applied to long-range electrostatic interactions. Trajectory processing and analysis were performed with cpptraj module of AmberTools 17. Trajectory convergence was monitored following Cα-RMSDs, and flexibility was examined by means of per-residue Cα-RMSF. Snapshots of the trajectory were clustered into 5 clusters—each one with a representative structure—using a hierarchical agglomerative algorithm. Binding free energies of HIF-1α to the VBC complex were calculated using the MM-PB(GB)SA methods (50). For those calculations, the first 50 ns of the trajectories were discarded, then 100 snapshots separated by 3.5 ns were used. Representative structures of clusters with appreciable population (>10%) were used to calculate the electrostatic potential of VBC using the APBS software (51) implemented in the APBS/PDB2PQR web server (52).

For real-time PCR analysis, one-way ANOVA was used with a Tukey test post evaluation. The chi-square test was used to analyze the differences in tumor incidence, and crosstabs were created. Statistical significance was defined as a p-value <0.05, and all data were graphed as mean ± standard deviation unless indicated otherwise.

We analyzed the effect of P138R and L163R novel variants on VHL protein stability using Venus-tagged proteins. Human 786–0 RCC cell line (VHL-deficient) was infected with retroviral vectors to stably express VHL-P138R-Venus, VHL-L163R-Venus, and VHL-WT-Venus. Protein levels for both variants were significantly lower than those for VHL-WT-Venus ( Figure 1A ). Assessed by RT-qPCR, mRNA levels showed that VHL-P138R-Venus and VHL-L163R-Venus variants were similar and even higher than VHL-WT-Venus mRNA levels ( Figure 1B ), suggesting that transcription levels are not responsible for the differences in protein levels evidenced by Western blot.

Cell lines were treated with cycloheximide to inhibit protein translation and enable the determination of half-lives for both VHL variants and WT pVHL. After 6 h, results showed that VHL-P138R-Venus and VHL-L163R-Venus have a significantly shorter half-life (≈1.2 h and 1 h, respectively) compared to that of VHL-WT-Venus (≈3.4 h) ( Figure 1C ).

Inhibiting the proteasome with MG132 (proteasome inhibitor) significantly increased both variants’ protein levels, achieving quantities comparable to WT pVHL levels after MG132 treatment for the case of P138R and slightly lower for L163R ( Figure 1D ).

To date, pVHL’s most described function is its interaction and consequent downregulation of HIF-α protein subunits (53). To this end, pVHL needs to form the VBC complex (pVHL-Elongin B-Elongin C). Immunoprecipitation of GFP Trap showed a specific band of 25 kD for GFP alone and 50 kD on cells expressing GFP-pVHL-Venus Tag ( Figure 2A ). Consistent with previous results ( Figure 1A ), pVHL levels are different for the WT and P138R and L163R variants, resulting in less coimmunoprecipitation of Elongin B and C for the variants compared to WT VHL cell line ( Figure 2A ). We calculated the ratio between the bands obtained: Elongin C/pVHL and Elongin B/pVHL for WT pVHL, P138R and L163R pVHL-expressing cell lines. Ratios were normalized to WT pVHL’s set as 1, and we observed that P138R immunoprecipitates less Elongin B and Elongin C (approximately 0.6) and L163R manages to immunoprecipitate a similar proportion of Elongin C but a lower quantity of Elongin B (0.25).

Since VBC complex was evidenced for both variants, we sought to evaluate its functionality. Firstly, the capacity of pVHL variants to downregulate HIF-2α was assessed. HIF-2α is overexpressed in the parental cell line used (786-0) (54), and its levels decrease significantly in the derived cell line expressing VHL-WT-Venus ( Figure 2B , lanes 1 and 2). Protein levels for both P138R and L163R cell lines ( Figure 2B , lanes 3 and 4) were intermediate for HIF-2α assessed by Western blot, although mRNA levels did not change in the different cell lines ( Figure 2B ). To evidence the consequence of these intermediate levels of HIF-2α protein, we quantified mRNA levels of two of its downstream targets: VEGF-A and GLUT1 using qRT-PCR in normoxic and pseudohypoxic conditions ( Figure 2C ). Despite different HIF-2α protein levels, mRNA levels in normoxia for VEGF-A and GLUT1 were similar among cell lines expressing WT and P138R and L163R pVHL ( Figure 2C , upper panel). Under pseudohypoxic conditions, we found significantly higher levels of VEGF-A and GLUT1 mRNAs on the variant cell lines compared to the one expressing WT pVHL ( Figure 2C , lower panel).

pVHL is known to regulate fibronectin mRNA levels, although the underlying molecular mechanism has not been yet described. We assessed fibronectin mRNA levels in the 786-0 and 786-0-derived cell lines expressing VHL-WT-Venus, VHL-P138R-Venus, and VHL-L163R-Venus by RT-qPCR. Cells expressing WT-VHL have higher fibronectin mRNA levels than the parental 786-0, which is pVHL null ( Figure 3A ). Regarding the variants, P138R expression shows similar fibronectin mRNA levels to that of WT-VHL-expressing cell line. On the other hand, L163R expression resulted in diminished fibronectin mRNA levels and significantly different to the WT-VHL but comparable to the levels obtained for 786-0 cell line ( Figure 3A ).

Fibronectin expression per se does not ensure its proper extracellular matrix organization. Using immunofluorescence, we evidenced fibronectin deposition in the 786-0 cell line as a dotted pattern, while in VHL-WT-Venus resulted in fibrillar network of fibronectin deposition ( Figure 3B ). Both variants, P138R and L163R, failed to generate this fibrillar organization, demonstrating a pattern similar to that observed in the parental 786-0 cell line where pVHL is absent ( Figure 3B ).

To test the tumor suppressor role of the novel variants, we injected the cell lines expressing WT-VHL and P138R and L163R pVHL into male nude mice. Also, 786-0 cell line was injected as an internal control for the experiments. In our hands, visible tumors were developed, on average, 9 weeks after injection for all the tested cell lines ( Figure 4A ).

As expected, the ratio between the number of tumors developed and the number of sites injected was significantly higher in 786-0 compared to the cells expressing the WT-VHL protein. Moreover, P138R and L163R pVHL-expressing cells developed more tumors when compared to WT-VHL cell line ( Figure 4B ). Contingency tables were obtained, showing a significant difference between P138R, L163R, or 786-0 cells with WT pVHL, where tumors developed in 55% (11/20 for both variants) or 40% (4/10 for 786-0 cells) of the sites injected compared to a 10% for WT pVHL (3/30) ( Figure 4B ). Also, the variants showed a similar ratio of developed tumors to that of the parental cell line.

H&E staining confirmed that developed tumors had histological characteristics that are compatible with clear cell renal carcinoma ( Figure 4C ). These solid tumors were composed of atypical polyhedral cells that have a large, acidophilic, or optically empty cytoplasm with large nuclei where its membrane was observed thickened and a prominent central nucleolus. Cells are grouped into clusters separated by thin collagen tracts through which small blood vessels pass ( Figures 4C, I, II). Tumors had infiltrating growth toward neighboring tissues ( Figures 4C, III) and showed histological signs of proliferative activity, evidenced by the numerous mitotic figures found ( Figures 4C, IV).

pVHL protein expression was verified on tumors developed by 786-0 cells, WT pVHL, P138R, and L163R cell lines by Western blot. As shown in Figure 4D , pVHL was not detectable on 786-0 cells and had higher levels on WT pVHL-expressing cells compared to both variants (P138R and L163R).

The results of our database and online prediction tools are summarized in Table 1 .

Our variants were not found in the Genome Aggregation Database (gnomAD) that includes thousands of genomes and exomes; this information allows us to infer that they have a very low allelic frequency. Most of the effect prediction tools used suggest that both variants are deleterious. L163R was previously reported by our group and reported in ClinVar by a genetic testing laboratory that classifies it as a variant of unknown significance (VUS). Using VarSome to follow the ACMG guidelines for classification of new variants, they are classified as likely pathogenic (P138R) and pathogenic (L163R).

MD simulations enabled us to inspect at a molecular level the effects of introducing P138R and L163R pVHL variants in the VBC: HIF complex structure ( Figure 5A ) and stability, flexibility of the protein components, and other features relevant toward molecular recognition of pVHL by HIF (here represented by a 559-577 peptide fragment from HIF-1α containing either hydroxyproline Hyp564 or P564 in a carboxyl-terminal oxygen-dependent CODD motif, as representative of normoxia and hypoxia, respectively) in the VBC complex and by other possible interactors ( Figure 5A ).

All of the six MD 400-ns simulations promptly converged, showing formation of structurally stable complexes in all the cases. Introducing variants P138R and L163R in pVHL ( Figure 5B ) appears not to considerably disrupt HIF-1α binding to VBC under normoxic conditions: as shown in Table 2 , the three complexes display similar binding strength values. Although VBC: HIF-1α complexes still form as evidenced in vitro ( Figure 2A ), binding strength is significantly reduced in all the cases under hypoxia, particularly for variant P138R ( Table 2 ).

Global structural fluctuations in protein backbones appear to be smaller under hypoxia (when HIF-1α Hyp564 is replaced by P564) with respect to normoxia (See Figure S1 in the Supplementary Material ). Differences in dynamic behavior among WT and P138R and L163R variants of pVHL are more pronounced under conditions representative of normoxia and accompanied by side-chain shifts in residues relevant for the pathophysiological functions of pVHL.