Apolipoprotein L1-Specific Antibodies Detect Endogenous APOL1 inside the Endoplasmic Reticulum and on the Plasma Membrane of Podocytes

Significance Statement

Specific variants of APOL1, G1 and G2, are associated with CKD in the Black population. Overexpression of these variants kills cells, through different proposed mechanisms in different subcellular compartments. The localization of endogenous APOL1 has not been conclusively established because all studies have used antibodies that crossreact with APOL2. Generation and use of APOL1-specific antibodies show that endogenous podocyte APOL1 localizes mainly inside the endoplasmic reticulum, with a few molecules on the cell surface. These findings potentially support the endoplasmic reticulum stress or cell surface cation channel models of cytotoxicity.

Abstract

Background

APOL1 is found in human kidney podocytes and endothelia. Variants G1 and G2 of the APOL1 gene account for the high frequency of nondiabetic CKD among African Americans. Proposed mechanisms of kidney podocyte cytotoxicity resulting from APOL1 variant overexpression implicate different subcellular compartments. It is unclear where endogenous podocyte APOL1 resides, because previous immunolocalization studies utilized overexpressed protein or commercially available antibodies that crossreact with APOL2. This study describes and distinguishes the locations of both APOLs.

Methods

Immunohistochemistry, confocal and immunoelectron microscopy, and podocyte fractionation localized endogenous and transfected APOL1 using a large panel of novel APOL1-specific mouse and rabbit monoclonal antibodies.

Results

Both endogenous podocyte and transfected APOL1 isoforms vA and vB1 (and a little of isoform vC) localize to the luminal face of the endoplasmic reticulum (ER) and to the cell surface, but not to mitochondria, endosomes, or lipid droplets. In contrast, APOL2, isoform vB3, and most vC of APOL1 localize to the cytoplasmic face of the ER and are consequently absent from the cell surface. APOL1 knockout podocytes do not stain for APOL1, attesting to the APOL1-specificity of the antibodies. Stable re-transfection of knockout podocytes with inducible APOL1-G0, -G1, and -G2 showed no differences in localization among variants.

Conclusions

APOL1 is found in the ER and plasma membrane, consistent with either the ER stress or surface cation channel models of APOL1-mediated cytotoxicity. The surface localization of APOL1 variants potentially opens new therapeutic targeting avenues.

ESKD disparately affects Blacks compared with European Americans, in large part due to risk variants G1 and G2 in the APOL1 gene (wild type being designated G0) that are associated with nondiabetic kidney diseases.^1–2 Generation of therapeutics for these APOL1 nephropathies first requires a better understanding of the mechanism of action of APOL1.^3,4

Evidence supports that expression of APOL1 in kidney podocytes, rather than circulating APOL1, is likely responsible for kidney disease.^3–7 In vitro overexpression of either APOL1 risk variant is usually more cytotoxic than G0 in many cell types, including podocytes.^8–11 More compellingly, APOL1 variants induced to high levels specifically in mouse podocytes in vivo lead to proteinuria and glomerular pathology resembling human FSGS.⁸

A variety of putative mechanisms have been proposed to explain APOL1 variant-dependent cytotoxicity, all involving different subcellular compartments.⁴ These include endoplasmic reticulum (ER) stress; mitochondrial dysfunction^12,13; enhanced cell surface suPAR/integrin binding¹⁴; surface cation efflux channel activity^11,15; cholesterol accumulation¹⁶; and defects in endosomal maturation, autophagosome flux, or lysosomal permeability.^{8–9,11,17–19} Accordingly, APOL1 has been variously immunolocalized to the ER,^12,20,21 the mitochondria,^12,13,22 lipid droplets,²³ early and late endosomes,^8,22,24 and by biochemistry to the plasma membrane.^11,15 However, most studies used overexpressed APOL1, which may not traffic normally, and the antibodies were not characterized for crossreactivity with related APOL family members. Those antibodies reportedly stained endogenous APOL1 in podocytes, endothelial cells, and proximal tubules of human kidneys,^25,26 but we found that they also recognize APOL2, thereby calling into question whether APOL1 is truly expressed there.

Here, we characterized 80 antibodies specific to APOL1 by immunofluorescence (IF) and demonstrate that APOL1 is expressed in endothelial cells and podocytes, specifically in the podocyte ER and plasma membrane, although two N-terminal splice isoforms are cytoplasmic. Our data thus potentially favor the ER stress or surface cation channel models of cytotoxicity.

Methods

Cell Culture

Human immortalized podocytes (AB 8/13) were obtained under license from Moin Saleem²⁷ and were not genetically confirmed in our laboratory, although we periodically verified that they stained normally by IF for podocin and synaptopodin after differentiating at 38°C for 14–15 days. They were grown at 33°C, feeding 3× a week with fresh growth medium (RPMI with 10% FBS [IXL9/06807/GEN; Seradigm]; 1% glutamine; 1% pen/strep [15140–122; Gibco]; 1% insulin, transferrin, selenium [41400–045; Gibco]), and discarded after passage 29 due to loss of APOL1 expression with time. Induction of endogenous APOL1 with 100 ng/ml human IFNγ (285-IF, IFNγ; R&D Systems) was done for approximately 24 hours unless otherwise stated; the level of induction was similar whether induced for 24, 48, or 72 hours (see Figure 8A).

Figure 8.

APOL1 isoform expression in podocytes. (A) Western blot of WT and APOL1 KO podocytes stimulated for 0–4 days with 100 ng/ml IFNγ. Lysates were immunoblotted for APOL1 as in Figure 7D, then reprobed with 1A2 anti-tubulin as a loading control. IFNγ treatment increases the level of APOL1 and a faint upper band also is reproducibly detected on 10% or 12% Tris-Glycine gels (12% here) at the right loading level. Note that there is no band smaller than the major band, suggesting that vC is not detectable. (B) The upper APOL1 band in IFNγ-stimulated podocytes is the same size as vB3. Western blot of APOL1 isoform stable podocyte lysates from Figure 7D loaded adjacent to 9 or 17 µg of WT IFNγ-stimulated podocyte lysate and a larger amount of vB3 in the last lane. The upper band of endogenous APOL1 is similar in size to that of APOL1.vB3, although we cannot rule out the possibility that it is a post-translational modification. From the molecular weight predictions (Supplemental Table 1B) it could not represent any of the other isoforms. The band beneath vA in WT podocytes (*) is probably a degradation product, because it is only sporadically detected (compare [A]).

APOL1 CRISPR knockout (KO) podocytes and generation of doxycycline-inducible iAPOL1-G0 podocyte and CHO stable cell lines are described in the accompanying manuscript (Gupta et al.²⁸). The same methods were used to generate iAPOL1 podocytes stably transfected with doxycycline-inducible cDNAs (i.e., nonspliceable) for each N-terminal isoform (see Supplemental Table 1B) and APOL1-G1 and -G2. These podocytes require tetracycline-free FBS and 5 µg/ml puromycin in the medium and were induced with 5 or 10 ng/ml doxycycline unless otherwise stated.

STR-validated COS7 cells were obtained from the Genentech cell culture facility and maintained in high-glucose DMEM, 10% FBS, 1% glutamine, and 1% nonessential amino acids (M7145; Sigma). Transient transfection of APOL cDNAs (listed in Supplemental Table 1) was achieved by plating 0.9×10⁴ COS7 cells/well in eight-well LabTekII slides for 24 hours, then adding premixed 0.7 µl of Fugene HD (E231A; Promega) with 0.25 µg of DNA per well for 2 days. JHH-1 cells were cultured as in Gupta et al.²⁸

APOL Family cDNA Cloning

All available APOL cDNAs were obtained as summarized in Supplemental Table 1 (APOL5 was not commercially available). APOL1 Δss consists of APOL1.vA with its 27-aa signal sequence deleted, whereas ssAPOL2 comprises APOL2 with the 27-aa signal sequence of APOL1.vA fused to its N terminus. Where indicated, subcloning into mammalian expression vectors with C-terminal myc-FLAG tags (Origene) or C-terminal myc only for APOL4 (Genecopoeia) was achieved using EcoRI and XhoI restriction sites. The APOL1 splice isoforms were all untagged. All constructs were verified by Sanger sequencing.

Commercial Antibodies and Their Validation

Commercial anti-APOL1 antibodies were rabbit polyclonals from Sigma (HPA018885, lots E103963 and E105900, used at 0.4 µg/ml for IF and 2 µg/ml for western) and Proteintech (11486–2-AP, lot 00048412, used at 1 µg/ml for IF and 0.44 µg/ml for western), and rabbit monoclonal (rabmab) EPR2907(2) lot GR145110–2 from Epitomics at 2 µg/ml (for both IF and western). Compartment marker antibodies were rabbit anti-calnexin (Ab25595, 0.25 µg/ml; Abcam [validated in our laboratory by colocalization with GFP-Calnexin (data not shown) and subsequently shown by Abcam to give no signal in calnexin KO cells]), rabbit anti-PDI (protein disulfide isomerase; 1:200 SPA-890; Stressgen [validated by colocalization with calnexin]), mouse anti-transferrin receptor H68.4 (1:250 Invitrogen 13–6800 [validated by colocalization with internalized Alexa488-transferrin ligand (data not shown)]), rabbit anti–perilipin-2 (15294–1-AP at 1.74 µg/ml for IF and 0.08 µg/ml for western; Proteintech [validated in Supplemental Figure 7]), mouse anti–perilipin-2 clone ADFP-5 (3.3 µg/ml, SAB4200452; Sigma [validated in Supplemental Figure 7]), goat anti-EEA1 (1 µg/ml C15, Sc-6414 lot K1904 for early endosomes [validated by colocalization with mouse anti-EEA1]), rabbit anti-LAMP1 cytoplasmic domain (1 µg/ml Novus NB120–19294 Lot MF159260 [validated by colocalization with mouse anti-LAMP1]), mouse anti-mitofusin 2 clone 6A8 (Ab56889; Abcam) at 1 µg/ml, rabbit anti-FACL4 (Ab137525; Abcam) at 1 µg/ml (both confirmed by IF to localize to mitochondria [Supplemental Figure 16B]), mouse anti-KDEL antibody clone 10C3 for immunoelectron microscopy (iEM) of the ER lumen (420400; Calbiochem) at 1:70, and the following mouse monoclonals all from BD Transduction Laboratories: calnexin clone 37 (2.5 µg/ml 610523 [validated by colocalization with rabbit anti-calnexin and Calnexin-GFP]), LAMP1 H4A3 (0.5 µg/ml 555798 [validated by colocalization with GFP-LAMP2a²⁹]), cytochrome c 6H2 (2.5 µg/ml 556432 [validated by its surrounding of mitochondrially targeted GFP by us²⁹ and by colocalization with MitoTracker³⁰]), mouse anti-TOM20 (612278 at 1:1000, not validated other than by its enrichment in the mitochondrial fraction [Supplemental Figure 11A]), GM130 clone 35 (1 µg/ml 610823), p230/Golgin-245 (1 µg/ml 611280 [validated by proximity to GM130]), and EEA1 (early endosome antigen 1) clone 14 (1 µg/ml 610457 [validated by overlap with a 5-minute pulse of EGF and transferrin]). Other antibodies were rabmab anti-myc 71D10 (1:200, 2278; Cell Signaling Technology), mouse anti-FLAG M2 (F3165, 1 µg/ml; Sigma), mouse anti-myc 9E10 (2 µg/ml; Genentech), β-Actin–HRP 13E5 (5125, 1:1000; Cell Signaling Technology), mouse anti-tubulin 1A2 (T9028, 1:10,000; Sigma), and rabbit anti-calnexin (1:3000 SPA-860 for western blot; Stressgen).

Anti-APOL1 Mouse mAb Generation

Mouse monoclonals were generated to his₆-APOL1 (aa 61–398 of NM_003661) as described in the accompanying paper.²⁸ The best 35 mouse monoclonals were sequenced and cloned into a murine IgG2a expression vector with effectorless L234A, L235A, P329G (LALA-PG) mutations as previously described,^31,32 with 22 cloned antibodies successfully recognizing APOL1. To enable triple labeling with other species, the Fab regions of some of the best-cloned antibodies were subcloned into rabbit IgG and/or rat IgG2b backbones for expression in CHO cells and purification by protein-A affinity chromatography. The unpaired cysteines in the reformatted rabbit antibodies were paired by mutation of residue 80 to cysteine (P80C in rabbit 3.7D6, and A80C in rabbit 3.1C1 and 3.1C7).

Anti-APOL1 Rabmab Antibody Generation

Two rabbits were immunized five times each (at Lampire Biological Laboratories) with 0.5 mg of the above his₆-APOL1 antigen, except the fourth and fifth boosts which were with fixed antigen in order to maximize IF reactivity. For fixation, paraformaldehyde from a 16% stock was diluted into 1 mg of his₆-APOL1 to 3% final for 20 minutes at room temperature and 10 minutes on ice, then PFA was removed by dialysis in APOL1 buffer (20 mM Tris-Cl pH 7.5, 150 mM NaCl, 0.1% N-dodecyl-β-d-maltoside) in a 10,000-MWCO membrane (Pierce).

Rabmabs were generated from APOL1-immunized rabbits by the B-cell cloning technology. PBMCs were isolated from rabbit whole blood after Ficoll density centrifugation. The PBMCs were washed in PBS, then resuspended in FACS buffer (2% FCS in PBS) and stained with FITC-labeled anti-rabbit IgG (AbD Serotec; BioRad). After washing, the cells were resuspended in FACS buffer containing propidium iodide. Live, single IgG-positive B cells were sorted into 96-well culture plates using a FACSAria sorter (BD Biosciences). After the B cells were cultured in B-cell culture medium (Genentech) for a week, the supernatants were screened by ELISA for APOL1 binding. The B cells in each well were lysed in 100 µl of RLT buffer (Qiagen, Hilden, Germany) and immediately frozen at −80°C for storage until molecular cloning. Variable regions (VH and VL) of each ELISA-positive mAb were cloned into expression vectors from extracted mRNA as previously described.³³ Individual rabbit antibodies were recombinantly expressed in 1 ml of Expi293 cells and were subsequently purified with a protein A–type resin in a high-throughput manner. IF-positive clones were selected for scale up (30 ml) and were purified on a protein A column followed by size exclusion chromatography to remove any protein aggregates.

Antibody Screening

Mouse mAbs that recognized APOL1 by ELISA were screened for IF by testing at 1 µg/ml on COS cells transiently expressing APOL1 or APOL2 with the methanol protocol (−20°C for 5 minutes), and the positives retested with PFA/Triton X-100. ELISA-positive rabmabs were primarily screened with PFA/Triton X-100 instead of methanol because they were raised against PFA-fixed APOL1. To rank the antibodies by sensitivity, lower-expressing iAPOL1-CHO stables²⁸ were stained with the PFA/Saponin method, because that works best for most subcellular organelles. The strongest few were used to stain endogenous APOL1 in podocytes.

RT-PCR

Flash-frozen normal postmortem human kidneys were obtained under agreement from the University of Michigan (three women, aged 47, 50, and 62; and two men, aged 62 and 63). Total RNA was extracted from tissue or cells using RNeasy mini kit (Qiagen, Valencia, CA). On-column genomic DNA digestion was performed during purification using an RNAse-free DNAse set (Qiagen) and cDNA synthesis was done using the High Capacity Reverse Transcription Kit (Applied Biosystems, Foster City, CA) with the provided random hexamer primers. Quantitative PCR reactions were performed by Taqman analysis in triplicate, using APOL-specific probes (ThermoFisher Scientific) on an ABI PRISM sequence detection system (Applied Biosystems). We validated the APOL family probe specificities (catalog numbers: L1 is Hs1066280_m1; L2 is Hs00603146_m1; L3 is Hs00600896_m1; L4 is Hs00540930_m1; L5 is Hs00229052_m1; and L6 is Hs00229051_m1) on cDNAs encoding all of the available family members (Supplemental Figure 1A) and the isoform specificities of the different APOL1 probes are shown in Supplemental Figure 20B. Note that the APOL1 probe is somewhat weak, only 10zz, unlike an earlier 20zz version that was rejected due to crossreactivity with APOL2. Relative mRNA was calculated using the 2^-ΔCt method normalized to RPL19. Data were plotted using PRISM.v7 software.

In Situ Hybridization (RNAscope)

Formalin-fixed and paraffin-embedded (FFPE) normal postmortem kidneys (from men aged 47 and 59) were obtained under agreement from the MT group (Van Nuys, CA) with ethics committee approval and informed consent. Kidneys or cell pellets sectioned onto slides at 5 µm were loaded into a Leica BondRx autostainer and baked and dewaxed. Slides were pretreated with ER2 at 95°C for 15 minutes and Leica Enzyme at 40°C for 15 minutes. Probes (all RNAscope LS 2.5 probes from Advanced Cell Diagnostics) were hybridized at 42°C for 120 minutes: APOL1 (Hs-APOL1–02, cat# 569168 [green] or 569168-C2 [red], nt 768–1431 of NM_003661.3, 10zz), Hs-NPHS2 (cat# 556538, nt 268–1172 of NM_014625.3, 20zz), and Hs-EGLF7 (cat# 314008 [green] or 314008-C2 [red], nt 14–1356 of NM_016215.4, 14zz). RNAscope 2.5 LS Duplex Reagent kit (cat# 2005835) was used for detecting signal. We verified that the APOL1 probe did not crossreact with APOL2 and was capable of detecting endogenous APOL1 in podocytes (Supplemental Figure 1B). Note that an earlier version of the APOL1 probe (cat# 411578 nt 2–902 of NM_001126541.1) did crossreact with APOL2 (and has thus been discontinued) and that probe did stain glomerular endothelia (data not shown). PLVAP protein was stained with antibody 10A2 (Genentech). Sections were counterstained with light Mayer’s hematoxylin, dried at 60°C for 1 hour, and mounted with Histomount.

IF

Cells were plated on LabTekII (154534; Nalge Nunc) slides and fixed at 70%–100% confluency. For mAb screening, cells were fixed/permeabilized for 5 minutes in −20°C methanol (EMD MX0485–6). For all other experiments, cells were fixed in 3% paraformaldehyde in PBS (EMS 15710) for 20 minutes at room temperature, quenched with 50 mM NH₄Cl in PBS for 10 minutes, then permeabilized with 0.1% TX-100 in PBS (or digitonin in KHM, see below) for 4 minutes or for 1 hour in saponin buffer (0.4% saponin [S7900; Sigma], 1% BSA [A2153; Sigma], 2% FBS [Seradigm] in PBS]. First antibodies (stored sterile at 4°C) were applied for 1 hour at room temperature, at 1 µg/ml for screening on transfected cells or at subsequently optimized concentrations for detecting endogenous APOL1. After 3× 10-minute washes in PBS (for methanol or TX-100 protocols) or saponin buffer (saponin protocol), secondary antibodies were applied at 1.88 µg/ml for 1 hour at room temperature, washed 3×, and mounted under No. 1.5 coverslips in ProLong Gold with DAPI (P36931, 4′,6-diamidino-2-phenylindole; Life Technologies). Secondary antibodies (stored at −20°C in 50% glycerol), all highly crossadsorbed donkey F(ab′)₂ anti H&L from Jackson ImmunoResearch, included Alexa488 anti-mouse (715–546–150), Alexa488 anti-mouse additionally crossadsorbed against rat (715–546–151), Alexa488 anti-rabbit (711–546–152), Cy3 anti-mouse (715–166–150), Cy3 anti-rabbit (711–166–152), Cy3 anti-rat (minimally crossreactive with mouse, 712–166–153), Cy3 anti-goat (705–166–147), Alexa647 anti-rabbit (711–606–152), Dy649 anti-rabbit (712–496–153), and Alexa647 anti-mouse (715–606–150, or 715–606–151 for costaining with rat antibodies). For triple labeling of ER and mitochondria, isotype-specific mouse secondaries were used: Alexa488 anti-IgG2a (A21131; Invitrogen) for murine 4.17A5, Alexa555 anti-IgG1 (A21127; Invitrogen) for cytochrome c, and Alexa649 anti-rabbit for calnexin. Alternatively rat 4.17A5 and Cy3 anti-rat were used (Supplemental Figure 15).

For selective permeabilization of the plasma membrane, the concentration of digitonin was first optimized by ensuring that free GFP leached out of the cells prefixation, whereas ER-targeted GFP-Calnexin with luminal GFP²⁹ was only detectable by GFP fluorescence and not anti-GFP antibodies in the red channel (data not shown). The optimal concentration for COS cells was 0.0025% (20 µM) and was applied in KHM buffer (110 mM potassium acetate, 20 mM Hepes pH 7.4, 2 mM MgCl₂) for 4 minutes in place of TX-100 after fixation (see above).³⁴

COS cells were imaged on an AxioM2 imaging system (Zeiss) with a 60× PlanAPO NA 1.4 objective, standard DAPI/FITC/Cy3/Cy5 filter sets, and a PhotoMetrics HQ2 camera managed by SlideBook (v5.5). Podocyte and CHO cells were imaged by spinning disc confocal microscopy (3i W, Zeiss AxioObserver M1 microscope with a Yokogawa W1 spinning disc) with a 63× PlanAPO NA 1.4 oil objective and 405-, 488-, 561-, and 640-nm lasers powered by SlideBook (v6). The camera was a Hamamatsu FLASH 4.0 sCMOS. Images were assembled in Adobe Photoshop CC2019. RGB profiles were generated in Fiji using the Color Profiler plug-in.³⁵

Immunohistochemistry

For antibody screening, a mini array of FFPE cell pellets was generated, including untransfected CHO cells, doxycycline-induced iAPOL1-CHO stable cell lines,²⁸ transiently transfected APOL2 CHO cells, WT podocytes without and with IFNγ stimulation (26 hours at 100 ng/ml), and IFNγ-stimulated APOL1 KO podocytes. To generate the cell pellets, ten 15-cm dishes of CHO cells were Lipofectamine transfected with 12 µg of APOL2 DNA per dish for 48 hours; iAPOL1-CHO stable cell lines were induced for 48 hours. Cells were detached with 5 mM EDTA in PBS, pelleted, then formalin fixed and paraffin embedded (FFPE) using standard methods. The generation of APOL1-G0 BAC transgenic C57BL/6 mice (under the human APOL1 promoter, which express APOL1 similarly to humans in liver, kidney, and lung) will be described elsewhere (see also ref.³⁶). All of the murine hybridoma supernatants and cloned rabmabs were screened at 5 µg/ml on the Thermo Fisher AutoStainer with DAKO Target Retrieval solution and ABC-HRP detection.

The optimized protocol for IHC was performed on a Thermo Fisher AutoStainer. After 20 minutes of heat-induced antigen retrieval at 99°C in DAKO Target Retrieval solution, primary anti-APOL1 rabmab 5.17D12 was incubated at 0.5 µg/ml for 60 minutes at room temperature, followed by a secondary biotinylated donkey anti-rabbit antibody (711–066–152; Jackson ImmunoResearch) at 5 µg/ml for 30 minutes, then Streptavidin-HRP for 30 minutes and a 3-minute amplification step with TSA (Perkin Elmer). The chromogen was DAB. FFPE normal human kidneys were obtained under agreement from Folio and Cureline from two women and three men, aged 64–89. Podocytes were identified morphologically as lining the glomerular basement membrane and having flat cytoplasm and a plump nucleus with prominent nucleoli. Glomerular endothelia instead follow the shape of capillaries and have long, flat nuclei and long, flat cytoplasm lining the capillary lumen.

Immunocytochemistry

First, 5-µm cryosections from OCT frozen normal human (52-year-old man) kidney were thawed at room temperature, washed three times with PBS, and then blocked in PBS with 10% normal donkey serum for 1 hour at room temperature. Samples were then coincubated with 1 µg/ml anti-APOL1 4.17A5 and 1 µg/ml rabbit anti-synaptopodin H-140 (Sc-50459; Santa Cruz Biotechnology) or 1 µg/ml rabbit anti-WT1 C-19 (Sc-192; Santa Cruz) in PBS overnight at 4°C. After washing 3× 5 minutes in PBS, samples were incubated with 1.87 µg/ml Alexa488 anti-mouse and Cy3 anti-rabbit (715–546–150 and 711–166–152, respectively; Jackson ImmunoResearch) for 1 hour at room temperature, washed 3× in PBS, and mounted in ProLong Gold with DAPI (P36931). Samples were imaged by spinning disc confocal microscopy as for IF (see above), except with the 20× (NA 0.8) and 40× (NA 1.3) objectives.

Electron Microscopy (iEM)

For iEM, 1.4×10⁶ HEK-293 cells in T25 flasks were reverse transfected with 5.4 µg of APOL1 or APOL2 cDNA with 12 µl of Fugene 6 (11814443001; Roche) for 50 hours or 65 hours, respectively. Podocytes and transiently transfected HEK-293 cells were fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (PB), pH 7.4, for 2 hours at room temperature, then overnight at 4°C, and with 2% paraformaldehyde, 0.2% glutaraldehyde in 0.1 M PB, pH 7.4, for 2 hours at room temperature, respectively. Fixations were continued in 1% PFA in 0.1 M PB at 4°C for several days. The cells were scraped, pelleted, embedded in 12% gelatin, cryoprotected with 2.3 M sucrose, mounted on aluminium pins, and frozen in liquid nitrogen. Ultrathin cryosections were cut at −120°C and stained with anti-APOL1 or anti-KDEL antibodies followed by protein-A gold particles (10 nm). Where indicated, incubation with rabbit anti-mouse IgG antibody, as a secondary antibody, was inserted between the primary antibody and Protein-A gold steps to amplify the signal. For quantifications on immunogold-labeled sections, the criteria of gold particle scoring described below were established, taking into account that the primary antibody–Protein A–10-nm gold particle complex (without rabbit-anti-mouse amplification) is approximately 25-nm long. Hence, 10-nm gold particles on a section can be located with their centers at a maximal distance of 20 nm from the epitope to which they are bound.

For the quantitation of the distribution of APOL1/2 with respect to the ER membranes of APOL1/2-expressing HEK-293 cells, random images of APOL1/2-expressing HEK-293 cells were made. The positions of at least 50 gold particles per antibody were scored as (1) outer membrane face: overlapping on the outer ER membrane surface and cytosol, or located in the cytosol at a distance of ≤15 nm between the outer ER membrane surface and the 10-nm gold particle surface; (2) inner membrane face: overlapping on the inner ER membrane surface and ER lumen, or located in the lumen at a distance of ≤15 nm between the inner ER membrane surface and the gold particle surface; or (3) lumen: within the ER lumen, with a distance of >15 nm between the 10-nm gold particle surface and the nearest inner ER membrane surface (see also Supplemental Figure 19A for an explanatory schematic). Gold particles were scored only if the ER membrane at the spot of the gold particle was transversely cut and consequently clearly visible. Two investigators independently performed the quantification procedure. Their percentage scores of gold particles on the ER were averaged and the results were plotted as a line chart for each antibody.

For the quantification of the APOL1 and KDEL gold particles in the ER lumen or associated with the outer membrane face or inner membrane face, sections of iAPOL1-G0 podocytes were labeled either with 3.6E10 anti-APOL1 or with anti-KDEL antibody, directly followed by Protein-A–conjugated 10-nm gold particles. Gold particles were scored in the categories as above, but only when located at a position in an ER cistern with transversally cut, and hence clearly visible, membranes and a local lumen width of ≥60 nm (at this width the “ER lumen,” according to the above criteria, is ≥20 nm wide [see Supplemental Figure 19A]). In total, 81 APOL1 and 100 KDEL gold particles were scored from 55 and 23 random images of APOL1- and KDEL-labeled sections, respectively.

For the quantification of the plasma membrane gold particle density, plasma membrane lengths were measured using ImageJ software on electron micrographs of randomly sampled cell profiles of iAPOL1-G0 (n=13) and KO (n=12) podocytes in 5.17D12-labeled ultrathin cryosections. Along these cell profiles, anti-APOL1 10-nm gold particles were scored as present at the plasma membrane if their centers were located within a distance of 20 nm from the plasma membrane.

For the quantitation of gold particle densities on ER, nuclear envelope, and mitochondria, areas of profiles of ER cisterna, nuclear envelopes, and mitochondria were measured by means of Fiji (ImageJ)³⁵ software, making use of a line raster overlaid on random electron micrographs of anti-APOL1–labeled sections of iAPOL1 and KO podocytes. Gold particles associated with ER, nuclear envelope, and mitochondria were counted and the gold particle density was calculated as gold particle number per µm² of ER, nuclear envelope, or mitochondrial area in each sample. In total, we analyzed 26 images of iAPOL1-G0 podocytes (corresponding to 11.8 µm² of ER/nuclear envelope and 7.8 µm² of mitochondrial area) and 20 images of KO podocytes (10.9 µm² of ER and 10.3 µm² of mitochondria).

For the precise localization of APOL1 at the mitochondrial-associated ER membrane (MAM), we quantified APOL1 immunogold particles on 17 randomly sampled MAMs in two samples of the iAPOL1-G0 podocytes labeled with the 5.17D12 antibody. MAMs were identified as segments of ER cisterna in which the membrane was located ≤30 nm from a mitochondrial outer membrane. At MAMs, 10-nm gold particles were assigned to the ER cisterna segment or to the mitochondria if they were overlapping with the lumen or limiting membranes of the respective compartment. The areas of both the ER cisterna segment of the MAM and the associated mitochondria were measured using ImageJ (Fiji) software,³⁵ and, subsequently, gold particle numbers per ER and per mitochondrial area were calculated.

Flow Cytometry

Live cells detached with 5 mM EDTA in PBS were stained on ice with 1 µg/ml 3.6D12 for 1 hour, washed, and detected with 1 µg/ml Alexa488 anti-mouse (A11029; Invitrogen) for 1 hour on ice, with three washes after each antibody. Flow cytometry was performed on a FACSCalibur as in Gupta et al.²⁸

Western Blotting

Cells were lysed in native RIPA (1% NP-40 buffer) as in Gupta et al.,²⁸ except that they were run on 10%, 12%, or 4%–20% Tris-Glycine gels with See Blue Plus 2 molecular-weight markers (LC5925; Invitrogen); or on 4%–12% Bis-Tris gels in MOPS buffer with Precision Plus protein Kaleidoscope markers (1610375; BioRad). Tris-glycine gels were transferred with an iBlot2 (P1 program) to nitrocellulose membranes and blocked in milk buffer (5% w/v nonfat milk and 0.1% Tween-20 in PBS) for 30 minutes at room temperature, then primary antibodies were incubated overnight in milk buffer at 4°C. Blots were washed 4× 15 minutes. Then, 1:3000 HRP-conjugated anti-mouse (NA931V; GE Healthcare) or 1:10,000 HRP anti-rabbit (711–036–152; Jackson) was incubated at room temperature for 1–2 hours, then washed 4× 10 minutes (all in milk buffer). Detection was with ECL (Amersham RPN2109) for 2 minutes or ECL Prime (Amersham RPN2232) for 5 minutes and blots were exposed to BioMax MR film (895 2855; Carestream). See Gupta et al.²⁸ for details of Bis-Tris gel immunolabeling. The best anti-APOL1 antibody for western blotting was a mixture of cloned rabbit 3.1C1 and 3.7D6 (at 0.05 µg/ml each), encompassing both ends of APOL1.

Podocyte Fractionation

First, 150× 15-cm dishes of 24-hour IFNγ-induced WT podocytes were harvested by trypsinization (because trypsin acts faster than EDTA); then manually homogenized in 10 ml (in 5× 2-ml batches) of IB_cells-1 buffer (225 mM mannitol, 75 mM sucrose, 30 mM Tris-Cl, 0.1 mM EGTA, pH 7.4) with 50 strokes of a Teflon Potter–Elvehjem #22 tissue grinder; sonicated (3× 10 seconds at speed 3.5 in a Branson 450 sonifier) to better separate ER, mitochondria, and MAM; and then fractionated according to the protocol of Wieckowski et al.³⁷ Briefly, the crude mitochondria from the 10,000 × g pellet were separated from the MAM by fractionation on a 30% Percoll gradient at 95,000 × g for 30 minutes, whereas the ER was collected from the mitochondrial supernatant by centrifugation at 100,000 × g. Then, 10 µg of each fraction (judged by the BCA assay, Pierce) was loaded on 4%–20% Tris-Glycine gels and probed with suitable compartment markers.

Results

Kidney Podocytes Express APOLs 1, 2, and 6

APOL1 is the only secreted member of the human APOL family, comprising APOL1–APOL6,^38–40 and is most closely related to APOL2 (53% protein identity; Supplemental Table 1A). RT-PCR confirmed that APOLs 1, 2, 3, and 6, but not APOL4 or 5, mRNAs are expressed in normal human kidneys (Figure 1A).^38–40 Dual in situ hybridization (ISH) pinpointed APOL1 mRNA to glomeruli, the latter in a subset of podocytes (Supplemental Figure 1D), as judged by partial overlap with NPHS2, but not a mesangial marker. Dual ISH of APOL1 and endothelial EGFL7 detected low APOL1 mRNA in extraglomerular endothelial cells, but not in glomerular endothelia.

Figure 1.

APOLs 1, 2, and 6 are expressed in kidney podocytes. (A) RT-PCR of five normal human kidneys with probes specific for each APOL family member. Data are normalized to RPL19 and the mean and SD of the five kidneys are plotted over the individual data points (each kidney is assigned a unique symbol and the mean of the relative expression [2^−dCT] of up to three independent RT-PCRs is plotted for each sample). See Supplemental Figure 1A for APOL probe specificities. (B) RT-PCR of immortalized differentiated (Diff, blue, grown at 38°C) or undifferentiated (Undiff, red, grown at 33°C) human podocytes (mean±SD of three separate passages, each analyzed in duplicate) using the same probes as in (A). The variability is likely attributable to our observation that APOL1 levels decline during passaging. Open triangles, untreated; closed triangles, IFNγ treated (100 ng/ml for 24 hours). APOL3 became detectable after IFNγ treatment in undifferentiated, but not differentiated, podocytes. Asterisks indicate APOLs whose expression was significantly elevated by IFNγ treatment according to the two-tailed unpaired t test (**P<0.01; ***P<0.001). (C) Representative western blot on a 4%–12% Bis-Tris gel of immortalized undifferentiated (UD) and differentiated (Diff) podocytes ± IFNγ stimulation using the Proteintech polyclonal that recognizes APOL2 as well as APOL1 (see Figure 2B). The asterisk denotes a nonspecific band not seen with other anti-APOL1 antibodies. Actin served as the loading control.

In differentiated, immortalized human podocytes,²⁷ only APOL1 and APOL2 mRNAs were detected (Figure 1B), both increasing >10-fold with IFNγ stimulation, as expected.¹⁰ APOL6, but not APOLs 3–5, also became detectable after IFNγ treatment. Similar trends were seen in undifferentiated podocytes, although APOL1 induction by IFNγ was less pronounced. By western blotting, APOL2 (37 kDa) appeared more abundant than APOL1 (41 kDa), and both increased with IFNγ (Figure 1C). Thus, specific immunolocalization of APOL1 in podocytes requires antibodies noncrossreactive with APOL2 or APOL6.

Many Anti-APOL1 Antibodies Crossreact with APOL2

Unfortunately, all three commercial rabbit antibodies used in previous reports^{8,21,22,25,26} do crossreact with APOL2 (Figure 2, A and B), the Proteintech polyclonal additionally recognizing APOL3 (and weakly APOL4, not shown) by IF. Therefore, to identify APOL1-specific antibodies, we screened 135 mouse mAbs to APOL1²⁸ by IF of APOL1- and APOL2-transfected COS cells. Forty-four of the 107 mAbs that detected APOL1 crossreacted with APOL2 (data not shown). To expand epitope coverage, we generated 35 more monoclonals in rabbits (rabmabs), 23 of which immunostained APOL1, and six also APOL2. The high monoclonal crossreactivity with APOL2 (50 of 130 IF positive, 38%) was unexpected, because the antigen has only 61% homology (Supplemental Table 1A).

Figure 2.

Several anti-APOL1 antibodies crossreact with APOL2. (A) APOL-family reactivity of commercial anti-APOL1 antibodies by IF. COS cells were transfected with untagged APOL1-G0, APOL2, APOL3a-myc-FLAG, APOL4-myc, or APOL6-myc-FLAG (see Supplemental Table 1A for cDNAs). After 40–48 hours, cells were PFA fixed, Triton X-100 permeabilized, and stained with the indicated commercial rabbit anti-APOL1 antibodies (at 1 µg/ml; Ptech, Proteintech; Epito, Epitomics) followed by Alexa488 anti-rabbit. Transfection of APOLs 3–6 was confirmed by staining for the epitope tag (with mouse anti-FLAG-M2 or mouse anti-myc 9E10 for APOL4-myc, followed by Alexa647 anti-mouse), and any positive signals colocalized with the tag as expected (Supplemental Figure 2B). All results were verified in at least one independent experiment. (B) Western blots of APOL-transfected COS cells or endogenous APOL1 in podocytes after 24-hour IFNγ stimulation with commercial anti-APOL1 antibodies corroborate the IF data. Lysates were run on 4%–12% Bis-Tris gels. UT, untransfected COS; G0, G1 (I384M, S342G), and G2 (ΔN388,Y389) are the three APOL1 variants; L2, APOL2; L3, APOL3a-myc-FLAG; L4, APOL4-myc; L6, untagged APOL6; WT, wild-type podocytes; KO, APOL1 CRISPR KO podocytes. Anti-myc labeling confirmed expression of myc-tagged APOLs 3 and 4 and actin served as the loading control. (C) As in (A), except with in-house mouse anti-APOL1 monoclonals at 1 µg/ml. Secondary antibodies were Alexa488 anti-rabbit for 5.17D12 and Alexa488 anti-mouse for the others. 5.17D12 and 3.7D6 are APOL1-specific, and although mAb 4.17A5 crossreacted weakly with APOL4, it could be considered APOL1-specific for the purposes of kidney staining due to lack of APOL4 expression in this tissue (Figure 1). By contrast, the 3.6D12 mAb crossreacts with APOLs 2, 3, and 4 by IF. (D) The same lysates as in (B) were probed with the top in-house mouse monoclonals at 2 µg/ml, or a mixture of rabbit 3.7D6 and 3.1C1 at 0.05 µg/ml (bottom blot), which covers both ends of APOL1 and is our preferred reagent for western blotting. 3.1C1 maps to the linker domain, whereas the rest map to the N-terminal pore-forming domain, except for 4.12E5, which recognizes the C-terminal APOL1-G2 epitope (it barely recognizes APOL1-G2 by western blotting and does not crossreact with APOL2–6 [confirmed by IF; Supplemental Figure 2B]). The results corroborate the IF crossreactivities, except for APOL4 crossreactivity, which appears to be conformationally sensitive. The 3.6D12 antibody continues to recognize APOL2 in the APOL1 KO podocytes (generated in Gupta et al.²⁸). Note that the 4.17A5 western is shown at a longer exposure than the others.

The most sensitive APOL1-specific antibodies for IF were rabmab 5.17D12 and mAbs 4.17A5 and 3.7D6 (Figure 2C). mAb 3.6D12 was strongest, but crossreacted with APOLs 2, 3, and 4 (Figure 2C, Supplemental Table 2). Western blotting confirmed these APOL specificities, their sensitivities for detecting endogenous podocyte APOL1 (WT), and recognition of APOL1-G1 and -G2 (Figure 2D). The 41-kDa podocyte band was absent in APOL1 KO podocytes, confirming antibody specificities.

APOL1-Specific Antibodies Reveal APOL1 Is in Podocytes, Mainly Restricted to the ER

Of 170 antibodies screened for APOL1 specificity by immunohistochemistry, 5.17D12 was the best (Supplemental Figure 3). In human and APOL1 transgenic mouse kidneys, 5.17D12 stained podocytes and the luminal surfaces of glomerular and extraglomerular endothelial cells, but there was no clear staining of proximal tubules (Figure 3A, Supplemental Figure 3B), unlike with the commercial polyclonals (Supplemental Figure 3C).^22,25 APOL1 was robustly expressed in the same cells as synaptopodin and WT1 by immunocytochemistry (Figure 3B), confirming its presence in podocytes. Liver hepatocytes also expressed APOL1 (Supplemental Figure 4), but more weakly than glomeruli; their similar western blot signals²⁵ presumably reflect the far greater percentage of hepatocytes in liver than glomeruli in kidney (<2%).⁴¹

Figure 3.

Endogenous APOL1 is found in normal kidney podocytes and endothelial cells by immunohistochemistry. (A) Left panel, immunohistochemistry of FFPE normal human kidney stained with 0.5 µg/ml APOL1-specific rabmab 5.17D12 at 5× (200-µm scale bar), showing two positive (brown-stained) glomeruli (red arrows), multiple positive extraglomerular endothelia (black arrows), and numerous APOL1-negative proximal tubules (PT). Center, ×20 magnification (scale bar, 50 µm) of the APOL1-stained upper glomerulus from the left image. Red arrows point to APOL1 signal in podocytes; yellow and black arrows highlight APOL1 on the luminal surface of glomerular endothelial cells and extraglomerular endothelial cells (capillaries), respectively. The signal is primarily membrane and cytoplasmic staining. The different cell types were identified morphologically by a kidney pathologist (see Methods). This staining pattern was replicated by two other top anti-APOL1 antibodies to different epitopes and is representative of at least five kidneys examined (data not shown). Unlike a previous report,²² the staining was not predominantly in proximal tubules. G, glomerulus; PT, proximal tubule. Right, rabbit IgG isotype control on an adjacent section of the same glomerulus, also at ×20, showing no signal. (B) Dual IF confirms that APOL1 is strongly expressed in human kidney podocytes. A frozen normal human kidney was costained with 1 µg/ml murine 4.17A5 anti-APOL1 (green, left) and rabbit anti-podocyte markers (center, red): synaptopodin (SYNPO, at ×20 [top] and ×40 [middle row]) or nuclear WT1 (lower row, at ×40), all overlaid with DAPI (blue). The merged images on the right confirm that most APOL1 is found in synaptopodin- and WT1-positive podocytes and is predominantly intracellular (Figure 4 suggests that it is likely ER). Scale bars in all panels, 50 µm.

Subcellularly, APOL1-G0, -G1, and -G2 expressed transiently in COS cells appeared almost exclusively in the ER, including the nuclear envelope, with all 130 IF-positive antibodies, colocalizing with the ER-specific marker calnexin, but not other compartments (Supplemental Figures 5–8 and data not shown). Interestingly, APOL2 also colocalized with calnexin, despite lacking a signal sequence (Supplemental Figure 5). All three APOL1 variants likewise localized to the ER and not mitochondria in stably transfected doxycycline-inducible APOL1 (“iAPOL1”) podocytes (Supplemental Figure 9A), which was confirmed by iEM (Figure 4, A and B, Supplemental Figure 10). Although APOL1 occasionally appeared to partially overlap with a few mitochondria by IF in the perinuclear region, this is likely a resolution issue because 5%–20% of mitochondria are in close proximity to the ER,⁴² and triple labeling of APOL1, ER, and mitochondria revealed that APOL1 was never seen in mitochondria without concomitant ER overlap (Supplemental Figure 9B). More convincingly, mitochondria were consistently APOL1-negative by quantitative iEM, even when abutting positive ER cisternae (Supplemental Figure 10). Subcellular fractionation also corroborated that endogenous podocyte APOL1 was enriched in the ER and MAMs,⁴³ but absent from pure mitochondria (Supplemental Figure 11A). Quantitative APOL1 iEM of the MAM confirmed that APOL1 is within the ER cisternae (Supplemental Figure 11, B–G), but not at all in the mitochondria. iEM could also detect some APOL1 in the Golgi that was not readily discernible by IF (Figure 4C), but not in any other intracellular organelles, indicating that APOL1 is secretory.

Figure 4.

Endogenous and stably transfected APOL1 localizes to the ER of cultured podocytes. (A–C) iEM of iAPOL1-G0 podocytes. APOL1 KO podocytes re-expressing iAPOL1-G0 (after 5 ng/ml doxycycline induction for 48 hours) were 4% PFA fixed and immunolabeled with 25 µg/ml rabmab 5.17D12, followed by protein-A gold. Staining of APOL1 in the ER (A and B) and Golgi (C) is shown. All scale bars, 200 nm. *, ER lumen; M, mitochondrion; L, lysosome; E, early endosome; G, Golgi. APOL1 is seen in the ER and, to a lesser extent, the Golgi ([C], note that the image shown has above-average gold particles to illustrate that APOL1 is seen in this compartment), but not in mitochondria or the endolysosomal system. Examples of APOL1 staining in the more perinuclear region of the ER and its absence from the mitochondria and MAM are shown in Supplemental Figure 10, along with uninduced controls showing that the antibody staining is APOL1-specific. (D) Endogenous APOL1 in wild-type differentiated podocytes (upper row) is shown in PFA-fixed, saponin-permeabilized cells costained with 2 µg/ml APOL1 5.17D12 (green) and mouse anti-calnexin (red); merge with nuclear DAPI is on the right. IFNγ-treated (24 hours) WT podocytes (middle) show stronger ER labeling of APOL1, whereas APOL1 KO (lower) podocytes lacked this. Note that nuclear speckles (overlapping DAPI) are evident in many cells, which are clearly nonspecific (not APOL1) because they are found in KO as well as WT cells. Scale bar, 40 µm. APOL1 was similarly ER-localized with eight other antibodies in both differentiated and undifferentiated podocytes (see Supplemental Figure 12 for 4.17A5, and other data not shown).

Importantly, ER-localized APOL1 was not merely an overexpression artifact, because endogenous APOL1 was similarly ER-localized in both differentiated and undifferentiated podocytes (Figure 4D, Supplemental Figure 12) and JHH-1 (liver) cells (Supplemental Figure 13). Endogenous podocyte APOL1 increased after upregulation by IFNγ and was absent from APOL1 KO podocytes and from other compartments, including Golgi, endosomes, mitochondria, and lipid droplets (Supplemental Figures 14,15 and 8B), and showed only minor overlap with MAM markers by IF (Supplemental Figure 16). Notably, APOL2-crossreactive antibodies stain the ER of APOL1 KO podocytes almost as strongly as WT, indicating that most of the WT signal actually represents APOL2 (Supplemental Figure 17). In summary, endogenously, transiently, and stably expressed APOL1 appears ER-restricted in all cell types examined.

APOL1, but Not APOL2, Is on the Extracellular Face of the Plasma Membrane

Despite not detecting APOL1 at the plasma membrane by IF in any of these (permeabilized) cell lines, most antibodies recognized surface APOL1 by flow cytometry of unpermeabilized WT (but not APOL1-KO) podocytes and stable cell lines.²⁸ APOL2-crossreactive 3.6D12 likewise did not shift on APOL1 KO podocytes (which retain APOL2), indicating that APOL2 is not found on the cell surface (Figure 5A). Surface APOL1 is only a small proportion of the total, because in permeabilized cells the far-stronger ER signal obscures the surface signal. Staining of unpermeabilized podocytes showed that surface APOL1 was punctate, with fewer, less-bright puncta than in higher-expressing JHH-1 cells and iAPOL1-G0, -G1, and -G2 podocytes (Figure 5B, Supplemental Figure 18). In line with this, quantitative iEM of APOL1 on the plasma membrane of iAPOL1-G0 podocytes showed thinly dispersed signal, albeit specific, because it was 14-fold higher than the background on KO cells (Figure 5C and D).

Figure 5.

A small proportion of endogenous podocyte APOL1, but not APOL2, is at the plasma membrane. (A) Live WT (left) or APOL1 KO (right) podocytes were incubated on ice for 1 hour (without fixation or permeabilization) with 2.5 µg/ml 3.6D12 (an APOL2-crossreactive mAb). Gray, Alexa488 anti-mouse secondary antibody alone; blue, untreated; red, 24-hour IFNγ-treated podocytes. There is a shift on WT, but not KO cells (which express APOL2 but not APOL1). The lack of signal on APOL1 KO cells indicates that APOL2 is not on the cell surface, despite being more abundant than APOL1 (Figure 1C, Supplemental Figure 17A). Thus, the shift seen with 3.6D12 in WT cells must represent APOL1, and increases with IFNγ, as expected. Evidence that 3.6D12 has the ability to recognize unfixed APOL2 on cells is shown in Supplemental Figure 21A. (B) Live IFNγ-treated WT podocytes (upper), APOL1 KO podocytes (middle), and liver JHH-1 cells (lower) were incubated with 5 µg/ml APOL1-specific 3.7D6 (left) or APOL2-crossreactive 3.6D12 (right) on ice for 1 hour, then washed, PFA fixed, and detected (without permeabilization so as to avoid seeing the abundant intracellular ER signal) with Alexa488 anti-mouse (green; overlaid with nuclear DAPI in blue). Punctate signal is seen on the cell surface of WT podocytes and JHH-1 (but not APOL1 KO podocytes) with both APOL1-specific and APOL2-crossreactive antibodies, indicating that only APOL1 (not APOL2) is on the cell surface. Scale bar, 40 µm. Similar results on podocytes were seen with ≥8 other antibodies to different epitopes (data not shown). (C) iEM of APOL1-G0 podocytes, using the APOL1-specific antibodies 5.17D12 and 3.6E10, showing the dispersed distribution of APOL1 immunogold particles (arrows) at the cell surface. Note that the gold density on the plasma membranes visible in these images is 2–5 times above average in order to illustrate more than one particle per image. *, ER lumen; G, Golgi; N, nucleus. Scale bars (top to bottom), 100, 200, 200, and 500 nm. (D) Quantification of 5.17D12 anti-APOL1 immunogold particle density (protein A gold [PAG] particle number/µm membrane length) at the plasma membrane of iAPOL1-G0 podocytes (n=13 cells, totaling 945.8 µm plasma membrane analyzed, black triangles) is substantially (14.4-fold) above the background level in APOL1 KO cells (n=12 cells, totaling 1148.7 µm analyzed, open triangles). Means and SDs are overlaid on the results for the individual cells. Examples of negative-control KO podocyte staining are shown in Supplemental Figure 10, E and F.

APOL1 and APOL2 Localize to Opposite Faces of the ER Membrane

Because only APOL1 was found on the cell surface, but both APOL1 and APOL2 appeared ER-localized, we examined whether APOL2 localizes to the outer (cytoplasmic) face of the ER by permeabilizing APOL2-transfected COS cells with digitonin, which selectively permeabilizes the plasma membrane, leaving the ER membrane intact³⁴ (Figure 6A). All 44 APOL2-crossreactive mAbs exhibited a similar ER staining pattern with digitonin as with saponin or Triton X-100 (which permeabilize all cell membranes), indicating that APOL2 is indeed at least partially on the cytoplasmic face of the ER (Figure 6B). Fusing a signal sequence onto APOL2 completely abolished the reticular ER signal with digitonin; ssAPOL2 instead appeared in the nuclear membrane, which is contiguous with the ER lumen. This suggests that APOL2 is normally all on one face of the ER, rather than of mixed topology on both faces.

Figure 6.

APOL1 is associated with the inner face of the ER membrane. (A) Schematic of PFA-fixed cells permeabilized with 0.4% saponin for 20 minutes (or 0.1% Triton X-100 for 4 minutes; upper), with cell-surface and ER membranes permeabilized (dashed lines) versus cell-surface–specific permeabilization by 0.0025% digitonin for 4 minutes (ER membrane is intact, solid line; lower). Note that PFA fixation permeabilizes the nuclear envelope of some cells (Nuc, gray) irrespective of any detergent. (B) COS cells transiently transfected with APOL2, or APOL2 with the signal sequence of APOL1 at its N terminus (ssAPOL2), or doxycycline-induced (5 µg/ml for 20 hours) iAPOL1-CHO stables²⁸ were PFA fixed then permeabilized with saponin (top) or digitonin (bottom). The APOL2 proteins in COS cells are shown stained with 3.6D12 (a mAb to the N-terminal pore-forming domain [PFD]), and similar results were obtained with the other 43 APOL2-crossreactors (all to the PFD, data not shown). Anti-calnexin luminal domain staining for the APOL2-COS (monkey) cells is shown to the right of APOL2 as a control for permeabilization conditions (this antibody does not crossreact with CHO [hamster] calnexin). APOL1, in iAPOL1-CHO cells, is shown stained with APOL1-specific PFD mAb 3.7D6, membrane-addressing domain (MAD) mAb 3.3A8, or SRA-interacting domain (SRA-ID) mAb 3.1C1, each detected with Alexa488 anti-mouse (see Gupta et al.²⁸ for epitope mapping), and similar results were obtained with all of the other APOL1-specific antibodies to all three domains (data not shown). The reticular ER pattern persists in digitonin (plasma membrane only)-permeabilized cells for APOL2, but not APOL1 or ssAPOL2, indicating that APOL2 is on the cytoplasmic face of the ER, whereas APOL1 and ssAPOL2 are inside, and thus not on both faces at once. The nuclear membrane signal for APOL1 with digitonin is also seen without detergent in some cells because PFA semipermeabilizes the nuclear membrane (data not shown). With digitonin, some of the more sensitive antibodies, including the PFD (3.7D6) and SRA-ID (3.1C1) antibodies shown here (but not MAD antibody 3.3A8), additionally detected APOL1 on the plasma membrane (arrowheads), which is topologically equivalent to the ER lumen. (C) Immunoelectron micrographs of iAPOL1-G0 podocytes (induced for 24 hours with 5 ng/ml doxycycline) immunolabeled with 3.6E10 and amplified with rabbit anti-mouse secondary antibody for APOL1 (left) or the soluble luminal marker anti-KDEL (right), directly detected with 10-nm protein A–gold, both showing representative gold distributions in the ER cisternae. *, ER lumen; M, mitochondrion. Scale bars, 200 nm. APOL1 appears more closely associated with the inner ER membrane face than does the luminal KDEL. (D) Quantitation confirms that APOL1 is more associated with the inner face of the ER membrane than KDEL in iAPOL1-G0 podocytes. Quantitation of ER-associated gold particles was done on 55 images of APOL1 immunogold-labeled sections (81 gold particles) and 23 images (100 gold particles) from KDEL-labeled sections. Both labelings were performed without secondary antibody amplification in order to minimize the size of the antibody–Protein A–gold complexes and to render the quantitation more accurate. APOL1 is mainly associated with the inner ER membrane face (IMF), with a small amount attributed to the outer (cytosolic) face (OMF) of the ER membrane or central lumen by virtue of the approximately 25-nm size of the immunogold complex relative to the 7-nm ER membrane thickness (see Supplemental Figure 19A for details). By contrast, the KDEL reference marker for a luminal ER protein appeared equally distributed over the central ER lumen and the 20-nm luminal zone near the surface of the inner ER face, and, as expected, was minimally detected in the 20-nm zone of cytosol flanking the outer ER membrane surface. The few gold particles apparently lying in the cytosol away from any membrane were considered background and not included in the quantitation.

All 80 APOL1-specific antibodies gave nuclear membrane (plus plasma membrane if high affinity) instead of reticular ER signal with digitonin permeabilization (Figure 6B and data not shown), indicating that APOL1 is inaccessible inside the ER lumen. Thus, all three domains of APOL1 are luminal, with not one of our epitopes projecting into the cytoplasm. Quantitation of the iEM additionally revealed that APOL1 was associated with the inner ER membrane face of iAPOL1 podocytes, not soluble in the ER lumen like the luminal ER marker KDEL (Figure 6, C and D). APOL1 was similarly associated with the inner ER membrane face of transfected HEK-293 cells (Supplemental Figure 19), whereas APOL2 appeared more distributed toward the outer ER leaflet, supporting the differential topologies observed with digitonin.

Significantly, endogenous APOL1 and APOL2 were also on the luminal and cytoplasmic faces of the ER, respectively, in digitonin-permeabilized podocytes (Supplemental Figure 17). This suggests that any alternatively spliced cytoplasmic APOL1 isoforms^10,44 are either luminal or below the detection limit (see below). Thus, APOL1 and APOL2 are on the inner and outer faces of the ER membrane, respectively.

APOL1 Isoforms Have Different ER Topologies

It has been proposed that APOL1 risk variants bind less well than G0 to the vesicle-SNARE VAMP8, resulting in inhibition of endosomal trafficking as a putative mechanism of risk variant pathology.²² However, vesicle-SNAREs are all cytoplasmically oriented, whereas we clearly showed above that APOL1 is not. A caveat is that we likely detected the major reference isoform (vA or v1), whereas IFNγ stimulation of podocytes renders three minor splice variants of APOL1 (isoforms vB1, vB3, and vC)^10,44 detectable (Supplemental Figure 20, Supplemental Table 1B). These differ in their N termini so may not all be secreted, and thus their topology needed evaluating. For clarity, we call the splice forms “isoforms” and the G1/G2 risk alleles “variants.”

It has been shown that all four isoforms are secreted into the media upon transient overexpression in HEK-293 cells.⁴⁴ However, as the authors suggested,⁴⁴ this may be an overexpression artifact, because we obtained the same result in transiently transfected COS cells, but not in stable cell lines. In iAPOL1 podocytes individually expressing each of the APOL1 isoforms (all of which localized to the ER; Figure 7A), only the vA and vB1 isoforms that appeared luminal (with the digitonin technique) gave a large shift at the cell surface by FACS of live cells when induced to comparable levels, i.e., were transported along the secretory pathway (Figure 7, B–D). By contrast, in COS transients, vA with its signal sequence deleted (Δss) appeared on the cell surface and in the media equally well as normal vA, despite having the expected opposite ER topology by digitonin IF (Supplemental Figures 21 and 22). Thus, APOLs “secreted” into the media probably at least partially originate from overexpression-induced cell rupture or other artifactual (nonclassic) release when transiently overexpressed, as previously suggested.⁴⁴

Figure 7.

APOL1 isoforms localize to opposite sides of the ER membrane in podocyte stables. (A) Stable pools of APOL isoforms were PFA fixed and fully permeabilized with Triton X-100 to reveal total APOL distribution, and APOL1/2 was costained with rabbit anti-calnexin cytoplasmic tail (intracellular domain, Cnx ICD), followed by Alexa488 anti-IgG2a and Dy649 anti-rabbit. All of the APOL1 isoforms and APOL2 appear ER-associated and colocalize well with calnexin under these conditions. (B) As in (A), except with digitonin permeabilization to reveal only cytoplasmic APOLs, and mouse anti-calnexin extracellular (luminal/ECD) domain and isotype-specific secondaries (Alexa488 anti-IgG2a for APOL; Alexa647 anti-IgG1 for calnexin). APOL1.vA, vB1, and a few cells expressing vC exhibit luminal staining (noncytoplasmic, arrow; arrowhead indicates nuclear envelope [luminal] signal in the same cell as reticular [cytoplasmic] signal), whereas vB3, APOL2, and the majority of vC are cytoplasmically oriented because they retain a reticular ER staining pattern with digitonin. Note that the calnexin ECD antibody gives similar nuclear membrane staining to the luminal APOL1 isoforms, validating our digitonin method. (C) Flow cytometry of APOL1-G0 isoforms and APOL2 podocyte stable pools (gated on live [PI-negative] cells i.e., unfixed, unpermeabilized) stained with 1 µg/ml 3.6D12 and Alexa488 anti-mouse after 22 hours of induction at 10 ng/ml doxycycline (except UI, uninduced vA control). APOL1.vA and vB1 give a large FACS shift, indicating that they are secretory, vC gives a smaller shift (partially secretory), and vB3 and APOL2 (L2) give almost no shift (i.e., nonsecretory), in accordance with the topologies identified by digitonin IF in (B). The y axis is % maximum and the x axis is Alexa488 fluorescence intensity; gray lines are secondary antibody alone and black lines are 3.6D12 histograms. Note that even the secretory isoforms (vA and vB1) were not actually secreted into the media due to being anchored to the cell surface, presumably via their predicted transmembrane domains⁵³ (data not shown). (D) The different APOL1 isoforms are expressed at comparable levels in stable podocytes. A portion of the cells used in (C) were western blotted on a 10% gel with 0.05 µg/ml 3.1C7/3.7D6 (validated in Figure 2D) and calnexin as a loading control. The similar total expression of the different isoforms implies that the greater FACS shifts with vA and vB1 (C) are due to greater secretory transport to the cell surface rather than higher expression. Furthermore, the similar size of APOL1.vA and vB1 suggests cleavage at the same VRA/EE site and thus that the 43-aa signal sequence of vB1 is functional, in contradiction to the signal sequence program predictions (Supplemental Table 1B). APOL1.vB3 is larger than vA, consistent with lack of signal sequence cleavage, which was confirmed by retention of an N-terminal tag (Supplemental Figure 22C), but also has a smaller, potentially cytoplasmically cleaved band. By contrast, vC is smaller than vA, despite predictions that it should be 1-kDa larger because it is mostly nonluminal (Supplemental Table 1B), and thus it may also be cytoplasmically clipped; indeed at high expression levels it appears as a doublet (Supplemental Figure 22B).

The ER-luminal localization of stable APOL1.vB1 in podocytes agrees with its identical size to vA by western blot (Figure 7D), which implies that their different signal sequences are cleaved at the same site. Conversely, vB3 runs larger, consistent with no signal cleavage (Supplemental Figure 22C, Supplemental Table 1B) and its cytoplasmic location; the lower band may be a cytoplasmic degradation product similar to vC. Intriguingly, APOL1.vC is predominantly cytoplasmic by IF, but also shows a little luminal/cell surface signal, sometimes within the same cell (arrowhead in Figure 7B).

WT podocytes were western blotted to determine which endogenous APOL1 isoforms are detectable. The major band is the size of vA, as expected, but longer exposures revealed a minor upper band uniquely after IFNγ treatment, which could only represent vB3 if it were a splice isoform, and is indeed similar in size (Figure 8). RT-PCR confirmed previous reports^10,44 that the two cytoplasmic isoforms (vB3 and vC) are 30× less abundant than the two luminal ones (vA and vB1; Supplemental Figure 20C); hence, it is unsurprising that no cytoplasmic endogenous APOL1 could be detected by IF. In summary, APOL1.vA predominates endogenously and is inside the ER, and any minor cytoplasmic variants (vB3 and vC) localize to the outer face of the ER rather than endosomes (Figure 7B, Supplemental Figure 23).

Discussion

We characterized multiple APOL1-specific antibodies to investigate the localization of APOL1. At the tissue level, in agreement with previous reports,^{8,22,25,26,45} we found APOL1 mRNA and protein in liver hepatocytes and podocytes and extraglomerular endothelia of kidneys. By contrast, we found no APOL1 protein or mRNA in proximal tubules, suggesting that proximal tubule staining with the commercial polyclonals^22,26 may be APOL2 crossreactivity or simply nonspecific, underscoring the need for specific antibodies. APOL1 protein was strongly expressed on extraglomerular endothelial cells lining capillaries and arterioles, as well as glomerular endothelia, by IHC, in agreement with earlier findings.^22,26 However, the mRNA signal was weak in extraglomerular endothelia and undetectable in glomerular endothelia, in contrast to Ma et al.²⁵ Our shorter probe (or RNAscope method, the first attempted for APOL1) is probably less sensitive at the expense of its specificity for APOL1. If not cell-autonomous, APOL1 protein could be bound to, or endocytosed into, glomerular endothelia from the blood.

Subcellularly, the major (reference) isoform, APOL1.vA, was consistently (endogenously, transiently, and stably expressed) found within the secretory, but not endocytic, pathway. It was mostly in the ER, with a small proportion (≤18%)²⁸ transported via the Golgi to the cell surface (Figure 9). This is not unexpected, because it possesses a definitive signal sequence^39,44 and is secreted from hepatocytes to associate with circulating HDL particles.²⁰ In immortalized podocytes, however, secretory APOL1 remains attached to the cell surface, perhaps due to lack of HDL particle production machinery.⁴⁶ Surface APOL1 appeared punctate all over the cell and was not restricted to focal adhesions, unlike exogenously added APOL1,¹⁴ suggesting that it is not complexed with integrins, at least in the absence of integrin stimulation. Plasma membrane APOL1 is thus not merely an overexpression artifact,¹¹ and our findings support evidence that it may act as a cation channel at the cell surface.^11,15

Figure 9.

Summary of APOL1 isoform and APOL2 subcellular locations. Diagram of a cell depicting localization and relative abundance of APOL1 isoforms (although note that APOL1 and APOL2 are grossly under-represented compared with the minor variants). APOL1.vA (maroon), vB1 (pink), and a little of vC (tan) are within the ER lumen (associated with the inner membrane) and a small proportion of each is transported to the plasma membrane via the Golgi (not shown), i.e., the classic secretory pathway. By contrast, APOL1.vB3 (gray-brown), APOL2 (blue), and the majority of APOL1.vC (tan) are on the cytoplasmic face of the ER and not on the cell surface. This sketch was created with BioRender.

There was no overlap by either IF or iEM of endogenous podocyte or hepatocyte APOL1 or transfected APOL1.vA with mitochondria, lipid droplets, or endosomes, organelles reported by others to harbor APOL1.^8,13,22,23 Although we cannot exclude the possibility that ER APOL1 “drowns out” the signal in other compartments, we suspect that the far more abundant APOL2 and/or suboptimal staining procedures, epitope tags, transfection artifacts, or different cell types may have contributed to previously published results. At least some of the APOL1 within the ER cisternae is adjacent to (although not specifically enriched in) the MAM, a subdomain of the ER associated with lipid synthesis and ER stress signaling,⁴³ which is interesting because APOL1 has been reported to affect lipid metabolism,⁴⁷ reverse cholesterol efflux,¹⁶ and ER stress.^21,28

Because acidic pH is required for membrane insertion of APOL1,^49–51 we expected APOL1 to be soluble in the neutral⁵² ER lumen. However, quantitative iEM revealed that APOL1.vA is on the inner ER leaflet, although some was released into the cytosolic fraction by sonication, suggesting that it might only be peripherally associated with phosphatidylinositols.^19,23 We hypothesize that full membrane integration⁵³ may occur in the Golgi at pH ≤6 en route to the cell surface.⁵²

That GFP-APOL1 (without a signal sequence [Δss]) previously localized to the ER was interpreted as signal-sequence–independent ER targeting,¹² but we demonstrate that APOL1 Δss (like isoform vB3 and APOL2) is topologically distinct, associating with the outer, not inner, face of the ER membrane. Conversely, APOL2 becomes secretory when given a signal sequence; thus, APOL1.vA and APOL2 appear to normally bind opposite leaflets of the ER. APOL1.vC was the only isoform detected on both ER faces, presumably because loss of exon 4 does not fully disrupt its signal sequence. The localization of APOL1.vA to the inner ER leaflet agrees with its apparent absence from mitochondria and lipid droplets, because mitochondrial import sequences are recognized in the cytoplasm⁵⁴ and lipid droplets bud from the outer leaflet of the ER.⁵⁵ Thus, we propose that the recent reports of APOL1 import into mitochondria⁵⁶ and onto lipid droplets²³ may be transfection-related artifacts.

The absence of detectable endogenous cytoplasmic APOL1 in podocytes could be because these podocytes²⁷ lose APOL1 in culture; cytoplasmic isoforms vB3 and vC could conceivably be expressed at meaningful levels in vivo, particularly if upregulated by inflammatory cytokines.¹⁰ However, stably overexpressed vB3 and vC localize to the outer face of the ER (perhaps binding the abundant phosphatidylinositols in that leaflet),⁵⁷ rather than VAMP8-positive early endosomes or secretory granules.^58–60 Because APOL1-mediated cytotoxicity is reportedly restricted to the secretory (exon 4 positive) isoforms (vA and vB1),⁴⁴ in which the SRA-ID crosses neither the ER nor the plasma membrane²⁸ into the cytoplasm, we argue that stable VAMP8 binding^8,19,22 is unlikely a direct mechanism of APOL1 variant–mediated cytotoxicity.

Our localization data rather favor either the ER stress^21,48 or cell surface cation channel models^11,15,51 for APOL1 variant–mediated cytotoxicity. There was no differential localization between variants, in agreement with other studies,^12,13,22 so the function or magnitude of variant effects presumably differs. Determining where APOL1 resides within podocytes in vivo will be informative; for example, whether it colocalizes with basement membrane integrins.¹⁴ Studying the subcellular localization of APOL1 in primary podocytes from patients with APOL1 nephropathy that have sustained a “second hit,” and establishing whether APOL1 protein is upregulated or changes localization in APOL1 risk variant kidneys with and without disease, will also be important.

In summary, we have generated the first APOL1-specific antibodies sensitive enough to detect endogenous podocyte APOL1. The successfully cloned antibodies will be useful tools for future investigations of APOL1 localization and identification of interaction partners without interference from APOL2 binding that has plagued previous studies.

Disclosures

All current and former Genentech employees were compensated financially by Genentech and own Roche stock. S. Scales, A. Peterson, and N. Gupta have a patent “Apolipoprotein L1-specific antibodies and methods of use” pending.

Funding

A. De Mazière, G. Posthuma, and J. Klumperman report grants from Genentech (a member of the Roche group), during the conduct of the study.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2019080829/-/DCSupplemental.

Supplemental Table 1. (A) APOL family cDNAs and homologies. (B) APOL1 isoform cDNAs, molecular weights, and signal sequence predictions.

Supplemental Table 2. Anti-APOL1 antibody reactivity summary.

Supplemental Figure 1. APOL specificities of qRT-PCR and in situ hybridization probes.

Supplemental Figure 2. APOL specificities and colocalization with epitope tags of the anti-APOL1 antibodies.

Supplemental Figure 3. Validation of rabmab 5.17D12 for IHC staining and comparison with commercial antibodies.

Supplemental Figure 4. Expression of APOL1 in liver.

Supplemental Figure 5. Transiently transfected APOL1-G0, -G1, and -G2 and APOL2 are all associated with the ER of COS7 cells.

Supplemental Figure 6. Transiently transfected APOL1 is not appreciably detected in other compartments than the ER in COS7 cells by IF.

Supplemental Figure 7. Validation of the anti–perilipin-2 antibodies.

Supplemental Figure 8. APOL1 is not seen on lipid droplets in oleic-acid–treated cells.

Supplemental Figure 9. Stably transfected APOL1-G0, -G1, and -G2 and APOL2 are all associated with the ER in podocytes.

Supplemental Figure 10. Further examples of APOL1 localization to ER but not mitochondria in podocytes by iEM.

Supplemental Figure 11. Confirmation by fractionation and iEM that APOL1 is in the ER of the MAM and not in mitochondria.

Supplemental Figure 12. Endogenous APOL1 is also ER-localized in undifferentiated podocytes.

Supplemental Figure 13. Endogenous liver cell APOL1 is not detected within compartments other than the ER by IF.

Supplemental Figure 14. Endogenous podocyte APOL1 is not detected within compartments other than the ER by IF.

Supplemental Figure 15. Triple IF labeling confirms that endogenous APOL1 in podocytes is not specifically targeted to mitochondria.

Supplemental Figure 16. Endogenous podocyte APOL1 is adjacent to the MAM.

Supplemental Figure 17. Endogenous APOL1 and APOL2 are on opposite sides of the membrane in podocytes, with APOL2 comprising the majority of the signal.

Supplemental Figure 18. APOL1 reaches the cell surface in iAPOL1 podocytes and increases with expression level.

Supplemental Figure 19. Quantitation of inner versus outer ER membrane face localization of APOL1 and APOL2 by iEM.

Supplemental Figure 20. Quantitation of relative abundance of APOL1 isoforms in differentiated WT podocytes by RT-PCR.

Supplemental Figure 21. All APOL1 isoforms and APOL2 appear to reach the cell surface when transiently transfected in COS cells, despite differential ER topology.

Supplemental Figure 22. Transient transfection artifact in COS cells: APOL2 and signal sequence-free APOL1 isoforms appear in the media despite not being luminal.

Supplemental Figure 23. APOL1 is not associated with endosomes of digitonin-permeabilized podocytes.