PNMA2 forms immunogenic non-enveloped virus-like capsids associated with paraneoplastic neurological syndrome

Abstract

The paraneoplastic Ma antigen (PNMA) proteins are associated with cancer-induced paraneoplastic syndromes that present with an autoimmune response and neurological symptoms. Why PNMA proteins are associated with this severe autoimmune disease is unclear. PNMA genes are predominantly expressed in the central nervous system and are ectopically expressed in some tumors. We show that PNMA2, which has been co-opted from a Ty3 retrotransposon, encodes a protein that is released from cells as non-enveloped virus-like capsids. Recombinant PNMA2 capsids injected into mice induce autoantibodies that preferentially bind external "spike" PNMA2 capsid epitopes, whereas a capsid-assembly-defective PNMA2 protein is not immunogenic. PNMA2 autoantibodies in cerebrospinal fluid of patients with anti-Ma2 paraneoplastic disease show similar preferential binding to spike capsid epitopes. PNMA2 capsid-injected mice develop learning and memory deficits. These observations suggest that PNMA2 capsids act as an extracellular antigen, capable of generating an autoimmune response that results in neurological deficits.

Keywords: Gag; PNMA; autoimmune disorder; capsid; intercellular signaling; paraneoplastic disorder; retrotransposon; retrovirus; virus.

Figure 1.. Ty3/mdg4 retrotransposon-derived PNMA2 is expressed in mammalian neurons.

A. Evolution of PNMA2. Using data derived from the Vertebrate Multiz Alignment and Conservation track at the UCSC Genome Browser (hg38), we show the phylogenetic relationship (left) and sequence conservation (right) for a subset of vertebrate genomes across the mRNA sequence and shared promoter region of the PNMA2 (red) and DPYSL2 (green) genes. We find three transposable elements annotated by RepeatMasker within the PNMA2 mRNA: a Mam-Gyp-int element (purple box) that gave rise to the Gag-like coding sequence, and two Alu elements (grey boxes) embedded within the 3’ UTR. The arrows depict the predicted transcription start sites for PNMA2 and DPYSL2. The conservation track shows that PNMA2 mRNA sequence is deeply conserved across placental mammals (except for the two Alu elements which are primate-specific insertions), but not other vertebrates, while the DPYSL2 and its promoter are conserved more deeply across vertebrate evolution. B. Violin plots visualizing the density and distribution of single-cell expression of PNMA2 mRNA in the mouse cortex (left panel), mouse hippocampus (middle panel), and human cortex (right panel). The solid black dots in each plot represent each lineage’s median value of the PNMA2 expression. C. Representative image of single molecule fluorescent in situ hybridization (RNAscope) of PNMA2 mRNA in the hippocampus of a 2-month-old mouse. Scale bar: 400μm. See also Figure S1.

Figure 2.. PNMA2 proteins self-assemble into T=1 icosahedral virus-like capsids.

A. Representative negative-stained EM images of purified recombinant mouse PNMA2 (mPNMA2) capsids and human PNMA2 (hPNMA2) capsids. Scale bar: 50nm. B. cryoEM resolved surface representation of a mPNMA2 capsid, viewed down the two-fold axis. The spike densities are not resolved. C. Atomic model of the T=1 mPNMA2 capsid. CA_NTD is depicted in cyan and the CA_CTD in purple. D. External view of the isolated five-fold capsomere. E. One of sixty mPNMA2 monomers required to form the T=1 capsids. F. The two-fold CA_CTD interface axis is formed between helices 5 and 7 through a hydrophobic interaction between V252 and L255, flanked by charged and neutral residues. We detect a potential salt bridge between R247 and Q292. G. Helix 8 constitutes the threefold CA_CTD interfaces, where E314 and C310 form the interface between the individual CA molecules. H. The five-fold axis is constituted by residues 160–164 from each of the five CA molecules which form a short asymmetrical five-stranded beta-barrel, with one Y162 occupying the center of the capsomer. I. The CA_NTD:CA_NTD interfaces are formed between helices 1–3. Residues in helix 1, E186 and E194, may form salt bridges with R204 and H219 of helices 2 and 3, respectively. J. The CA_NTD:CA_CTD interface forms by docking of CA_CTD helix 6 residues L270 and L325 into a hydrophobic cavity formed by V217,F241 and V240 of helices 3 and 4 of the CA_NTD. K. The N-terminal residues F168 and M165 preceding helix 1 dock into two distinct hydrophobic grooves of CA_NTD. See also Figure S2–4 and Movies S1–3.

Figure 3.. Endogenous PNMA2 is released from cells as non-enveloped capsids.

A. Media was collected from primary cultured rat cortical neurons (DIV15–16) and fractionated using size exclusion chromatography (SEC). Fractions were run on a gel and blotted for PNMA2 and ALIX, a canonical EV marker. PNMA2 protein is released in early fractions that contain EV proteins. B. The early fractions (1–4) from primary cultured cortical neuronal media SEC were pooled and blotted for Arc, PNMA2, and ALIX. Fractions were incubated with Proteinase K (7μg/mL) with or without detergent present (1% Triton X-100) for 10 mins. Representative Western blots show that PNMA2 protein was sensitive to Proteinase K degradation without detergent present. C. Quantification of western blots in (B) (n=3 cultures). Error bars indicate mean ± S.E.M. D. The early SEC fractions (1–4) from primary cultured cortical neuronal media were further fractionated using ultracentrifugation, and an iodixanol gradient was used to separate proteins by density and size. PNMA2 protein was enriched in fraction 6, while ALIX was enriched in fractions 3 and 4. E. Representative negative-stained EM image of non-enveloped PNMA2 capsids isolated from iodixanol gradient fraction 6 in (D). Scale bar: 50nm. F. Media was collected from cultured NTERA-2 testis cancer cells and fractionated using SEC. Fractions were run on a gel and blotted for hPNMA2 and ALIX. hPNMA2 is released in early fractions that contain EV proteins. G. The early SEC fractions (1–3) from NTERA-2 media were pooled and blotted for hPNMA2, and ALIX. Fractions were incubated with Proteinase K (7μg/mL) with or without 1% Triton X-100 for 10mins. Representative western blots show that hPNMA2 protein was sensitive to Proteinase K degradation without detergent present. See also Figure S5.

Figure 4.. mPNMA2 capsid assembly mutants show reduced release from cells.

A. Representative Coomassie SDS-PAGE gels show peak fractions eluted from the SEC column for purified mPNMA2 WT, mPNMA2 L270Q/L325Q (L/Q), and mPNMA2 Y162A (Y/A) mutant proteins. mPNMA2 L/Q and Y/A proteins are shifted to later fractions, indicating lack of assembled capsids. B. Representative negative-stained EM images of mPNMA2 WT, L/Q, and Y/A proteins from pooled SEC fractions 15–17 and 19–21. mPNMA2 L/Q and Y/A proteins do not form capsids. C. Mass distribution histograms of purified mPNMA2, L/Q, and Y/A proteins, measured by mass photometry. mPNMA2 L/Q and Y/A proteins are not found in high-molecular-weight assemblies (> 3000 kDa). D. Western blot of mPNMA2 WT, L/Q, and Y/A expression in cell lysates and media from transfected HEK 293T cells. L/Q and Y/A proteins are released less than WT mPNMA2 protein (n=4 cultures for each group; ****One-way ANOVA with Dunnett’s multiple comparisons, P <0.0001. WT vs. Y/A: ** P=0.0014; WT vs. L/Q: ****P<0.0001). E. LDH cytotoxicity assay to test the viability of HEK 293T cells in (D). No differences in toxicity were observed (One-way ANOVA with Dunnett’s multiple comparisons, P= 0.9271; WT vs. Y/A: P= 0.9925; WT vs. L/Q: P= 0.9028). Error bars indicate mean ± S.E.M.

Figure 5.. PNMA2 capsids induce autoantibody production.

A. Mice were injected intraperitoneally with vehicle (n=4), 5μg purified mPNMA2 capsids (n=4) or 5μg mPNMA2 L/Q protein (n=4) and blood sera collected 3 weeks after injections. Sera were analyzed for antibody production using ELISA, using 2μg/mL mPNMA2 WT capsids or mPNMA2 L/Q protein coated on the plates as the antigen. mPNMA2 capsid-injected mice produced robust PNMA2 autoantibodies, whereas vehicle and mPNMA2 L/Q injections did not elicit autoantibody production. (Mann-Whitney test. Vehicle vs. WT (using WT capsids as antigen): P=0.0286; Vehicle vs. L/Q (using L/Q as antigen): P=0.8857). B. Representative negative-stained EM images of purified mPNMA2 capsids immunogold labeled with mouse PNMA2 autoantibodies in serum collected from mPNMA2 capsid-injected mice. C. Mice sera collected 3 weeks after the second injection of vehicle or 5μg mPNMA2 L/Q were analyzed for antibodies against mPNMA2 L/Q protein by ELISA. Antibodies were not produced after the second injection of L/Q. (Mann-Whitney test, P=0.1143). D. Purified mPNMA2 spike fragments and capsid shell fragments (see schematic) were used as ELISA antigens to determine the epitopes of PNMA2 autoantibodies from mPNMA2 capsid-injected mice. The schematic shows mPNMA2 protein regions located on the spike and the capsid domains of capsids, as determined by the cryoEM structure. PNMA2 autoantibodies preferentially bind to the spike fragments. (****One sample t-test, P<0.0001, null hypothesis of 50% binding). E. An ELISA, using purified hPNMA2 capsids as the antigen, was used to quantify PNMA2 autoantibodies in CSF from patients diagnosed with or without PNMA2-related paraneoplastic neurological syndrome. (n=3 for each group, *Mann-Whitney test, P=0.05). F. Representative negative-stained EM images of hPNMA2 capsids labeled with immunogold using CSF from control or paraneoplastic patients. G. ELISAs using hPNMA2 spike and capsid shell fragments as antigens show patient CSF PNMA2 autoantibodies preferentially bind to the spike fragment (*One sample t-test, P=0.0356, null hypothesis of 50% binding). Error bars indicate mean ± S.E.M. Abbreviation. WT: wild type; L/Q: L270QL325Q; Veh: vehicle. See also Table S1 and Table S2.

Figure 6.. mPNMA2 capsids activate bone marrow-derived dendritic cells.

A. Murine bone marrow cells were isolated and cultured with 20 ng/mL GM-CSF for 7 days to generate bone marrow-derived dendritic cells (BMDC), which were then incubated with varying concentrations of LPS, mPNMA2 capsids, or L/Q mPNMA2 protein for 24 hours prior to flow cytometric analysis. Data show quantification of matured (CD11c+MHCII++) BMDC following stimulation, graphed as fold increase relative to unstimulated BMDC of the same experiment. n=4 independent experiments. Both WT and L/Q mPNMA2 protein activates BMDCs. B. Mean fluorescence intensity (MFI) of CD80 (left), CD83 (middle), and CD86 (right) within the matured BMDC population graphed as fold increase relative to unstimulated BMDC of each independent experiment with 50μg antigen stimulation. Both WT and L/Q mPNMA2 protein induce costimulatory receptor expression. C-G. Cell culture supernatants of BMDC from (A) were collected and analyzed for expression of cytokines CXCL1 (C), CCL4 (D), CCL11(E), IL18(F), IL5(G) using the ProcartaPlex immunoassay. n=3 biological replicates. WT mPNMA2 capsids induce significantly higher levels of cytokine release than L/Q mPNMA2 protein. Error bars: mean ± SEM. Statistics were determined using a two-way ANOVA with Dunnett’s multiple comparisons test using unstimulated values as a control (A), one-way ANOVA with Dunnett’s multiple comparisons test using unstimulated values as a control (B) or unpaired t test (C-G), *p<0.05, ***p<0.001, ****p<0.0001. See also Figure S6.

Figure 7.. mPNMA2 capsid-injected mice develop learning and memory deficits.

A. Schematic of the fear conditioning behavior paradigm. Mice were injected with 5μg of mPNMA2 capsids or capsid mutant L/Q mPNMA2 protein, followed by a booster injection of 5 μg protein three weeks later. Three months after the second injection, mice were conditioned to associate mild foot shocks with a specific context. Abbreviations: ITI: inter-trial interval. B. We measured the percentage of mice freezing during fear conditioning. mPNMA2 capsid-injected mice showed decreased freezing during the training. Baseline was assessed 30s before the 1^st shock. ITI 1–4 were assessed as 30s bins from the 30^th second after the shock. Two-way ANOVA with Tukey’s multiple comparisons test. Main effect, shock: F (4, 55) = 15.67, ****P<0.0001; injected antigen: F (2, 55) = 9.691, ***P=0.0002; Non-injected vs. Capsids, ***P=0.0001; Non-injected vs. L/Q mutant, P= 0.1190; Capsids vs. L/Q mutant, P=0.0846. Abbreviations: Base: baseline. ITI: inter-trial interval. C. The maximum motion of mice during the shock (measured 5s from the beginning of shock presentation). There was no difference between PNMA2 capsid and L/Q injected mice, indicating normal pain perception of the shock. Two-way ANOVA. Main effect, shock: F (3, 44) = 2.110, P=0.1125; Injected antigen: F (2, 44) = 0.5415, P=0.5857. D. A memory retention test was performed 24h after training. mPNMA2 capsid-injected mice showed a significant decrease in freezing, compared with non-injected and L/Q injected mice, which showed robust memory recall. One-way ANOVA test with Tukey’s multiple comparisons test, ****P<0.0001. Non-injected vs. L/Q mutant: P= 0.1405; Non-injected vs. Capsids: ****P<0.0001; L/Q mutant vs. Capsids: **P=0.0012. B-D, non-injected group: n= 6 mice (3 male+3 female); L/Q and capsids injected group: n=4 mice (2 male +2 female). Error bars indicate mean ± S.E.M. See also Figure S7.