ZCCHC3 is a stress granule zinc knuckle protein that strongly suppresses LINE-1 retrotransposition

Abstract

Retrotransposons have generated about half of the human genome and LINE-1s (L1s) are the only autonomously active retrotransposons. The cell has evolved an arsenal of defense mechanisms to protect against retrotransposition with factors we are only beginning to understand. In this study, we investigate Zinc Finger CCHC-Type Containing 3 (ZCCHC3), a gag-like zinc knuckle protein recently reported to function in the innate immune response to infecting viruses. We show that ZCCHC3 also severely restricts human retrotransposons and associates with the L1 ORF1p ribonucleoprotein particle. We identify ZCCHC3 as a bona fide stress granule protein, and its association with LINE-1 is further supported by colocalization with L1 ORF1 protein in stress granules, dense cytoplasmic aggregations of proteins and RNAs that contain stalled translation pre-initiation complexes and form when the cell is under stress. Our work also draws links between ZCCHC3 and the anti-viral and retrotransposon restriction factors Mov10 RISC Complex RNA Helicase (MOV10) and Zinc Finger CCCH-Type, Antiviral 1 (ZC3HAV1, also called ZAP). Furthermore, collective evidence from subcellular localization, co-immunoprecipitation, and velocity gradient centrifugation connects ZCCHC3 with the RNA exosome, a multi-subunit ribonuclease complex capable of degrading various species of RNA molecules and that has previously been linked with retrotransposon control.

Fig 1. ZCCHC3 protein expressed in 293T cells forms multimers and associates with L1 RNP complexes.

(A) ZCCHC3-FL and V5-TEV-ZCCHC3 proteins co-IP on α-FLAG-M2 affinity gel indicating their binding and multimerization. This is further supported by a higher molecular weight band visible at about 95 kD (top panel). In the lower panel we assume bands smaller than the 44 kD full-length ZCCHC3-FL are degradation products occurring during the IP and elution protocol. Vector only control was pcDNA6 myc-his B (pcDNA6). α-DYKDDDDK is an antibody product of Cell Signaling Technology and recognizes FLAG epitope tag. (B) FLAG-tagged ZCCHC3 IPed from HEK 293T cell lysates binds untagged L1 ORF1p complexes coexpressed from a full-length L1 (construct pc-L1-RP). ORF1p was detected by monoclonal α-ORF1-4H1 antibody [38]. Interaction was lost following treatment of lysates with RNase (+, last lane). (C) FLAG-HA-tagged ORF1p expressed from a full-length L1 (pc-L1-1FH) co-IPs V5-TEV-ZCCHC3 complexes from 293T cells. Interaction was diminished by treatment with RNase (+, last lane). (D) FLAG-tagged ZAP, MOV10, and ZCCHC3 co-IP with endogenous L1 ORF1p following affinity purification with α-FLAG-M2 antibody bound to Protein G Dynabeads (left). For this Western we used polyclonal α-L1Hs-ORF1p antibody. The bar chart (right) shows fold-enrichment of bound ORF1p after bands were quantified with ImageJ software (IP/ Input compared with empty vector). Association of ORF1p with the zinc-finger mutant ZCCHC3-FL-7AA was reduced 2-fold compared with wild-type ZCCHC3-FL. Empty vector pcDNA6 was used in duplicate as a negative control. (E) ZAP-L-FL, FL-MOV10, ZCCHC3-FL, and ZCCHC3-FL-7AA co-IP endogenous L1 RNA following affinity purification with α-FLAG-M2 antibody bound to Protein G Dynabeads. RT-qPCR was used to quantitate in the input and IP fractions L1 mRNA (blue bars; detected with the N51 primer pair targeting ORF1 sequence), and as controls β-actin mRNA (red bars) and GAPDH mRNA (green bars). Analyses by real-time RT-qPCR used the Relative Standard Curve Method. For each construct, we calculated the ratio of RNA of the IP fraction/ RNA of the input fraction compared with the empty vector (shown as fold enrichment on the y-axis). Results shown are the average of triplicate RT-qPCR reactions for one IP experiment. (F) α-ZCCHC3-CS antibody bound to Protein G Dynabeads was used in two separate IP reactions [1 and 2] to show that endogenous ZCCHC3 protein binds endogenous L1 ORF1p. Interaction was lost with RNaseA treatment (+, lanes 6 and 7). Normal rabbit IgG linked to Dynabeads was used as negative control and did not bind ORF1p (last two lanes). n = number of biological replicates.

Fig 2. Expression of ZCCHC3 protein is not toxic to cells and alters retrotransposition in cell culture reporter assays.

To confirm lack of cytotoxicity due to ZCCHC3 overexpression, 293T cells transfected with tagged ZCCHC3 were (A) stained on day 4 post-transfection with trypan blue and counted using a Cellometer Auto T4 Cell Viability Counter (Nexcelom) or (B) assayed with the MultiTox-Fluor Multiplex Cytotoxicity Assay (Promega). (C) ZCCHC3 exogenous expression inhibits cell culture L1 retrotransposition. The L1 reporter construct 99-PUR-RPS-EGFP [45] was cotransfected in 293T cells with empty vector (pcDNA6 or pcDNA3) or constructs expressing wild-type or mutant ZCCHC3 or ZCCHC7. Five days later, percentages of EGFP-positive cells (i.e., cells with a retrotransposition event) were determined by flow cytometry. Each construct pair was tested in four replicate wells in at least three replicate experiments, and results are normalized to empty vector control (lighter bar). All constructs significantly reduced retrotransposition compared with vector only control (t-test p-values are shown above each bar, *: p<0.05, **: p<0.01, ***: p<0.0001). Bottom panels: Western blots show expression of the tagged test constructs. (D) Loss of endogenous ZCCHC3 expression increases retrotransposition over three-fold. 293T cells were transfected with a scrambled shRNA cloned in vector pLKO.1-TRC (CONT shRNA-16) [46] or shRNA sequences directed against ZCCHC3 [20] and tested for retrotransposition competency of 99-PUR-RPS-EGFP (normalized to empty vector control, lighter bar). Bottom panels: Western blots showing that ZCCHC3 shRNA-2 and ZCCHC3 shRNA-4 strongly decreased endogenous ZCCHC3 protein levels in 293T cells, but had no effect on levels of HSP90. Bottommost panel: following antibody detection, Western transfer blots were stained with Ponceau S as an additional loading control. (E) KO of ZAP expression in 293T cells by CRISPR causes 99-PUR-RPS-EGFP retrotransposition to increase greater than 2-fold. Results are normalized to vector control (lighter bar). Bottom panel: Western blotting shows loss of ZAP protein expression in KO lines. (F) Overexpression of ZCCHC3 strongly inhibits mouse IAP LTR element retrotransposition in HeLa-JVM cells in a neomycin phosphotransferase cell culture reporter assay using vector IAP-neo^TNF [48]. The y-axis shows percentage of G418-resistant (retrotransposition-positive) colonies normalized to vector control (lighter bar). n = number of biological replicates.

Fig 3. ZCCHC3 protein limits endogenous L1 protein but not L1 RNA expression and co-localizes in stressed cells with markers of cytoplasmic stress granules.

(A) Overexpression of ZCCHC3-FL protein significantly (p<0.01) reduces levels of endogenous ORF1p in 293T cells. A sample Western blot from 4 experiments is shown, along with Ponceau S staining of the blot as a loading control (bottom right). Using ImageJ analysis, endogenous ORF1p band intensity was calculated as a percentage of summed Ponceau S-stained bands and normalized to empty vector control (lighter bar, histogram left). n = number of biological replicates. (B) On the other hand, neither overexpression (left) nor shRNA KD (right) of ZCCHC3 protein in 293T cells significantly alters endogenous L1 mRNA levels assayed by RT-qPCR and ΔΔCt analysis. The graph shows L1 RNA levels detected by primer pairs, 5UTR L1Hs (targeting the L1 5’UTR), N51 (targeting ORF1) and N22 (ORF2), relative to the empty vector condition. For each case, GAPDH RNA was used for normalization, and data shows the average of three biological replicates ± s.d. Experimental data were analyzed by two-way ANOVA followed by Sidak´s multiple comparison to calculate statistical significance. ns: not significant. (C and D) Endogenous ZCCHC3 protein, detected by an Aviva Systems Biology antibody (α-ZCCHC3-Av), does not form obvious SGs in untreated 2102Ep or U2OS-WT cells but (E and F) does when cells are stressed with 0.25 mM Na-arsenite. (G) Abclonal α-ZCCHC3-Abl antibody also detects endogenous ZCCHC3 in SGs of arsenite-treated U2OS-WT cells. eIF3η and TIA-1 are canonical SG marker proteins. Cell nuclei were stained with Hoechst 33342 (right-most panels). Size bars are 10 μm.

Fig 4. Quantification of cytoplasmic granule formation by tagged ZCCHC3 in wild-type or G3BP1/2 DKO U2OS cell lines.

(A) Endogenous ZCCHC3 protein enters SGs marked by endogenous G3BP Stress Granule Assembly Factor 2 (G3BP2) in stressed HEK 293T cells. (B) Summarizing the data of micrograph experiments (C-E), bar graphs were generated by counting at least 200 cells for each transfection. n = number of experimental replicates. (C) FLAG-tagged ZCCHC3 does not form SGs in unstressed U2OS-WT cells, (D) but does when the cells are stressed with Na-arsenite. (E) However, ZCCHC3-FL-containing granules do not form in G3BP1/2 U2OS-DKO cells, even when stressed with Na-arsenite. (F) Similarly, endogenous ZCCHC3 detected by α-ZCCHC3-Av antibody does not form granules in stressed U2OS-DKO cells. NT: no treatment. Cell nuclei were stained with Hoechst 33342 (right-most panels). Size bars are 10 μm.

Fig 5. Analysis of cytoplasmic granule formation by ZCCHC3 and L1 ORF1p.

(A) Endogenous ORF1p forms granules in untreated, (B) and larger granules in arsenite-stressed 2102Ep cells. However, endogenous ZCCHC3 colocalizes with LINE-1 ORF1 granules in treated 2102Ep cells only (B). (C) However, unlike in untreated 2102Ep cells, endogenous LINE-1 ORF1p does not enter cytoplasmic granules in untreated U2OS-WT cells, (D) but only after Na-arsenite stress. (E) Bar graphs summarize the data of micrograph experiments (F-I). Data were generated by counting at least 200 cells for each transfection. n = number of experimental replicates. GFP-tagged L1 ORF1p expressed from the construct ORF1-EGFP-L1-RP is able to form cytoplasmic granules in both unstressed or stressed (F,G) U2OS-WT or (H,I) U2OS-DKO cells. NT: no treatment. Cell nuclei were stained with Hoechst 33342 (right most panels). Size bars are 10 μm.

Fig 6. ZCCHC3 protein associates with MOV10 and ZAP proteins.

(A) Immunoprecipitated FLAG-tagged ZCCHC3 protein associates with both exogenously expressed (top panels) and endogenous (bottom panels) MOV10 protein in an RNA-dependent manner (last two lanes, -/+ RNaseA treatment). Control vectors were pcDNA5 FRT/TO (top panels) and pcDNA6 myc/his B (bottom panels). (B) α-ZCCHC3-CS antibody bound to Protein G Dynabeads was used in two separate IP experiments [1 and 2] to show that immunoprecipitated endogenous ZCCHC3 protein binds endogenous MOV10 protein and that this interaction is lost with RNaseA treatment (lanes 6 and 7). Normal rabbit IgG was used as negative control and showed no binding of endogenous MOV10 protein (last 2 lanes). (C) FLAG-tagged ZCCHC3 IPs endogenous ZAP protein in an RNA-independent manner (last two lanes, -/+ RNaseA treatment). To generate this data, the Western blot of Fig 6A (lower panel) was stripped. (D) FLAG-tagged ZAP-L long isoform (PARP13.1) co-IPs V5-TEV-ZCCHC3 protein in an RNA-independent manner (last two lanes, -/+). Experiments in A-D were performed in 293T cells. (E and F) Tagged V5-TEV-ZCCHC3 colocalizes with endogenous MOV10 and ZAP proteins in cells stressed with Na-arsenite. (G) Exogenously expressed ZCCHC3-FL and V5-TEV-TRIM25 proteins colocalize in cytoplasmic granules of stressed U2OS-WT cells. (H) Multiple 293T cell lines deleted for ZAP expression by CRISPR protocol (in red text, KO cell lines AE-BB-B8-P3 and AE-CC-D8-P9 are shown as examples) have diminished expression of endogenous ZCCHC3 compared with control lines (top panels). Control lines (ex., AB-B-B5-P5 and AE-CC-E4-P9) were similarly subjected to CRISPR treatment but without successful KO of ZAP expression. Full-length (ZAP-L) and the short (ZAP-S) isoforms of ZAP are indicated by arrows on the Western blot. Reprobing with α-HSP90 antibody and staining of Western transfer blots with Ponceau S (bottom panels) are controls for equal sample loading. Bands were quantitated by ImageJ software for 5 ZAP KO and 5 control cell lines. For each pair of cell lines (KO and control), endogenous ZCCHC3 band intensity was calculated as a percentage of summed Ponceau S-stained bands between the two arrows indicated (bottom right panel) and normalized to empty vector control (histogram, left).

Fig 7. FLAG-tagged ZCCHC3 co-IPs coexpressed tagged EXOSC5 and DIS3L exosome subunits but not other selected (A) cytoplasmic SKI complex components (SKIV2L, TTC37, WDR61, and HBS1LV3), (B) exosome catalytic subunits (EXOSC10, DIS3, and DIS3L), or (C) subunits of the exosome core (EXOSC3, EXOSC5, EXOSC7, and EXOSC8).

In (A), the HBS1L short splicing isoform HBS1LV3 reportedly links the SKI complex and cytoplasmic exosome in humans (the canonical HBS1LV1 variant does not associate with the exosome) [83]. In (B), the DIS3L isoform 2 used differs from the canonical sequence by lacking the first 83 amino acids (Acc. # NP_001310865.1). The control empty vectors were pcDNA5 in (A) and pcDNA6 in (B and C). (D) Endogenous ZCCHC3 binds coexpressed V5-TEV-EXOSC5 when immunoprecipitated by α-ZCCHC3-CS antibody bound to Protein G Dynabeads. The control empty vector used was pcDNA3.1/nV5-DEST. The amount of immunoprecipitated EXOSC5 protein increased with RNase treatment (+, last lane). Alpha-tubulin is shown as a loading control (bottom panel). All experiments were performed in HEK 293T cells.

Fig 8. ZCCHC3 protein colocalizes in SGs with exosome component EXOSC5 and cosediments in a sucrose gradient with selected proteins of the exosome as well as MOV10 and ZAP.

(A,B) Both ZCCHC3-FL and the short ZAP isoform tagged with GFP (GFP-PARP-13.2) colocalize with tagged exosome component EXOSC5 in cytoplasmic granules of stressed 293T cells (see arrows). C) On the other hand, V5-TEV-DIS3L protein does not colocalize with ZCCHC3-FL in cytoplasmic granules of stressed U2OS-WT cells, (D) nor does endogenous ZCCHC3 protein colocalize with endogenous DIS3L nuclear homolog DIS3 in granules of 293T cells. (E) V5-TEV-ZCCHC3 does not colocalize with endogenous EXOSC3 in SGs of stressed 2U2OS-WT cells. Cell nuclei were stained with Hoechst 33342 (right-most panels). Size bars are 10 μm. (F) HEK 293T cells expressing ZCCHC3-FL were lysed, and lysates were fractionated by sucrose velocity gradient centrifugation. Aliquots of each 900 μl fraction were subjected to SDS-PAGE and Western blotting with antibodies against endogenous MOV10, ZAP, and exosome components. The ~42 kD band above the expected 30 kD EXOSC3 band (marked by an arrow, lower panel) appears more in HMW fractions and may reflect post-translational modification, possibly ubiquitination [123]. (G) Untransfected HEK 293T cell lysates were analyzed as in (F). (H) In 293T cells, protein expression from the V5-TEV-EXOSC3 construct is strongly inhibited by cotransfection of the shEXOSC3 shRNA construct, but not in untransfected or scrambled shRNA (shN3)-transfected controls [34]. (I) HEK 293T cells expressing ZCCHC3-FL and shN3 control (top) or shEXOSC3 (bottom) shRNA constructs were lysed and subjected to sucrose velocity gradient centrifugation as above and analyzed by α-FLAG-M2 antibody. Cotransfection of shEXOSC3 caused redistribution of ZCCHC3 protein throughout the gradient suggesting its dissociation from HMW complexes.