Elimination of HIV-1 Genomes from Human T-lymphoid Cells by CRISPR/Cas9 Gene Editing

We first tested the ability of our modified CRISPR/Cas9 gene editing system to eliminate the HIV-1 genome in a human T-lymphocytic cell line, 2D1011. These cells harbor integrated copies of a single round HIV-1_PNL4-3 whose genome lacks sequences encoding the majority of the Gag-Pol polyprotein, but encompasses the full-length 5′ and 3′ LTRs, and includes a gene encoding the marker protein green fluorescent protein (GFP) replacing Nef protein in the latent state (Fig. 1A). Thus, 2D10 is a suitable cell line to first establish proof-of-principle of HIV-1 eradication because of the uniform nature of the integrated provirus. Treatment of clonal 2D10 cells stably expressing Cas9, but not gRNAs, with proinflammatory agents such as phorbol myristate acetate (PMA) and/or the HDAC inhibitor trichostatin A (TSA) profoundly stimulates HIV-1 promoter activity, leading to production of the viral proteins and GFP in over 90% of treated cells (Fig. 1B, left panels), providing a convenient cell culture model for studying viral latency and reactivation. Co-expression of Cas9 along with gRNAs A and B, designed respectively to target the highly conserved sequence among all viral isolates spanning the LTR U3 region at nt −287/−254 (gRNA A) and nt −146/−113 (gRNA B) (Fig. 1A) completely eliminated PMA/TSA-induced GFP production, indicating inhibition of HIV-1 gene expression in the pre-selected mixed clonal population of T-cells expressing both Cas9 and gRNA expression plasmids (Fig. 1B, right panels). Expression of gRNAs and Cas9 was verified by RT-PCR and Western blot, respectively (Fig. 1C,D). HIV-1 expression was completely eliminated from the cells expressing both Cas9 and gRNA expression plasmids, shown by flow cytometry detection of GFP production by randomly-selected Cas9-positive clonal cells with or without gRNA expression (Fig. S2A). Also, we found that GFP production was effectively blocked in many clones that expressed only a single gRNA (A or B), to levels similar to those elicited by co-expression of both A and B (Fig. S2B; also see Fig. S2A), suggesting that expression of either gRNA in single configuration can initiate cleavage at both LTRs to achieve eradication of proviral DNA.

We verified the site(s) of HIV–1 proviral DNA integration by whole–genome sequencing (WGS) of 2D10 cells. We used CREST calling27 of the structural variation (SV) to identify breakpoints caused by proviral DNA integration in the host genome, and used the hg19 genome and the HIV–1 genome, KM390026.1↗ as reference genomes for reading the DNA sequences. We identified four inter–chromosomal translocations, designated by CTX (Fig. S3), that are related to HIV-1 DNA. Breakpoints between the HIV-1 5′ LTR and P163.3:1991382 and the HIV-1 3′ LTR P613.3:1991378 were detected, mapping to exon 2 of the methionine sulfoxide reductase B1 MSRB1 gene (NM_01332), and corresponding to a previously mapped location for the provirus in the 2D10 cells11,28. In addition, two CTXs were mapped to chromosome 1 with the breakpoint between P13.2:114338315 and the HIV-1 5′ LTR, and other breakpoints between HIV-1 3′ LTR and P13.2:114338320. Also, we noted that four nucleotides, TAAG, were deleted between the two breakpoints in chromosome 1P13.2. The HIV-1 provirus in chromosome 1, which was previously undetected by linker-addition mapping, was integrated in the second intron (114339984–114320431) of the round spermatid basic protein 1 (RSBN1) gene (NM_018364↗). A schematic presentation of identified consensus sequences for sites of HIV-1 DNA integration in chromosomes 1 and 16 are shown in Fig. S3.

Short-range amplification assay of LTR DNA revealed an expected 497-bp DNA fragment in control cells and a second DNA fragment of similar size (504 bp) after treatment with Cas9/gRNAs A and B (Fig. 2A). Results of direct DNA sequencing of the PCR amplicon suggested that the observed 504-bp DNA fragment in Cas9/gRNA-treated cells was created by joining of the residual 5′ LTR to the remaining 3′ LTR after cleavage by Cas9/gRNA B (Fig. 2B). An Indel mutation with a seven-nucleotide insertion was also detected in the junction of the 5′ and 3′ fusion site of the clonal cells (Fig. 2B). The 257-bp PCR amplicon corresponding to the Rev response element (RRE), which is positioned in the center of the viral genome, was absent, verifying that Cas9/gRNAB removed the DNA sequences spanning between the two terminal repeats (Fig. 2A). Long-range PCR analysis of 2D10 control cells expressing Cas9 but not gRNAs, using a pair of primers derived from the second intron of RSBN1, verified the presence of a 6130-bp DNA fragment corresponding to the integrated HIV-1 genome plus its chromosome 1-derived flanking DNA sequence (Fig. 2C). The 264-nucleotide DNA fragment that represents host cell DNA sequence from the other copy of chromosome 1 was also present (shown at the bottom of the gel). In cells treated with Cas9/gRNAs A and B, a DNA fragment of 6130 nucleotides corresponding to the integrated HIV-1 genome was completely absent. Instead, PCR amplification produced a smaller DNA fragment of 909 nucleotides. Sequencing of the amplicon verified excision of the integrated viral DNA, spanning between the B domain of the 5′ LTR and the B domain of the 3′ LTR (Fig. S4A,B). Again, we detected a 264-nucleotide DNA fragment amplified from the host genome from the other chromosome (Fig. 2C).

We examined chromosome 16 for presence of HIV-1 proviral DNA using long-range PCR using a primer pair corresponding to the second exon of MSRB1 gene and compared its status in Cas9/gRNA A/B-treated cells. The results showed that the expected 5467-bp DNA fragment of the HIV-1 genome and its flanking host DNA in chromosome 16 was absent. Instead we detected a smaller 759-bp DNA fragment that reflected joining of the residual U3 region of the 5′ LTR after cleavage by gRNA A to the remaining U3 region of the 3′ LTR upon cleavage by gRNA B (Fig. 2D). Direct sequencing of the 759-bp DNA fragment identified the sites of viral DNA excision (Fig. S5A,B). A smaller, 110-bp DNA fragment found resulted from amplification of host DNA from the other copy of chromosome 16. These observations provide strong evidence that the gene editing molecules used effectively eliminate multiple copies of the integrated proviral DNA of the HIV-1-genome, which are scattered among various chromosomes.

To further validate the efficiency of our Cas9/gRNA treatment–based gene editing strategy in eliminating HIV– 1 proviral DNA from latently–infected T–cells, we analyzed the occurrence of insertion/deletion (InDel) and single nucleotide polymorphisms (SNP) in the HIV–1 genomes of control and HIV–1 –eradicated cells, using GATK calling29 against reference HIV–1 DNA (GenBank accession #KM3900261). Consistent with the results shown in Fig. 2, the reads from the whole-genome sequencing mapped to the 5′- and 3′-LTRs and to the proviral genome in the control 2D10 cells, supporting the precision and reliability of the deep coverage of the HIV-1 DNA by genome sequencing (Fig. 3A). In Cas9/gRNA-treated 2D10 cells, the integrated genomics view (IGV) revealed complete removal of a large DNA fragment corresponding to the HIV-1 proviral DNA with reads that map to the 3′ LTR (Fig. 3B). Reads mapping to the entire proviral genome between the two LTRs were completely absent, suggesting that both copies of the integrated HIV-1 genome in host cells were fully eliminated, and that Cas9/gRNA expression in a single clonal cell can attain 100% gene editing/elimination, perhaps attributable to repeated genome editing by stably-expressed Cas9/gRNA. Further, these results suggest that, after cleavage of viral 5′ and 3′ LTRs and excision of the viral genome, viral DNA is likely degraded, such that no transposition or re-integration occurs into the host genome.

To determine the repair events after Cas9/gRNA A/B–induced cleavage of both LTRs, we used BWA calling27 of the structural variant (SV) in the DNA from cells with HIV–1 excision, and identified the breakpoints of large insertions and/or deletions. The results verified that no excised HIV–1 DNA from one chromosome was inserted in the host genome and/or in the integrated copy of proviral DNA on the other chromosome further ruling out the notion of re-integration of the excised viral DNA into host cell genome. However, we identified three breakpoints caused by deletion of the DNA fragments corresponding to sites of viral DNA integration into the host genome. One left breakpoint positioned at the end of the 5′ LTR at nucleotide 636 (=HIV: 9710) as supported by 10 reads. One right breakpoint exhibited two patterns, one at HIV: 9073 (=HIV:-3) supported by 6 reads with 2 C → G and 4 C → T conversions; and HIV: 9075 (=HIV:-1) supported by 63 reads (Fig. 3C,D). Of note, these two breakpoints can actually reflect the presence of the entire 634 nucleotides of the LTR after the excision of full proviral DNA by Cas9/gRNAs A or B at the 5′ and 3′ LTRs followed by precise rejoining of the DNA at the cleavage site. A third breakpoint is located in the middle of the 3′ LTR at nucleotide 9389 (=HIV: 313) with C insertion supported by 87/161 reads and CTAAGTT insertion supported by 69/161 reads (Fig. 3E). This breakpoint represents the joining of DNA after cleavage at sites A and B of the 5′ and 3′ LTRs (Fig. 3F).

We also investigated the impact of CRISPR/Cas9-mediated excision of HIV-1 proviral DNA from the RSBN1 gene on the level of RNA production from RSBN1 and several of the other cellular genes positioned in close proximity of the proviral insertion sites, as shown in Fig. 4A. Results from RT-PCR of five controls and five HIV-1 eradicated single cell clones indicated no significant effect on the level of expression of RSBN1, although smaller variations of less than 0.4-fold were detected in the levels of neighboring RNA (Fig. 4B), which may not be attributed to the gene editing strategy and may not impact the overall expression of their proteins. Similarly, elimination of the HIV-1 genome from chromosome 16 showed no significant impact on the expression of the site of integration, i.e. MSRB1 gene and its surrounding gene (Fig. 4C,D). We also investigated the effect of Cas9/gRNAs A and B on several parameters related to the health of the cells, including cell viability, cell cycle progression and apoptosis using several clonal cells after the eradication of HIV-1 by Cas9/gRNAs A and B. We found no persistent and significant deleterious effects on the host cells vital signs after elimination of the HIV-1 proviral DNA by the Cas9/gRNAs A and B (Figs S6–8).

To expand the scope of analysis of potential off-targets, we bioinformatically compared the InDel results from whole-genome sequencing of the 2D10 cells after treatment with the Cas9/gRNAs A/B system that elicited complete eradication of the proviral HIV-1 DNA. To improve InDel-calling confidence, we aimed the whole-genome sequencing at 100× coverage, but statistical analyses revealed that we actually achieved total coverage of 109.3× for control cells and 112.7× for HIV-1 eradicated cells (). Coverage levels varied for each chromosome, ranging >96× for chromosome 1 and >110× for chromosome 16 (). Using human (hg19) genome as a reference sequence, we identified 1,361,311 InDels (<50 bp insertions/deletions) in Control (+Cas9/−gRNA) and 1,358,399 in HIV-1-eradicated cells (), and 3,973,098 single nucleotide polymorphisms (SNP) in Control and 3,961,395 in HIV-1-eradicated cells (). Comparative bioinformatics analysis between the control and the HIV-1-eradicated cells identified 32,399 somatic InDels (small insertion/deletion called by Strelka), 46,614 somatic SNVs (single-nucleotide variations, called by MuTect) and 52 SVs (structural variations including large InDels called by CREST) between the latter two groups, that were distributed in different genomic regions (). Table S1 Fig. S9 Table S2 Table S3 Table S4

After discarding the small InDels we found in the public database dbSNP, we identified 30,156 InDels and 43,858 SNVs in HIV-1-eradicated cells. Filtering out heterozygous mutations, reduced this number to 989 InDels. To determine if these filtered InDels are de novo mutations caused by our Cas9/gRNA A/B editing system, we extracted ±30 bp, ±300 bp or ±600 bp sequences flanking each filtered InDel and used Blastn (e-value cutoff: 1000) to compare them vs. the potential gRNA off-target host genome sites predicted by sequence similarity at 0–7 mismatches, and vs. HIV-1 on-target sequences. Without any mismatches to targets of gRNAs A and B, we found no off-target site around the extracted 60, 600 and 1200 bp sequences of the filtered InDels. Within the extracted 60-bp sequences, we found no off-target site even with 7 mismatches at alignment lengths >12 nucleotides from PAM NRG (which must be 100% matched). Within the extracted 600-bp sequences, we found no off-target site with 3 mismatches for targets of gRNA A or B. With 4–7 mismatches, we found only one potential off-target site with 6 mismatches at an alignment length of 20 bp from PAM and another with 3 mismatches at 12 bp alignment length from PAM for Target A, and one additional potential off-target site with 4 mismatches at 16 bp length from PAM for Target B. Within the extracted 1200 bp sequences for 3 mismatches, we found no off-target sites for Target A but one potential off-target with 2 mismatches at 13 bp from PAM for Target B. With criteria of 3–7 mismatches against the 1200-bp sequences, we found only six potential off-target sites for Target A and two potential off-target sites for Target B (Fig. 4E). Together, these data strongly suggest that none of the indels detected in the cells with the excised HIV-1 genome lie within 60 bp of Targets A or B of any potential off-target sites, as predicted by search criteria allowing up to 7 mismatches. By expanding the searching sequences to 600 or 1200 bp, relatively rare off-target sites were identified, including various numbers of mismatches and aligned length. With perfect match to the last 12 bp seed sequence plus PAM NRG, none of the indels fell within the search area of 60–1200 DNA sequences. Our overall interpretation of these data verifies the preceding Surveyor assay results in these cells, as well as in the other cell types25, and establishes by very stringent analysis that no off-target effects upon the host T cell genome is elicited by our Cas9/gRNAs HIV-1 DNA-excising system.

We selected several T-cell clones whose proviral DNA was eliminated by Cas9/gRNAs and maintained at various levels, expression of Cas9 as well as the gRNAs to assess the extent of new infection by HIV-1. As seen in Fig. S10A, clone C7 expresses Cas9 but not gRNA B, whereas clone AB8 shows no detectable level of Cas9, yet contains gRNA B. Two additional clones AB9 and AB5 with an equal amount of gRNA B and different levels of Cas9 expression were selected for a re-infection study. Infection of these cells by HIV-1_NL4-gfp followed by longitudinal evaluation of viral replication by flow cytometry showed that cells expressing either Cas9 or gRNA B alone were infectable by HIV-1, and supported viral replication throughout the course of these studies (day 18 post-infection) (Fig. S10B). In contrast, cells expressing both Cas9 and gRNA B were resistant to infection by HIV-1 and failed to support viral replication. AB5, which expressed a higher level of Cas9, appeared to be more resistant to viral replication than AB9, which showed reduced Cas9 expression (Fig. S10A,B). Figure S10C summarizes the quantitative values of the results shown in Panel B. The results demonstrate that the intracellular presence of both Cas9 and the LTR-directed gRNAs can effectively protect culture of human T-cells against new infection by HIV-1.

We tested the ability of Cas9/gRNAs to suppress HIV-1 infection of CD4+ T-cells prepared from healthy individuals. We chose a lentivirus vector for delivering Cas9 and gRNA expression DNAs because of its high transduction efficiency and low toxicity. Results of the LV transduction showed efficient cleavage of the HIV-1 LTR DNA by the LVs expressing both Cas9 and gRNAs, but not in control cells transduced with LV expressing only Cas9 (Fig. 5A). Of note, our gRNAs do not cleave the LVs LTR, which lacks the U3 modulatory region, thus they have no effect on the expression of the LV genome. Accordingly, flow cytometry analysis revealed functional inactivation of the integrated HIV-1 genome in latently-infected T-cells upon transduction with LV-Cas9/gRNA (Fig. 5B). Again, we found no evidence for cell death that may be associated with Cas9/gRNAs in the primary cells, corroborating our observations shown in Fig. S6. Once the efficacy of the gene delivery of editing molecule by LV was verified in the T-cell line, we infected primary cultures of CD4+ T-cells with HIV-1_JRFL or HIV-1_PNL4-3, then transduced them with either control LV Cas9 or LV Cas9 plus LV gRNA (Fig. 5C). Compared to controls, a substantial decrease in HIV-1 copy number was seen in the CD4+ T-cells treated with LV Cas9/gRNA (Fig. 5D). Amplification of viral DNA revealed the expected 398-bp amplification in the control cells and a similar-sized DNA fragment with lesser intensity in cells transduced with LV Cas9/gRNA in CD4+ T-cells (Fig. 5E).

We also assessed the ability of lentivirus delivered Cas9/gRNA in editing the HIV-1 genome present in PBMC’s and CD4+ T-cells obtained from HIV-1+ patients during routine visits to the Temple University Hospital AIDS clinic. For this proof-of-concept study, initially we sought to prepare PBMCs and CD4+ T-cells from four patients (TUR0001 to TUR0004;Cases 1–4) who were undergoing antiretroviral therapy and exhibited diverse responses to treatment as determined by viral load assay and percentage of CD4+ cells (). The procedure we used to prepare PBMCs and CD4+ T-cells and examine CD4+ T-cells, and the timeline for lentivirus treatments and cell harvest are shown in. The purity of the CD4+ T-cells was confirmed by flow cytometry of FITC conjugated anti-CD4 antibody (). Fig. S11A Fig. S11B Fig. S11C

Results of transducing PBMC’s with lentivirus-Cas9 and lentivirus-Cas9/gRNA revealed a substantial decrease, 81% in Case 1 and 91% in Case 2, in the viral copy number of cell populations expressing Cas9 and gRNA (Fig. 6A). Similar results were obtained after lentiviral transduction of CD4+ T-cells, which showed >92% reduction in viral copies in Case 1 and 56% for Case 2 upon expression of both Cas9 and gRNA, compared to control cells expressing only Cas9 (Fig. 6B). Standard curves and amplification plots served for absolute quantification of β-globin and Gag gene copy number is shown in Fig. S13. Examination of Gag p24 gene production in the CD4+ T-cells confirmed viral replication was decreased in Case 1 (71%) and Case 2 (62%) upon single transduction of the cells with lentivirus-Cas9/gRNA compared to that seen with lentivirus-Cas9 (Fig. 6C). Also, we examined the level of Gag p24 in PBMCs obtained from Cases 3 and 4 after delivery of Cas9/gRNA by lentivirus. Results from this study showed 39% and 54% decrease in HIV-1 p24 production from Cases 3 and 4, respectively, after transduction of the cells with therapeutic lentivirus (Fig. S12).

Next, we assessed the nature of mutations introduced by Cas9/gRNAs in the patient samples, by amplifying and sequencing of viral DNA. Our initial gene amplification of the CD4+ T-cells using primers spanning −374/+43 failed to detect any band in Case 1 and in Case 2 a DNA band was observed in the control sample that lacked gRNA expression (Fig. 6D). This observation suggests that the HIV-1 genome sequence in case 1 may differ from those of the primers we used for gene amplification (Fig. 6D). In case 2, where we detected the expected DNA fragment in untreated cells, the mutations that were introduced by Cas9/gRNA may have eliminated the recognition of DNA sequence by the PCR primer, thus interfering with DNA amplification. The use of an alternative set of primers that recognizes different regions of the LTR led to production of the expected 398-nucleotide amplicon in all samples (Fig. 6E). It is possible that similar to the results from 2D10 cells after treatment with Cas9/gRNAs (shown in Fig. 2A), some of the 397 nucleotide DNA fragments seen in the presence of gRNA expression result from joining of the remaining 5′ and 3′ sequences of the viral LTR after excision of the entire HIV-1 coding sequence. Sequencing of the amplicon verified the effect of Cas9/gRNA on editing of the viral genome at the expected positions and showed the presence of InDel and single nucleotide variation (SNV) mutations within and/or next to the PAM sequence within the LTR (Fig. 6F).

Jurkat 2D10 reporter cell line has been described previously11 and cultured in RPMI medium containing 10% FBS and gentamicin (10 ug/ml). 2 × 10⁶ cells were electroporated with 10 ug control pX260 plasmid or pX260 LTR-A and pX260 LTR-B plasmids, 5 ug each (Neon System, Invitrogen, 3 times 10 ms/1350V impulse). 48 h later medium was replaced with one containing puromycin 0.5 ug/ml. After one week selection, puromycin was removed and cells were allowed to grow for another week. Next, cells were diluted to a concentration of 10 cells/ml and plated in 96 well plates, 50 ul/well. After two weeks, single cell clones were screened for GFP tagged HIV-1 reporter reactivation (12 h PMA 25 nM/TSA 250 nM treatment) using Guava EasyCyte Mini flow cytometer. The non-reactive clones were used for further analysis.

Buffy coat was obtained through CNAC Basic Science Core I (Lewis Katz School of Medicine at Temple University, Philadelphia). PBMCs were isolated from human peripheral blood by density gradient centrifugation using Ficoll-Paque reagent. (For details, see S1 Experimental Procedures).

Cloning lentiviral constructs as well as lentivirus packaging and purification, and transduction of primary culture are detailed in S1 Experimental Procedures.

HIV-1_JRFL crude stock used was prepared from supernatants of PBMCs infected with HIV-1 for 6 days, clarified at 3000 RPM for 10 minutes and 0.45 um filtered. HIV-1_NL4-3-EGFP-P2A-Nef reporter virus was prepared by transfecting HEK 293 T cells with pNL4-3-EGFP-P2A-Nef plasmid and processed like lentiviral stocks (see above). HIV-1_JRFL was titered using Gag p24 ELISA, HIV-1 NL4-3-EGFP-P2A-Nef by GFP-FACS of infected HEK 293 T cells.

In vitro HIV-1 infection, and viral detection and quantification including p24 ELISA, are described in S1 Experimental Procedures.

The single subclone control C11 and experimental AB5 from parent 2D10 T cells were validated for target cut efficiency and functional suppression of HIV-1 EGFP reporter reactivation. The genomic DNA was isolated with NucleoSpin Tissue kit (Macherey-Nagel) according to the protocol of the manufacturer. The genomic DNA was submitted to Novogene Bioinformatics Institute (http://www.novogene.com/en/↗) for WGS and bioinformatics analysis. Briefly, DNA quality was further verified on 1% agarose gels, DNA purity was checked using the NanoPhotometer® spectrophotometer (IMPLEN, CA, USA) and DNA concentration was measured using Qubit® DNA Assay Kit in Qubit® 2.0 Fluorometer (Life Technologies, CA, USA). The detailed procedures for genomic DNA preparation, sequencing and bioinformatics, and SURVEYOR assays are described in S1 Experimental Procedures.

Total RNA was extracted from Jurkat cells using RNeasy kit (Qiagen) with on column DNAse I digestion. Next 0.5 ug of RNA was used for M-MLV reverse transcription reactions (Invitrogen). For gRNA expression screening specific reverse primer (pX260-crRNA-3′/R, Table I.3 in S1 Experimental Procedures) was used in RT reaction followed by standard PCR using target A or B sense oligos as a forward primers (Table I.5 in S1 Experimental Procedures) and agarose gel electrophoresis. For checking neighboring genes expression oligo-dT primer mix was used in RT and cDNA was subjected to SybrGreen real time PCR (Roche) using mRNA specific primer pairs and β-actin as a reference (Table I in S1 Experimental Procedures).

GFP and RFP expression in Jurkat 2D10 cells was quantified in live cells using Guava EasyCyte Mini flow cytometer (Guava Technologies). For HIV-1 reporter virus titer, cells were trypsinized 48 h after infections, washed and fixed in 4% paraformaldehyde for 10 minutes, then washed 3 times in PBS and analyzed for GFP FACS. CD4 expression in primary T cells was checked by direct labeling with CD4 V5 FITC antibody (BD Biosciences) followed by FACS.

Jurkat cells were washed, counted and diluted to a density of 1 × 10⁵ cells/ml in PBS. For each sample, 100 uL of cells in suspension was mixed with 100 ul of room-temperature annexin V-PE staining reagent (Guava Nexin Reagent) and incubated for 20 minutes at room temperature in the dark. After incubation, samples were acquired using a Guava EasyCyte Mini flow cytometer.

Cell viability was assessed using propidium iodide staining. To 200 ul of live cells in suspension PI solution was added to final concentration 10 ug/ml. Samples were incubated for 5 minutes at room temperature in the dark. After incubation, samples were acquired using a Guava EasyCyte Mini flow cytometer.

Cells were washed with 1× PBS and then resuspended in 250 ul of 1× PBS at room temperature. This suspension was added drop-wise to 1ml of −20 °C 88% ethanol, for a final concentration of 70% ethanol. Cells were fixed overnight at −20 °C then washed, incubated with 10 ug/ml of propidium iodide and RNase A solution 100 ug/ml in 1× PBS for 30 minutes at 37 °C then samples were cooled to 4 °C and acquired using a Guava EasyCyte Mini flow cytometer.

Standard methods for protein expression including Western blot and immunocytochemistry were used. For details see S1 Experimental Procedures.