Decreased mitochondrial NAD+ in WRN deficient cells links to dysfunctional proliferation

Loss of WRN in mesenchymal stem cells disrupts cellular proliferation and metabolic pathways

The helicase WRN has an essential role in DNA repair and genomic maintenance, as evidenced by the disruption of cellular proliferation, accelerated senescence, and stem cell dysfunctions seen in primary cells from WS patients [2–5, 7–9, 26]. Additionally, metabolic abnormalities are common in WS patients, where mitochondrial dysfunction and compromised mitophagy play important roles [5, 6]. Using our recently published RNA sequencing data from MSCs with a constituent knock out of WRN (WRN^−/−) and an isogenic control [5], we evaluated significantly altered pathways between WT and WRN^−/− (Figure 1A). Multiple pathways related to cellular metabolism and mitochondrial function changed in WRN^−/− cells compared to WT cells, including the pentose phosphate pathway, oxidative phosphorylation, and pyruvate metabolism (Figure 1A). Also, proliferation related pathways were disrupted upon WRN deficiency. Interestingly, only 24 h treatment with 1 mM nicotinamide riboside (NR), an NAD⁺ precursor, rescued multiple pathways in the WRN^−/− cells, including increased expression of genes driving mitochondrial and metabolism-related pathways, as well as proliferation-related pathways (Figure 1B). This indicated that NR treatment targets mitochondria in WRN^−/− MSCs. The rescue of proliferation-related pathways supports our previous findings showing complete restoration of stem cell proliferation following NR or NMN treatment in a Drosophila model of WS, with the mechanism unknown [6].

Heatmap analysis of differentially expressed genes related to proliferation and NAD⁺ metabolism using fpkm (fragments Per Kilobase of transcript per Million mapped reads) values with a cut off at fold-change > 1.5 and p-value < 0.05 analyzed with DESeq2, showed a clustering of the WRN^−/− cells (WRN^−/− (Veh)), compared to another clustering of WT together with WRN^−/− with NR treatment (WT (Veh), WRN^−/− (NR)) (Figure 1C). This suggested that NR treatment of WRN^−/− cells at least partially restored the expression of this selection of genes to WT levels (Figure 1C). Using a correlation matrix, we saw that the expression of NAD⁺ related genes such as the rate limiting enzyme in the Salvage pathway, Nicotinamide phosphoribosyl transferase (NAMPT), which recycles nicotinamide (NAM) to NMN, and NAD Synthetase 1 (NADSYN1), which catalyzes the conversion of nicotinic acid adenine dinucleotide (NAAD) to NAD⁺, the final step in the de novo pathway correlated positively with one another, but negatively to both NMNAT1, NMNAT2, QRPT (Kynurenine pathway) and NNMT (Salvage pathway). NMNAT1 correlated positively with SIRT2, COX5A and NADK (not significant after adjusted p-value), while NMNAT2 correlated positively with SIRT7, NAPRT, PARP1, and NMRK1, all important in NAD⁺ biosynthesis (Figure 1D and Supplementary Table 1).

Due to premature senescent features of WS cells, and the essential role of WRN in DNA repair, we examined the correlation between NAD⁺ related genes and senescence/proliferation associated genes. We observed a positive correlation between NAMPT and NADSYN1 to pro-senescence genes including CDKN1A and CDKN2A, and a negative correlation with LMNB1, MCM2, and MCM3, previously shown to be downregulated during senescence and proliferative arrest [27, 28].

On the other hand, we observed a negative correlation between the NMNAT1-2 and proliferation-associated genes such as TP53, CDKN1A, and CDKN2A known to be upregulated during senescence [29, 30] (Figure 1D). Moreover, the NMNAT1-2 correlated positively with LMNB1, MCM2, and MCM3 previously shown to be downregulation during senescence and proliferative arrest [27, 28]. Opposite, to the other two NMNATs, NMNAT3 did not show any significant correlations with adjusted p-value. However, the pattern of NMNAT3 correlations were opposite to that of the other two NMNATs, resembling the patterns of NAMPT and NADSYN1. Combined, these results suggest a regulatory link between NAD⁺ related genes and proliferation.

To validate the results from the RNA sequencing analysis, we evaluated the effect of NR treatment on senescence markers in the WT and WRN^−/− MSCs and in primary fibroblasts derived from healthy control donors and WS patients (Supplementary Table 2) [9]. In line with the RNA seq results, we observed decreased senescence, measured with senescence associated β-Galactosidase (SA-β-Gal) staining, in WRN^−/− MSCs at passage 12 after both 11 and 18 days with 1 mM NR treatment (Figure 1F, 1G and Supplementary Figure 1A–1F). No change was observed in WT MSCs with NR treatment (Figure 1E and Supplementary Figure 1E, 1F). An increase SA-β-Gal staining and a decreased nuclear localization of HMGB1 indicating accelerated senescence was observed in WS patients-derived fibroblasts compared to WT fibroblasts. Interestingly, 10 days treatment with 1 mM NR decreased the senescence markers, as indicated by significantly decreased β-Gal positive area in WS fibroblasts and a significant increased percentage of cells with nuclear HMGB1 localization compared to WS Veh (Figure 1H, 1I and Supplementary Figure 1G–1K). Neither in the primary fibroblasts was a difference observed in WT cells treated with NR compared to vehicle (Figure 1H, 1I and Supplementary Figure 1G–1K). Combined with previous findings [1, 2, 6, 26], these data suggest that WRN is essential in proliferation and senescence inhibition, while NAD⁺ replenishment diminished senescence in MSCs and primary fibroblasts.

NAD⁺ synthesis-related proteins are dysregulated by WRN-KD in HEK293 cells

Via the salvage pathway, the NAD⁺ precursor NR is first converted to NMN which is further catabolized to NAD⁺; the latter step is mediated by the central proteins NMNAT1, NMNAT2 or NMNAT3, depending on subcellular localization [25]. We evaluated the change in the expression level of both NMNAT1 and other proteins involved in NAD⁺ metabolism in HEK293 cells treated with siRNA targeting WRN (WRN-KD). First, we confirmed the knockdown efficiency of WRN to be 99% compared to Scramble siRNA (Scr) with western blotting (Figure 2A, 2B). Next, the expression level of NMNAT1 was found to be significantly lower in WRN-KD (Veh) HEK293 cells compared to Scr (Veh) (Figure 2A, 2C) supporting our previous findings in primary fibroblasts [6]. We also tested the detection of NMNAT2 and NMNAT3, but the commercially available antibodies tested did not show bands on the expected sizes or did not work (not shown). The protein levels of NAMPT and NADSYN1 were significantly increased in WRN-KD (Veh) cells compared to Scr (Veh) (Figure 2A, 2D, 2E). To increase intracellular NAD⁺ [6, 18], we treated the cells with 1 mM NR for 24 h prior to protein extraction. Interestingly, we did not see an effect of NR treatment on the expression of either NMNAT1 or NADSYN1 in Scr or WRN-KD cells. The expression of NAMPT was significantly increased in Scr with NR treatment compared to Scr (Veh) (One-way ANOVA, Tukey’s multiple comparisons test, p-value = 0.0325) (Figure 2A, 2D). There was a tendency towards increased expression of NAMPT in WRN-KD cells with NR compared to WRN-KD (Veh) (One-way ANOVA, Tukey’s multiple comparisons test, p-value = 0.1048) (Figure 2A, 2D).

Based on our findings of decreased protein expression of NMNAT1 in WRN-KD cells, we wondered whether the regulation of WRN and all the NMNATs were connected. Due to the participation of both WRN and the NMNATs in transcriptional regulation [26, 31], we explored ChIP sequencing datasets from the ENCODE Transcription factor target database for potential co-regulation by comparing transcription factors known to bind to specific gene promoters, as we described previously [6]. Our analysis showed that WRN has 110 targets, NMNAT1 has 140 targets, NMNAT2 has 53 targets, and NMNAT3 has 74 targets (lists of all targets in Supplementary File 2). We previously showed that NMNAT1 and WRN share most of their known targets including SIRT6, RAD21, BACH1 and more [6] (Figure 2F). Additionally, more than 40% of the targets are shared between NMNAT2 and WRN (Figure 2G), and more than 50% are shared between NMNAT3 and WRN (Figure 2H). Interestingly, 39 targets are shared between WRN and all three NMNATs (Figure 2I). The shared targets involved chromatin regulators (CHD1, CHD2, RAD21, CTCF, CHD7SMC3, GTF2F1), transcriptional regulators (SIN3A, REST, YY1, ZNF143, BACH1, JUND, TBP), and histone modifiers (KDM4A, KDM5B, EP300, H2AFZ, HDAC2, HDAC6) (Figure 2I, Supplementary Table 3 and Supplementary File 2), suggesting that the co-regulatory axis between WRN and the NMNATs are important for chromatin stability and DNA maintenance. Combined, these results suggest that loss of WRN results in dysregulation of the NAD⁺ machinery, which likely affects transcriptional regulation.

Mitochondrial NAD⁺ is decreased in WRN-KD cells

Due to the severe mitochondria-related dysfunctions seen in WS [6], we examined the effect of loss of WRN on mitochondrial NAD⁺. Mitochondrial NAD⁺ was measured indirectly by using HEK293 cells stably expressing mito-EGFP-PARP1cd-myc, a mitochondrially localized PARP1 protein (termed mitoPARP cells onwards) [19]. The mitoPARP cells overexpress PARP1 fused with a mitochondrial localization signal (MLS), resulting in localization of MLS-PARP1 within the mitochondria [18, 19]. Based on the function of PARP1 to construct chains of poly-ADP-ribose (PAR) by use of NAD⁺, the level of PARylation in the mitoPARP cells correlates with the amount of freely available NAD⁺, as reported and validated previously [19, 20]. Therefore, the amount of PAR signal in the mitoPARP cells can be used as a reporter of mitochondrial NAD⁺ [19, 20]. We silenced WRN with siRNA (WRN-KD) in the mitoPARP cells and treated them with either Veh or 1 mM NR for 24 h in the presence of the PARP inhibitor 3-aminobenzamide (3-AB) to avoid mitochondrial NAD⁺ depletion by the mitoPARP overexpression [19]. To measure NAD⁺ levels, we released the cells from the PARP inhibitor and collected samples at timepoints 0 h, 3 h or 6 h following release (Figure 3A, 3B). We observed a significant increase of PARylation when comparing 0 h to 3 h as expected, with no difference between Scr and WRN-KD cells. Six hours after 3-AB release, we saw a significant increase of PAR-signal in Scr compared to the 3 h timepoint, but no further change in the PAR-signal in WRN-KD cells from 3 h to 6 h (Figure 3A, 3B). Moreover, there was statistically significant less PAR-signal in WRN-KD cells at 6 h post 3-AB compared to Scr. This indicates that WRN-KD resulted in decreased mitochondrial NAD⁺ compared to Scr cells. Due to the role of WRN in DNA repair, total cellular PAR levels, have previously been shown to increase upon WRN loss, which we also observed in the parental HEK293 cells 6 h post 3-AB release (Supplementary Figure 2). We did not observe an increase in PAR-signal in either Scr or WRN-KD cells after 24 h with 1 mM NR treatment (Figure 3A, 3B).

NR treatment or overexpression of SLC25A51 did not restore cell proliferation in WRN depleted immortalized cells

In the center of WS disease etiology is loss of cell proliferation and premature senescence. We have previously shown that treatment with the NAD⁺ precursors NR or NMN could rescue stress-induced intestinal stem cell proliferation in a Drosophila melanogaster model of WS [6]. Combined with our findings on decreased senescence markers post NR treatment in WRN^−/− MSCs and decreased mitochondrial NAD⁺ in WRN-KD cells, we examined the link between mitochondrial NAD⁺ and cell proliferation in WRN-KD cells. To increase mitochondrial NAD⁺ levels, we overexpressed SLC25A51, which has recently been identified as an NAD⁺ transporter localized in the mitochondrial membrane [20]. First, we used primary fibroblasts from WS patients as well as sex- and age-matched controls; however, we were unable to successfully transfect the SLC25A51 plasmid into the WS fibroblasts, possibly due to high vulnerability (not shown).

We then used HEK293 cells to measure proliferation via a classic colony-formation assay (Figure 3C–3E). Cells were seeded on Poly(ethyleneimine) solution coated 6-well plates and left undisturbed for 11 days, whereafter the colonies were fixed, stained, and imaged. As seen in Figure 3E, knock down of WRN was confirmed by western blotting. In the parental HEK293 cells, WRN-KD led to decreased colony formation compared to Scr, as expected (One-way ANOVA, Tukey’s multiple comparisons test, p-value = 0.0189). A 24 h treatment with 1 mM NR prior to colony formation assay did not affect the proliferation capacity of either Scr or WRN-KD cells in the parental HEK293 cells (Figure 3C, 3D). Next, we used two additional cell lines: one cell-line overexpressing the mitochondrial NAD⁺ transporter SLC25A51 (SLC25a51-myc-FLAG, called SLC25A51 from now on) [20] in order to shift the equilibrium in the cells towards increased mitochondrial NAD⁺, and in addition the mitoPARP cells described above. After 6 h of growth without 1 mM 3-AB to deplete the cells from mitochondrial NAD⁺, the mitoPARP cells were seeded for colony formation in the presence of 3-AB to avoid further depletion of mitochondrial NAD⁺. The mitoPARP cells showed an almost complete disruption of colony formation in both Scr and WRN-KD cells with no statistical difference (Figure 3C, 3D). It is worth noting that the continued treatment with 3-AB, will inhibit all cellular PARP activity, which likely resulted in disrupted DNA damage response and hereby cell death.

Interestingly, overexpression of SLC25A51 led to increased colony formation in Scr cells compared to the parental HEK293 Scr cells (One-way ANOVA, Tukey’s multiple comparisons test, p-value = <0.0001). However, no statistically significant change was observed in the parental HEK293 with WRN-KD compared to SLC25A51 cells with WRN-KD. These results indicate that the proliferation defects observed in WRN-KD immortalized cells cannot be rescued by increasing cellular NAD⁺ (with NR treatment) or by increasing the potential import of NAD⁺ into the mitochondria (by overexpressing SLC25A51), however cell proliferation was completely abolished when mitochondrial NAD⁺ was depleted.

SLC25A51 overexpression combined with NR treatment does not rescue proliferation in WRN-KD cells

To further understand the effect of mitochondrial NAD⁺ on proliferation, we treated parental HEK293 cells and SLC25A51-overexpressing HEK293 cells with 1 mM NR for 11 days during colony formation. The cell culture medium was exchanged every 3 days with fresh medium with or without 1 mM NR (Figure 4A, 4B). Silencing WRN with siRNA led to decreased colony formation capacity in both the parental and SLC25A51 HEK293 cells (One-way ANOVA, Tukey’s multiple comparisons test, p-value = 0.0009 and p-value < 0.0001, respectively). Colony formation was increased with SLC25A51 overexpression without WRN silencing, but not in cells with WRN silencing. Eleven days of 1 mM NR treatment did not affect colony formation in either Scr or WRN-KD cells. This suggests that mitochondrial NAD⁺ is not sufficient to rescue proliferation in WRN depleted HEK293 cells.

We further evaluated the effect of both NR treatment and SLC25A51 overexpression on the protein levels of NAD⁺ metabolism-related proteins by treating parental HEK293 (HEK293P) cells (same blots as in Figure 2) or SLC25A51 overexpressing cells with 1 mM NR treatment for 24 h. The results showed that neither NR nor SLC25A51 overexpression affected the expression of NMNAT1 when compared to HEK293 Parental in either Scr or WRN-KD cells (Figure 4C–4E). NADSYN1 expression was increased in Scr cells with SLC25A51 overexpression compared to parental HEK293 cells, but not in WRN-KD cells (Figure 4C, 4F; One-way ANOVA, Tukey’s multiple comparisons test, p-value = 0.0048)), and NAMPT was increased in both Scr and WRN-KD cells with SLC25A51 overexpression (Figure 4C, 4G, One-way ANOVA, Tukey’s multiple comparisons test, p-value < 0.0001 and p-value = 0.0004, respectively). No significant effects were found with 24 h 1 mM NR treatment in any of the cell lines tested.

Mesenchymal stem cells

The mesenchymal stem cells (MSCs) were generated in the laboratory of H.H. Cheung as previously published [9]. Control MSCs (WT; GM440, BM-MSCs) and WS MSCs (WS797; WS780) were seeded on 24-well plates. The day after seeding, NR treatment (0.5 mM or 1 mM) was initiated, and media was exchanged with fresh media without or with NR every third day. After 11 days (WS797), 18 days (WS780) or 24 days (GM440; BM-MSCs) cells were stained with β-Gal, and then imaged on a bright field microscope. The number of β-Gal positive cells compared to the total number of cells was quantified using ImageJ on images from two biological repeats per cell line.

Primary fibroblasts

The primary skin fibroblasts (NF1, NF2, WFL8, WF9A) were obtained from two healthy individuals and two WS patients. Written informed consent was obtained from the individuals who provided the samples to conduct and publish this study, which was approved by the Institutional Ethics Review Board of the Chiba University Graduate School of Medicine (IRB-1145[1029]). The primary fibroblasts (AG10803A; AG16146; AG02602; AG12975; AG06300) were obtained from Coriell Institute (USA). Two WS patient derived cell lines (AG12975; AG06300) and three control lines from healthy sex- and age-matched controls (AG10803A; AG16146; AG02602) were used. All experiments with primary fibroblasts were performed in passage 14–16, and the same passages were used for both controls and WS lines. The cells were grown in DMEM/MEM (50:50) with 15% fetal bovine serum (FBS) and 1% penicillin-streptomycin (PS). For β-Gal evaluation, cells were grown without or with 1 mM NR in the medium for 10 days in total, with the media exchanged with fresh media every other day. SA-β-gal staining was performed using the Senescence β-Galactosidase Activity Assay Kit (Fluorescence, Plate-Based; #23833, Cell Signaling Technology) according to the manufacturer’s protocol. Cells cultured in 12-well plates were fixed with 1× fixative solution and stained overnight at 37°C. The cells were washed with PBS (−), stained with Hoechst 33342 (H342; Dojindo, Kumamoto, Japan) according to the manufacturer’s protocol, and photographed using a BZ-X700 microscope (Keyence, Osaka, Japan). For SA-β-Gal detection with Spider β-Gal stain (Dojindo), cells were seeded on poly-D-lysine coated coverslips. The next day the media was exchanged to media without or with 1 mM NR. The following day, Spider β-Gal stain (Dojindo) was prepared in DMSO according to the manufacturer’s protocol, then diluted in Mcllvaine buffer (pH 6.0) (0.1 mol/L citric acid solution, 0.2 mol/L sodium phosphate solution, adjusted to pH 6.0, and diluted 5 times in ultrapure water). The cells were fixed in 4% paraformaldehyde (PFA), stained for 30 min, washed in PBS, and mounted with ProLong DAPI mounting media (Invitrogen, USA). Cells were imaged with Zeiss LSM780 confocal microscope (DAPI (Ex. 405 nm), Spider β-Gal (Ex. 488 nm)) and the intensity of the Spider β-Gal signal was analyzed with ImageJ (four biological repeats, with two controls and two WS cell lines).

To evaluate senescence, the cellular localization of HMGB1, was used; decreased nuclear HMGB1, is a marker of increased senescence. Primary fibroblasts from either controls or WS patients were grown for 7 days in cell culture media without or with 1 mM NR with the media exchanged with fresh media every other day. Then, WT or WS-patient derived fibroblasts were seeded on PDL coated coverslips in normal media. The next day, media was changed to Veh or 1 mM NR. After additional 24 h, cells were washed in PBS, fixed in 4% PFA and permeabilized, then stained with anti-HMGB1 antibody (1:400, Abcam, #18256) and secondary anti-rabbit (1:500; Alexa-flour 546, Invitrogen). The cells were washed in PBS and mounted with Prolong DAPI mounting media (Invitrogen, USA). Cells were imaged with Zeiss LSM780 confocal microscope and cells expressing HMGB1 only in the nucleus compared to in the entire cell were analyzed with ImageJ (three biological repeats with three controls and two WS cell lines).

Cell lines, cell culture, and RNAi knockdown

The HEK293, stably transfected HEK293 cell lines (mito-EGFP-PARP1cd-myc (mitoPARP), and SLC25a51-myc-FLAG (SLC25A51)) were generated in the M. Ziegler laboratory as described previously [19, 20]. Cells were grown in DMEM with 10% FBS, 1% PS and 550 µg/ml Geneticin-418. For mitoPARP cells, 1 mM 3-aminobenzamide (3-AB) was included in the medium unless otherwise noted.

To induce knockdown of WRN, cells were treated with 30 nM WRN siRNA oligo duplex for 6 h (#SR322215, OriGene) in OPTI-MEM media, followed by addition of normal growth media. After 48 h cells were collected in (Cell Signaling, #9806S) containing 1x Halts protease and phosphatase inhibitor cocktail (Thermo Fisher, USA) and the knockdown efficiency was confirmed with western blotting as described below.

Poly-ADP-ribose assisted detection of mitochondrial NAD⁺

MitoPARP cells were seeded on poly-D-lysine (PDL) coated plates in normal growth media minus antibiotics. After 16 h, the media was replaced with OPTI-MEM media and cells were transfected with either scramble siRNA (Scr) or 30 nM WRN siRNA (OriGene, USA). After 24 h, the medium was replaced with normal growth medium +/− 1 mM NR for 24 h. In all instances the mitoPARP cell line was grown with 1 mM 3-AB in the media to avoid mitochondrial NAD⁺ depletion. To evaluate the level of mitoPARP in Scr vs. WRN-KD cells, the medium was replaced with medium without 3-AB, and cells were harvested at different timepoints as indicated in the figure (0 h, 3 h or 6 h after release from 3-AB). To harvest the cells, they were first washed once in 1 mL PBS, then lysed (20 mM TrisHCl (pH 7.4), 150 mM NaCl, 1% (v/v) SDS, 1 mM EDTA, 1 mM 3-AB) and collected. After passaging the lysate 10 times through a 23-gauge needle to shear the genomic DNA, protein concentration was determined using Pierce BCA protein assay kit (Thermo Fisher, #23225). Fifteen µg of lysate was separated by SDS-PAGE in a 4–12% bis-tris gel and subjected to immunoblotting using overnight incubation at 4^οC with primary antibodies (anti-PAR, Novus #4355-MC-100, anti-β-Actin, Sigma, #SAB1305546) followed incubation with HRP-conjugated goat anti-mouse secondary antibody for 1 h at room temperature. HRP-detection was performed using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher, #34096) and ChemiDoc XRS+ imaging system (Bio-Rad, USA). The PAR signal was quantified using ImageJ.

Colony formation assay

To evaluate proliferation capacity of WRN-KD cells when changing the mitochondrial NAD⁺ availability, we performed a colony formation assay. Two-hundred fifty cells were seeded in 6-well plates including 3-AB for mitoPARP cells. The next day, cells were transfected with scramble or WRN siRNA as described above. The following day, the medium was changed to normal DMEM +/− 1 mM NR. After 24 h, cells were trypsinized, counted, and re-seeded in coated 6-well plates for colony formation. 750 cells were seeded per well. In the experiment with continued NR treatment, the media was replaced with fresh media every three days.

On day 11, wells were washed in 1 mL PBS, fixed with ethanol and stained with 0.3% crystal violet (Sigma-Aldrich) in 70% ethanol for 30 min at room temperature following additional washes before drying. Colonies were imaged on ChemiDoc XRS+ imaging system and counted using ImageJ.

Western blotting

Western blotting was used to examine protein expression following the protocol we reported elsewhere [15, 42]. HEK293 and SLC25A51 cells were collected, and extracts prepared using 50 µl 1× RIPA buffer (Cell Signaling, #9806S) containing 1× Halts protease and phosphatase inhibitor cocktail (Thermo Fisher, USA). Proteins were separated on a 26-well 4–12% Bis-Tris gel (Invitrogen, USA) and run for 50 min at 200 V in the NuPage Midiblot Module (Invitrogen, USA). After SDS-PAGE, gels were transferred to the PuPage Midiblot transfer Module (Invitrogen, USA) at 25 V for 30 min, followed by blocking for 1 h at RT in 5% non-fat dried milk-TBST. Membranes were incubated with primary antibodies in milk-TBST ON at 4°C, and secondary antibodies in milk-TBST for 1 h at RT before detection on ChemiDoc XRS+ imaging system (Bio-Rad, USA). Quantification was performed using ImageJ. The original western blots can be found in the Supplementary Materials. Antibodies used were: β-Actin (Sigma, #SAB1305546), WRN (Santa Cruz, # sc-5629), NMNAT1 (Cell Signaling, #98354S), NAMPT (Abcam, #ab236874), NADSYN1 (Abcam, #ab171942), GFP (Roche #11814460001), FLAG (Sigma-Aldrich, #F1804-200UG), β-Tubulin (Abcam, #6046).

Bioinformatic analysis

Statistical analysis

Statistical analysis used in this study was Student’s two-tailed unpaired t-test for comparison between two groups, or ANOVA with multiple comparisons for comparisons among multiple groups. All data were presented as mean ± S.D. or mean ± S.E.M. as indicated, where a p-value < 0.05 was considered statistically significant.

Data availability statement

The RNA sequencing data were generated in [5], and is available from Gene Expression Omnibus with the accession number: GSE248012 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE248012). The data that support the findings of this study, including data in supplementary material of this article, are available from the corresponding author upon reasonable request.