Modeling paraquat-induced lung fibrosis in C. elegans reveals KRIT1 as a key regulator of collagen gene transcription

Paraquat increases collagen gene transcription in a KRI-1 dependent manner

To understand the genetic regulation of collagen accumulation in paraquat-induced lung fibrosis, we took advantage of the genetically tractable C. elegans model. By mining RNAseq data (GSE123531) from a previous study [31], we found that collagen genes were generally up-regulated by paraquat (Figure 1A). As shown in the volcano plot, among 181 genes encoding or predicted to encode collagens, 151 were picked up by the RNAseq (red color) and 48 have significant changes. Among 48 collagen genes, 46 were increased and only 2 were decreased by paraquat. Despite the general decrease in global transcription after paraquat treatment, the 48 collagen genes were 1.71 log2 fold changed, or 3.27 fold increased in expression (Figure 1B). COL-43, COL-80, and COL-139 were the top 3 genes affected by paraquat in this study. By examining cluster and protein structure on C. elegans collagens database (CeColDB, http://CeColDB.permalink.cc/.), we find that COL-43, COL-80 and COL-139 belong to different collagen subclasses, e.g. C23a, D07, and B16c, respectively. All contain several Col1-domains and 1 N-Propeptide. COL-139 and COL-43 each contain 1 transmembrane helix while COL-80 contains a Signal Peptide. Since the Col1-domain and N-Propeptide are widely present in many collagens, it seems that paraquat does not preferentially upregulate specific types of collagens.

Next, we asked if KRI-1 was required for collagen gene transcription. KRI-1 was recently shown to be a key mediator of paraquat toxicity in C. elegans [32] but its role in collagen transcription has not been reported. By using kri-1(ok1251) null mutant animals, we measured the mRNA levels of top-ranked genes from Figure 1A by real-time quantitative PCR (RT-qPCR). We raised WT and kri-1 mutant worms in 75 μM of paraquat (methyl viologen dichloride hydrate) as determined in [31] from L1 stage to day-1 of adulthood and extracted mRNA from whole worm. The RT-qPCR results showed that lacking KRI-1 largely prevented paraquat from increasing the transcription of col-43, col-80, and col-139 (Figure 1C). In addition, by using a GFP-tagged COL-19 fusion protein, we confirmed that COL-19 expression was significantly increased by paraquat and such increase required KRI-1, as kri-1 null mutant expressing COL-19::GFP was no longer responsive to paraquat treatment (Figure 1D, 1E). As kri-1 could develop slower than control in the presence of paraquat, we also examined COL-19::GFP at day-4 of adulthood. As shown in Supplementary Figure 1, kri-1 mutants remained much less responsive to PQ than the WT control. These studies identify a novel role of KRI-1 in regulating collagen gene transcription.

SKN-1 mediates KRI-1 regulation of collagen transcription in response to paraquat treatment

The oxidative stress responsive transcription factor SKN-1 positively regulates collagen transcription to delay aging in C. elegans [33]. We wondered if KRI-1 would regulate SKN-1 to promote collagen gene transcription in response to paraquat treatment. First, we knocked down SKN-1 transcription by RNAi and asked if such treatment would block paraquat from inducing collagen transcription. Worms were fed HT115 bacteria expressing double-stranded RNA targeting skn-1 gene from L1 stage to day-1 of adulthood in the presence or absence of 75 μM paraquat. RT-qPCR demonstrated that the transcription of col-43, col-80 and col-139 were all reduced by skn-1 knockdown (Figure 2A). skn-1 knockdown in kri-1 mutant did not further decrease the expression of above collagen genes (Figure 2B), suggesting that KRI-1 and SKN-1 function in the same pathway to regulate collagen transcription. By using Western blot to detect GFP levels, we further confirmed the genetic interaction between KRI-1 and SKN-1 on collagen gene transcription; paraquat-induced COL-19:GFP up-regulation was blunted by SKN-1 knockdown and largely blocked by kri-1 knockout, but was not additively affected by both (Figure 2C, 2D).

We next asked if SKN-1 hyperactivation could be epistatic to kri-1 mutant. WDR-23 is homologous to human KEAP1, a negative regulator of Nrf2; knocking down wdr-23 expression constitutively activates SKN-1 in C. elegans [34]. We knocked down wdr-23 from L1 in the presence and absence of 75 μM paraquat. Indeed, wdr-23 knockdown significantly increased the transcription of col-43 and col-80 in both WT and kri-1 mutant worms (Figure 3A, 3C). The normalized fold changes were higher in kri-1 mutant (Figure 3B, 3D) as compared to N2 WT, suggesting that Nrf2 are epistatic to KRI-1 in collagen transcription. Further supporting this result, western blot analysis showed that wdr-23 knockdown increased COL-19::GFP levels regardless of paraquat treatment in kri-1 mutant (Figure 3E, 3F), suggesting that KRI-1 regulates collagen transcription at least partly through SKN-1.

Collagen up-regulation by SKN-1 is required for protection from oxidative stress

Since collagen has been shown to improve health and extend lifespan in C. elegans [33], the transcriptional induction of collagen gene upon paraquat exposure could function as a protective mechanism against oxidative stress from paraquat. To test this, we first examined if loss of collagen could render worms sensitive to paraquat. Although there are 181 collagen genes in C. elegans, knocking down a single collagen gene has been shown to compromise the expression of other collagen genes [33]. We knocked down col-43 and col-80 individually and raised worms in the presence or absence of 75 μM paraquat. At day-1 of adulthood, we tested their tolerance to oxidative stress by challenging worms with 200 mM paraquat for 8 hours [35]. As shown in toxicity assay in Figure 4A, worms preconditioned with 75 μM paraquat tolerated much better than non-treated controls. However, knocking down either col-43 or col-80 significantly compromised the acquired tolerance to paraquat toxicity (Figure 4A).

Next, we tested if collagen overexpression would rescue the sensitive phenotype of kri-1 and skn-1 mutants to paraquat. Interestingly, although there are many collagen genes, overexpressing single collagen gene has been shown to promote longevity in C. elegans [33]. By using transgenic worms expressing high copy number of extrachromosomal collagen genes (Supplementary Figure 2), we showed that COL-43 and COL-80 overexpression significantly increased the resistance to paraquat toxicity in all WT kri-1 mutant and skn-1(RNAi) animals (Figure 4B, 4C, 4E, 4F). Consistent with a previous study [32], kri-1 mutant and skn-1(RNAi) animals were sensitive to paraquat, but such sensitivity was partly rescued by COL-43 and COL-80 overexpression. By measuring the area under the curve (AUC) and calculated the fold increase, we showed that both collagen overexpression preferentially increased paraquat tolerance in kri-1 and skn-1(RNAi) animals as compared to WT (Figure 4D, 4G), further confirming the specific roles of KRI-1 and SKN-1 in the regulation of collagen transcription.

We were also interested in knowing if promoting resistance to paraquat could result in extended lifespan. Overexpressing several collagen genes such as COL-10, COL-13 and COL-120 have been shown to extend lifespan [33]. Consistently, we found that COL-43 and COL-80 overexpression also extended lifespan of WT animals (Supplementary Figure 3B).

Conserved roles of Krit1 and Nrf2 in collagen transcription in human lung fibroblasts

We were interested to know if our findings in C. elegans would be conserved in human cells. In C. elegans, we observed that even without paraquat treatment, kri-1 mutation had already reduced collagen gene expression (Figure 1D, 2C), suggesting that kri-1 is also required for the basal collagen expression under normal condition. This prompted us to investigate into the public GEO dataset deposited at National Center for Biotechnology Information (NCBI). The GSE85657 contains RNAseq data of KRIT1 (human homolog of KRI-1) knockdown in mouse primary brain microvascular endothelial cells [36]. We took advantage of this resource and did a Volcano Plot of 11653 transcripts including 26 collagen genes. We found that in the KRIT1-depleted cells, 22 out of 26 collagen genes were decreased as compared to those in WT control (Figure 5A, red). The mean expression of all the 26 collagen genes was -0.71 in log2 fold change, which equals to 0.61 fold change as compared to those in WT control. These results support our findings in C. elegans, suggesting that KRIT1 could function in a conserved manner in human cells.

In human lung fibroblast cell line MRC-5, paraquat has been shown to increase collagen transcription [21]. We asked if KRIT1 and Nrf2 (human homolog of SKN-1) would be required for paraquat to increase collagen gene transcription. First, we knocked down KRIT1 and NRF2 individually, then treated MRC-5 cells with 300 μM paraquat for 48 hours as determined by the previous paper [21]. We extracted mRNA for RT-qPCR analysis of 2 top-ranked collagen genes COL28A1 and COL27A1 in Figure 5A (Arrows). COL28A1 and COL27A1 belong to their distinct collagen types, XXVIII and XXVII, respectively. COL28A1 belongs to a class of collagens containing von Willebrand factor type A (VWFA) domains and could be involved in collagen chain trimerization and degradation of the extracellular matrix [37]. COL27A1 is related to the "fibrillar" class of collagens and may play a role in development of the skeleton [38, 39]. Interestingly, they have been predicted to be functional partners based on STRING, the protein-protein interaction network. Consistent with our study in C. elegans, COL28A1 and COL27A1 expression were elevated after paraquat treatment in MRC-5 cells (Figure 5B, 5C). Interestingly, KRIT1 knockdown reduced basal expression of both COL28A1 and COL27A1, consistent with the RNAseq data shown in Figure 1A. Although paraquat still increased collagen gene expression in KRIT1- and NRF2-knockdown cells, the fold change was significantly reduced (Figure 5D), suggesting specific roles of KRIT1 and Nrf2 in paraquat toxicity.

We further tested if Nrf2 functions downstream of KRIT1 for collagen transcription. Similar to the study in C. elegans, we activated Nrf2 by knocking down its negative regulator KEAP1 and examined COL27A1 and COL28A1 mRNA expression. Consistently, KEAP1 knockdown upregulated COL27A1 mRNA levels in both WT and KRIT1-knockdown MRC-5 cells, rescuing the low expression in KRIT1-knockdown cells to WT levels (Figure 5E). A similar pattern was observed for COL28A1 except that the low expression in KRIT1-knockdown cells was not fully rescued (Figure 5F). However, after comparing the fold change, we found that KEAP1 siRNA knockdown had stronger effect on COL28A1 expression in KRIT1-knockdown cells than control cells (Figure 5G), suggesting KRIT1 regulates collagen at least partly through Nrf2 pathway. Interestingly, for unknown reasons, KEAP1 knockdown in paraquat-treated cells reduced the expression of both collagen genes (Figure 5E, 5F).

C. elegans maintenance and RNAi knockdown

Animals were maintained at 25° C on nematode growth medium (NGM) agar plates seeded with OP-50 bacteria. NGM agar plates were prepared as described before [47]. Information regarding N2 Bristol wild-type, kri-1(ok1251) mutant and strains expressing COL-19::GFP, COL-43 and COL-80 were detailed in the Supplemental Information section. RNAi clones were originally from RNAi library constructed by the Ahringer group [48]. RNAi knockdown was achieved by feeding worms HT115 bacteria expressing double-stranded RNA (dsRNA) of target genes on NGM agar plates. Specifically, RNAi bacteria were cultured to log phase in LB liquid medium containing 50 ug/mL carbenicillin then seeded on NGM agar plate containing 50 ug/mL carbenicillin and 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) for 2 days. L1 stage worms were transferred to and maintained on the RNAi plate until day-1 of adulthood.

MRC-5 cell culture and siRNA knockdown

MRC-5 cells were originally obtained from ATCC and maintained in Dulbecco’s minimal essential medium (DMEM) supplemented with 10% heat-inactivated fetal calf serum (FCS) at 37° C in 5% CO2 humidified incubator. Fibroblasts between passages 25 and 35 were used for all experiments. For siRNA knockdown, cells at 60% confluence were transiently transfected with siRNA by using lipofectamine 2000 (ThermoFisher Scientific). After 24 hours, cells were treated with paraquat at the final concentration of 300 μM for 48 hours. The siRNA information for KRIT1, KEAP1 and NRF2 is shown in Supplementary Table 1.

Real time quantitative PCR (RT-qPCR)

C. elegans day-1 adult animals were homogenized by sonication and total RNA was extracted by TRIZOL reagent. Total RNA from MRC-5 cells was obtained by purification from cell monolayers using the PureLink RNA Mini Kit. Total RNA was treated with DNase I to remove DNA contaminants, then reverse transcribed to cDNA with High-Capacity cDNA reverse transcription kit (Invitrogen) according to the manufacturer's instruction. RT-qPCR was conducted in SYBR Green PCR Master Mix (Applied Biosystems) in triplicate using the ABI Prism 7300 Sequence Detection System. Actin gene (act-1) was used as internal control for C. elegans and GAPDH for human cells. Primers sets were designed by ProbeFinder software (version 2.45) of the Universal Probe Library from Roche. The primer sets are listed in Supplementary Table 2.

Western blot

C. elegans day-1 adult animals were homogenized by sonication in RIPA Lysis and Extraction Buffer supplemented with Halt™ Protease Inhibitor Cocktail (ThermoFisher Scientific). The protein concentration was measured by Pierce™ BCA Protein Assay Kit and 100 μg of the total proteins for each sample was denatured by heating in SDS-loading buffer at 90° C for 5 min. 25ug of the total proteins were separated on SDS-PAGE and transferred to PVDF membrane. Western blot was carried out by blocking the membrane in 5% non-fat milk in PBS for 30 min, incubating with mouse anti-GFP or mouse anti-β-actin primary antibody (Promab Biotechnologies #20144, #20270) for 1 hour, washing extensively, and then finally incubating with an anti-mouse, HRP-conjugated secondary antibody for 30 min. HRP signal were detected with LumiGLO® reagent autoradiographed with X-ray film.

Paraquat toxicity assay

Paraquat toxicity assay in C. elegans is described in detail in [35]. Briefly, synchronized worms were raised to young adult stage and transferred to NGM agar plate containing 200 mM paraquat. 5 plates containing roughly 40 worms/plate were assayed each time and experiments were repeated 3 times. Live and dead worms were recorded every 3 hours until 12 hours. Death is defined by worms that were no longer moving when touched by a platinum tip. Percentage Survival were calculated and plotted with GraphPad Prism software.

Data analysis, visualization and statistics

RNAseq Datasets of GSE123531 and GSE85657 were downloaded from NCBI website and analyzed by using Rstudio software loaded with ggplot2 package. Volcano Plots were drawn in Rstudio using total data points and data points for collagen genes (red). COL-19::GFP signal and western blot signal were quantified in ImageJ software. Bar graph were generated using GraphPad Prism software. Statistical analysis methods were indicated in individual figure legends, which were also generated using GraphPad Prism software.