Comparative analysis of long noncoding RNAs in long-lived mammals provides insights into natural cancer-resistance

ABSTRACT

Mouse and rats are staple model organisms that have been traditionally used for oncological studies; however, their short lifespan and highly prone to cancers limit their utilizationsin understanding the mechanisms of cancer resistance. In recent years, several studies of the non-standard long-lived mammalian species like naked mole rat (NMR) have provided new insights of mechanisms in natural anti-cancer. How long-lived species genetically maintain longevity and cancer-resistance remains largely elusive. To better understand the underlying anti-cancer mechanisms in long-lived mammals, we genome widely identified long noncoding RNA (lncRNA) transcripts of two longevous mammals, bowhead whale (BW, Balaena mysticetus) and Brandt’s bat (BB, Myotis brandtii) and featured their sequence traits, expression patterns, and their correlations with cancer-resistance. Similar with naked mole rat (NMR, Heterocephalus glaber), the most long-lived rodent, BW and BB lncRNAs show low sequence conservation and dynamic expressions among tissues and physiological stages. By utilizing k-mers clustering, 75–136 of BW, BB and NMR lncRNAs were found in close relation (Pearson’s r ≥0.9, p < 0.01) with human ageing diseases related lncRNAs (HAR-Lncs). In addition, we observed thousands of BB and BW lncRNAs strongly co-expressed (r > 0.8 or r <-0.8, p < 0.01) with potential tumour suppressors, indicating that lncRNAs are potentially involved in anti-cancer regulation in long-lived mammals. Our study provides the basis for lncRNA researches in perspectives of evolution and anti-cancer studies.

Abbreviations: BW: bowhead whale; BB: Brandt’s bat; NMR: naked mole rat; LLM: long-lived mammal; HTS: human tumour-suppressors; PTS: potential tumour suppressor; ARD: ageing related diseases; HAR-Lncs: lncRNAs that related with human ageing diseases; Kmer-lncs: lncRNAs in long-lived mammal species that corelated (Pearson’sr ≥0.9, p < 0.01) with the 10 HAR-Lncs by k-mers clustering; All-lncs: all the lncRNAs in long-lived mammal species; SDE-lncs: significant differentially expressed lncRNAs

Introduction

Cancer is the most common cause of human death worldwide.In the past decades, mouse and rat as the traditional model organisms have been widely used in cancer studies. The long-term goal of cancer research is to develop a therapeutic or preventive strategy with both efficiency and non-toxicity [1,2]. However, more than 50% of aged mice and rats in previous studies die of cancer while this number in humans is approximately 23% [3], suggesting staple mammal models are limited for understanding and exploring the spontaneous occurring anti-cancer strategies. The cancer rates and ages of disease onset inmammal species have diverged dramatically during the lengthy process of evolution [1].There are several species include the naked mole rat, blind mole rat, Brandt’s bat, elephant and bowhead whales are known for their extremely longevity and cancer resistance [4]. It takes averagely two years for a mouse to develop cancer, whereas it takes decades in long- lived species [1,5–7], suggesting that longevous species are good animal models for studying the natural anti-cancer mechanisms.

Researchers have found multiple mechanisms contribute to the remarkable cancer resistance of naked mole rat (Heterocephalus glaber), the longest-live rodent. The cells of the naked mole rat secreted unique high molecular mass hyaluronan (HMM-HA) resulting in the activation of early contact inhibition (ECI) while cancerous cells lose ECI and continue to proliferate on top of each other [8]. Inactivation of either one of the tumour suppressors like RB1 or TP53 causes cell apoptosis in naked mole rat; however, it is in strikingly contrast to mouse or human cells, where inactivationof either RB1 or TP53 leads to more rapid proliferation [9].Bats live longer than any other order of mammal relative to their body mass [10]. Therefore, it is surprising that only a few incidences of tumours have been reported in bats [1]. In particular, the Brandt’s bat (Myotis brandtii) holds the longevity record with a small size of approximate 7 grams among the bats [11]. Recent studies suggested that an up-regulation of tumour suppressor miRNAs and the distinctive mitochondrial function in bats have played critical roles in their cancer-resistance [12–15]. On the other hand, the genetic variations ofthe highly conserved transmembrane domains in the growth hormone receptor (GHR) and insulin-like growth factor 1 receptor (IGF1 R)of long-lived bats are associated factors that contribute to the cancer resistance [13]. In addition to small sized long-lived mammals, bowhead whale (Balaena Mysticetus), the most longevous mammal species with large body size has evolved extraordinary tumor suppressor mechanisms as well. Previously comparative genomic and transcriptomic studies found the specific changes of gene expression include insulin pathway genes and some positive selected genes like ERCC1, PCNA and UCP1 in bowhead whale are linked to cancer and ageing [16,17].

Accumulated evidence has underlined the protein coding genes (PCG)-based genetic mechanisms of cancer-resistance in long-live mammals; however, the roles of epigenetic factors like long noncoding RNA (lncRNA) are remain largely unknown. The advance and accuracy of sequencing technology in recent years made lncRNAs the potential diagnose and prognostic markers in human cancers. Large number of lncRNAs were dysregulated in various cancer types, and many of them have been identified as tumour drivers such as LINK-A [18,19] or potential suppressors like Gas5 [20]. In human triple-negative breast cancer (TNBC), an upregulation of LINK-A was particularly observed comparing with non-TNBC breast cancer tissues and it is predictive of poor prognosis for patients with breast cancer [18]. In contrast to LINK-A, the expression of tumour suppressor, GAS5, is typically reduced in multiple cancers [21]. More crucially, the low GAS5 lncRNA levels are correlated with poor outcomes [22,23]. Recent studies revealed that overexpression of GAS5 can significantly suppress the growth of tumour cells in many cancer types [24–26], demonstrating that lncRNAs are promising therapeutic strategies for cancer treatment. For a better knowledge of lncRNAs in anti-cancer mechanisms, we identified the high-quality lncRNAs in two non-standard long-lived mammals organisms, Brandt’s bat and bowhead whales, at the genome wide scale; and performed comparative analysis with the published naked mole rat lncRNAs, aiming at mining the expressing features of long-lived mammals lncRNAs as well as their potential associations with natural cancer-resistance.

Results

Genome-wide identification of lncRNAs in long-lived mammals (LLM)

We genome-widely identified lncRNAs intwo long-lived species include bowhead whale (BW) and Brandt’s bat (BB). RNA-seq profiles containing 6 BW transcriptomes and 15 BB transcriptomes were used for lncRNAs identification (Table S1).

Following a stringent lncRNA identification pipeline, we separately annotated 8,004 and 6,824 high-quality lncRNAs transcripts from BW and BB genomes. To better understand the associations among lncRNAs, longevity and cancer-resistance, 4,422 lncRNA sequences from the most longevous rodent – naked mole rat (NMR) were acquired from our previous study for comparative analysis [27]. In comparison with protein coding gene (PCG), lncRNA in all three longevous species exhibit significant differences, with lower average lengths but higher GCcontents (Fig. 1). Average length of LLM lncRNAs is significant shorter than PCG (p < 0.01), especially in BB and NMR lncRNAs. Comparing with the other two longevous species, NMR possesses the shortest lncRNAs at average.

Figure 1.

Features of LLM lncRNAs comparing with protein coding genes (PCGs). (A) Average Length. (B) Average GC contents.

Conservation analysis among LLM lncRNAs

As suggested by previous phylogenic studies, the closest relative to the BW is minke whale (Balaenoptera acutorstrata), their divergence time is estimated by 25–30 million years (MYA) ago [17,28]. The life expectation of BB is 200 years; however, it is only 50 years for minke whale (Fig. 2). NMR has alike look and body size with mouse except for its barely nude skin, it is estimated that NMR divergent from rodent around 73 (60.2–80.3) MYA [29], and their life spans differ to 10 folds. To explore the conserved lncRNAs among these three longevous species, orthologous analysis was performed using OrthoMCL with BLASTN programme[30]. Only 2 pairs of BB and BW lncRNAs show homology (≥ 50% hits and E-value<1e⁻⁵), suggesting that LLM lncRNAs lack sequence conservation. The homologous lncRNA information are listed: (1) BW_00107607|location: scaffold_3048:284–646 vsBB_00066642|location: scaffold4:7,536,925–7,537,412; (2) BW_00071170|location: scaffold_21:1,920,428–1,923,505 vs BB_00072718|location: scaffold454:1,106,022–110,927.

Figure 2.

Phylogenetic tree of long-lived mammals and their relative species.

By analysing their FPKM values, we found that the first pair of homologs are completely non-detectable in all tissues except brain (cerebrum) in both BB and BW; however, the other pair of lncRNAs show less tissue-specific but express significant higher in brains as well. The functions of those homologous lncRNAs are currently unknown, but their high or unique expression in brains suggested that they may involve in neuronal related functionality.

Spatiality- and tissues-specific expression of lncRNAs in BB and BW genomes

The 6 BW transcriptomes were isolated and prepared from 6 tissues consisting of liver, cerebellum, heart, lung, kidney and muscle from a male bowhead whale [17]. A total of 15 RNA-seq profiles of BBwere obtained from 3 tissues (liver, kidney and brain) in bats at three behavioural periods include hibernating for 2 months (H2), hibernating for 6 months (H6) and summer active (A) [11].

Behavioural studies indicate someecological and physiological features in bats like hibernationcan increase their survival rates, thus linking with the longevity [31,32]. During hibernating, hundreds of lncRNAs in BB expressed differentially among tissues especially in livers (Fig. 3B–D). Comparing with summer active bats, there are 843 and 441 lncRNAs dysregulated in liver tissues of H2 and H6 bats, respectively; of which 229 are overlapped (Fig. 3C). In addition, 62 BB lncRNAs expressed significantly different acrossthree periods in livers (Fig. 3C); and a total of 26 dysregulated lncRNAs were commonly found in three kidneys (Fig. 3D).Similar with BB lncRNAs, 661 BW lncRNAs distribute specific expression manners across 6 tissues and the top-ranked 50 lncRNAs are shown in Fig. 3A.The expression of BB and BW lncRNAs displayed spatiality- and tissues- specificity, which is in common with NMR and non-longevous species [27].

Figure 3.

Significant differentially expressed (SDE) -lncRNAs in BB and BW genomes among tissues and physiological stages (‘A’ means summer active bats; ‘H2ʹ and ‘H6ʹ separately meansbats hibernate for 2 and 6 months). (A) Heatmap of the top 50 BWSDE-lncRNAs across 6 tissues (B) Heatmap of the top 50 BB SDE- lncRNAsacross 3 tissues at 3 physiological stages (‘A’, ‘H2ʹ and ‘H6ʹ) (C) Distribution of BB SDE-lncRNAsin livers (upper panel) and kidneys (lower panel) across 3 physiological stages(H2_vs_A: comparing ‘H2ʹ with ‘A’; H2_vs_H6: comparing ‘H2ʹ with ‘H6ʹ; H6_vs_A: comparing ‘H6ʹ with ‘A’) by Venny diagram: in livers, 62 SDE-lncRNAs are overlapped among three pairs, 295 specific SDE-lncRNAs in H2_vs_A, 151 specific SDE-lncRNAs in H2_vs_H6, and 115 SDE-lncRNAs in H6_ vs_A; in kidneys, 26 SDE-lncRNAs are overlapped among three pairs, 164 specific SDE-lncRNAs in H2_vs_A, 148 specific SDE-lncRNAs in H2_vs_H6, and 240 SDE-lncRNAs in H6_ vs_A.

Co-expression analysis among BB lncRNAs, BW lncRNAs and tumour suppressors

To investigate the associations between LLM lncRNAs and cancer resistance, human tumour suppressing genes were utilized for co-expression analysis. We organized a gene list that contains 1,217 experimentally verified human tumour-suppressors (HTSs) from TSGene2 [33], and then searched their orthologs in BB and BW from the annotated genomic files. In BW genome, 940 PCGs were identified as homologs of HTSs, and it was 1,015 in BB species. These homologs were considered as potential tumour suppressors in BB (PTSs-BB) and BW (PTSs-BW). Co-expression analysis of LLM lncRNAs and PTSs-LLM was performed by in-house R script.

The expression of lncRNA is usually at low level, which will probably result in some false positive results in co-expression analysis. To eliminate the bias, we only retained lncRNAs that expressed in all the tissues with FPKM values ≥0.3. In this way, a total of 3,961 BW lncRNAs and 4,070 BB lncRNAs were used for the subsequent analyses. Surprisingly and interestingly, both Pearson and Spearman correlation ranking analysis indicate that 89.7% (3,554/3961) BW lncRNAs show strong co-expression(r > 0.8 or r < −0.8, p < 0.01) with 646 PTSs-BW. For BB species, Spearman ranking exhibits a similar but tiny higher rate (99%, 4032/4070) of BB lncRNAs that strongly (r > 0.8 or r < −0.8, p < 0.01) co-expressed with 640 PTSs-BB than Pearson’s (98.9%, 4027/4070) (Table S2).

Recent studies found the additional copies of TP53 in elephants genome contribute to their low risk of cancers [34].However, in the genomes of BB, BW and NMR, non-similar extra copies of TP53 was observed. Here we found 106 BB and 432 BW lncRNAs strikingly coexpressed with TP53 and/or TP53INP1 genes(Pearson’s r ≥ 0.9, p < 0.01). InNMR, 133 lncRNA display striking co-expression with TP53, TP53BP1 and/or TP53INP1 genes (Pearson’s r ≥ 0.9, p < 0.01) (Table S2) [27]. Our co-expression analysis indicates that there exists large number of lncRNAs in long-lived mammals may relate with their natural anti-cancer mechanisms.

Kmers-analysis of LLM lncRNAs with human ageing diseases related lncRNAs (HAR-Lncs)

Unlike PCGs, lncRNAs usually lack linear sequence homology despite of similar functions, which makes it difficult to predict their potential function. Recently, Kirk et al developed a sequence comparison method to deconstruct linear sequence relationships in lncRNAs and evaluate similarity based on the abundance of short motifs called k-mers, and they found that lncRNAs of related functions often had the similar k-mers contents [35].

LLMs have been reported to have low incidences of ageing related diseases(ARD) [17,36,37]. To explore possible links that connect LLM lncRNAs and their anti-ARD, we utilized k-mers clustering method to identify LLM lncRNAs that are potentially related with ARD, we collected a list with 29 humanlncRNAs (HAR-Lncs) [38] like Hotair and Malat1 that have been experimental validated to be ARD related especially with cancers (Table S3). Considering those lncRNAs act as multiple roles in ARD and ageing process, we firstly performed a classification on 29 HAR-Lncs by k-mers clustering (Fig. 4). HAR-Lncs were clearly clustered into 7 modules, indicating that HAR-Lncs have 7 types of k-mers distributions (Fig. 4). The biggest HAR-Lnc module consists of 10 lncRNAs (Table 1). We subsequently synthesized the k-mers profiles of all the LLM lncRNAs and compared them with the k-mers contents of the biggest HAR-Lnc module. By calculating the Pearson correlation coefficient (Pearson’s r), 96 of NMR lncRNAs, 136 of BB lncRNAs and 75 of BW lncRNAs displayed extremely high correlation (Pearson’s r ≥ 0.9, p < 0.01) with the 10 HAR-Lncs (Fig. 5, Table S4). In order to illuminate lncRNAs of other interested species that potentially regulating ageing diseases, we generated the k-mers libraries of human and mouse lncRNAs. Using the same parameters, 169 of human lncRNAs and 153 of mouse lncRNAs are highly clustered (Pearson’s r ≥ 0.9, p < 0.01) with HAR-Lncs, which is much more than we identified in longevous species (Table S4).

Figure 4.

29 HAR-Lncs are clustered into 7 blocks by k-mers distribution: the biggest block contains 9 lncRNAs.

Table 1.

Description ofthe 10 HAR-Lncs that clustered into the biggest moduleby k-mers distribution.

LncRNA	Relevance to human ageing diseases/process
TERC	RNA component of telomerase involved in telomere maintenance
Linc-DMRT2	Low in fat tissue of obese humans
HI-LNC80	Elevated by high glucose
H19	High expressed in old muscle
YAM1	Implicated in muscle wasting, cancer, diabetes, chronic heart failure
PINK1-AS	Inactive muscle
RASSF1-AS1	Involved in p53 mediated cell cycle regulation and apoptosis
7SL1	Regulates p53 expression and ER mediated transport
MEG3	PAH-related vascular remodelling by p53 pathway
RMRP	Mitochondria functional

Figure 5.

Heatmap of the correlation between LLM lncRNAs and thebiggest k-mer block of HAR-Lncs: a total of 75–136 of LLM lncRNAs show strikingly correlation (Pearson’s r ≥ 0.9, p < 0.01) with these 9 HAR-Lncs. (A) NMR. (B) BB. (C) BW.

The 10 HAR-Lncs in the biggest module have been found to involve in tumorigenesis via p53 pathway or regulating metabolic process. For example, TERC is the telomerase RNA component which plays the decisive role of replicate senescence through telomere maintenance [39]. The telomerase activity in normal human somatic cells is either non-expressed or exhibiting at low levels whereas it is widely detectable in more than 85% of human tumours [40] . H19 is an imprinted oncofetal lncRNA that expressed in the embryo, down regulated at birth and then reappears in tumours and accumulated evidence has suggested its key function as an oncogene in multiple cancers [41]. In order to investigate the potential effects of the LLM lncRNAs that clustered with 10 HAR-Lncs (Kmer-lncs), we compared their average expression levels with it of all the LLM lncRNAs (All-lnc) by assessing FPKM values. In BB, Kmer-lncs expressed higher in brains and kidneys in both summer active (A) and hibernating periods (H2 and H6) but exhibited no difference in liver tissues (Fig. 6A). Similar with BB, the Kmer-lncs of BW averagely expressed higher in brain and kidneys than All-lncs (Fig. 6B). On the contrary with BB and BW, a lower expression level was observed in NMR Kmer-lncs especially in kidney tissues at three development stages that include 4-years-old (4y), 20-years-old (20y) and 4-years-old with low-oxygen-treated (LO) of naked mole rat (Fig. 6C). The function of the Kmer-lncs remains unknown, but our clustering analysis suggested that a lot of lncRNAs in the long-lived mammals are closely correlated with human cancer related lncRNAs in k-mers distribution and their dysregulated expression in tissues may potentially involve in natural cancer-resistance in long-lived mammals.

Figure 6.

Average expression of Kmer-lncs and All-lncs show significant difference (* indicates p < 0.05, ** indicates p < 0.01) in brains and kidneys at specific stages/ages of three LLMs. (A) BB (‘A’: summer active bats; ‘H2ʹ and ‘H6ʹ separately means bats hibernate for 2 and 6 months, Kmer-lncs). (B) BW. (C) NMR (‘NB’: new-born NMR; ‘4y’ and ‘20y’: NMR at 4-years old and 20-years old;‘LO’: 4-years old NMR with low oxygen treated).

Discussion

As the most longevous mammal, bowhead whale is always a mysterious cetacean with weight up to 100 tons. It has been proposed that the longevity of giant mammals, bowhead whale as well as elephant, stem from a lack of non-human predators and slow development [16]. Nevertheless, a specific expansion of TP53 copies in elephant genome has been associated with the evolution of increased body size and an enhanced DNA damage [34,42], which potentially delays their senescence and avoids cancers. Non-similar extra copies of TP53 was observed in bowhead whale, therefore, its underlying mechanisms that behind the cancer-resistance and longevity are still in urgent need to be recognized. Unlike bowhead whale and elephant, naked mole rat and Brandt’s bat are excluded from the canonical ageing theory of ‘bigger animals live longer’ [43] with surviving exceeds to > 30 years and a mouse-like body size. Natural selection favours them slow ageing, ageing-tolerance and cancer-resistance; however, the genetic mechanisms are rarely explored.

In recent years, lncRNAs have emerged as key modulators in regulating cancer and anti-cancer mechanisms and have been implicated as putative effective therapeutic strategies for cancer treatment [44]. However, the functions of the vast majority of lncRNAs are unknown due to very limited methods include co-expression analysis and k-mers clustering can used for their functional predictions. In order to understand the lncRNAs in non-standard longevous mammal organisms, here we assembled and identified thousands of high-quality lncRNA transcripts in two long-lived mammals, bowhead whale and Brandt’s bat, and performed the comparative analysis with naked mole rat to seek for the potential associations between lncRNAs and their natural cancer-resistance. The lncRNA transcripts that we identified in this study provide basic data for future ncRNA evolutionary study. We also observed that LLM lncRNAs similar with other non-longevous species, exhibit dynamic expression patterns spanningacross tissues and different behavioural stages. Using k-mers clustering and co-expression methods, wefoundthousands of them are strikingly co-expressed with tumour suppressors and identified hundreds themshare similarities with human ageing diseases related lncRNAs (HAR-Lncs).

Although lncRNA usually lacks canonical sequence conservation like protein coding gene does across species, accumulative evidence indicated that many lncRNAs are functional conserved [45,46]. Some functional conserved lncRNAs like LINC-PINT genes in human and mouse, harbour short sequence similarities that confer shared functions, even if conventional alignment algorithms do not detect the similarity [35,47]. We currently do not know whether there exists any anti-cancer related lncRNAs in long-lived mammals are playing conserved roles in human or not, the aim that we performed lncRNA identification and associated analysis is to provide basis for future lncRNA-related cancer preventive studies and trials. In summary, our cross-species investigation provides the first comprehensive genome-wide analysis of LLMs lncRNAs and reveals their prospective associations with cancer resistance, which is of significance for a better understanding of human anti-cancer study.

Materials & methods

Data accessibility

The raw BW and BB transcriptome datasets were downloaded from National Centre for Biotechnology Information (NCBI) database (PRJNA178678 and PRJNA194091). The assembled genome sequence and genomic annotation files were requested from GIGA database (http://gigadb.org/dataset/100022) and bowhead whale genome resources database (http://www.bowhead-whale.org/). NMR profiles were obtained from our previous study [27] and the naked mole rat genome resources database (http://www.naked-mole-rat.org/). Human (v29) and mouse (M19) lncRNA sequences were acquired from GENCODE database.

The fasta files, FPKM values and co-expression files of BW, BB and NMR lncRNAs by Pearson and Spearman methods can be accessed via:https://onedrive.live.com/?id=189757ACF263B018%21106&cid=189757ACF263B018, or https://drive.google.com/drive/folders/1-A2aXZgU8a8OGeSilf7_3YkC0EdFNt05?usp=sharingor clicking this link: https://pan.baidu.com/s/1e46odTZnr7H83IuBZalYJw with code: ‘yvx3’

LncRNA identification pipeline

A stepwise filtering workflow that based on our previous pipeline with minor revisions was used to identify long noncoding RNAs in two longevous mammals [27]. Briefly, raw RNA-seq reads were firstly trimmed by SolexaQA [48]. The cleaned data were then mapped to genome using Hisat2 [49] and assembled by Cufflinks packages [50]. Assembled and merged transcripts were filtered by length. Transcripts that longer than 200bp were retained for assessing their coding potential by Coding Potential Calculator 2 (CPC2) [51]. Transcripts that labelled with ‘noncoding RNA’ by CPC2 and their exon numbers larger than 2 were defined as long noncoding RNA.

Conservation analyses

OrthoMCL package-BLASTN programme [30] was employed to analysis the sequence conservation among LLM lncRNAs. BLASTN was selected and transcripts with ≥ 50% coverage and E-value < 1e-5 were clustered as homologous groups.

Functional prediction by K-mer analysis

SEEKR packages were obtained from GitHub [35] to synthesize the k-mers profiles and to calculate the similarities among lncRNAs from different species following the tutorials with default parameters.

Co-expression analysis

To investigate the potential correlations between lncRNAs and cancer-resistance in LLM, we performed co-expression analysis between LLM lncRNAs and PTSs-LLM by testing their FPKM values. LLM lncRNAs and PTSs that expressed in all tissues with FPKM ≥ 0.3 were used for the analysis. Cufflinks package was used to calculate the FPKM values. Pearson and Spearman correlation coefficient (r) and correlation P value (p) were used to assess the co-expression relationship by an in-house R script.r > 0.8 or r <−0.8 with pvalue < 0.01 was considered as strong correlation, and r ≥ 0.9 or r ≤ −0.9 with p value <0.01 was regarded as strikingly correlation.

Statistical analysis

In-house Perl scripts were used for calculating length and GC contents of LLM PCGs and lncRNAs. One-way ANOVA was used for pairwise comparisons. For all the comparison, * and ** indicate p < 0.05 and p < 0.01, respectively

Funding Statement

This work was supported jointly and equally by grants from the National Key R&D Program of China [No. 2018YFC2000400, 2018YFE0203700]. Additional support was received from the Key Research Program of Frontiers Science [QYZDB-SSW-SMC020] and Youth Innovation Promotion Association [to Q.P.K.] of the Chinese Academy of Sciences, National Natural Science Foundation of China [91749109, 81671404], and Yunnan Applied Basic Research Project [2017FA038].

Author contributions

KQP and JJJ conceived and designed the study. KQP and JJJ performed all analysis and drafted the manuscript. KQP and JJJ revised the manuscript. All authors read and approved the final manuscript.

Disclosure statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Supplementary material

Supplemental data for this article can be accessed here.