Forward–reverse mutation cycles in cancer cell lines under chemical treatments

Abstract

Spontaneous forward–reverse mutations were reported by us earlier in clinical samples from various types of cancers and in HeLa cells under normal culture conditions. To investigate the effects of chemical stimulations on such mutation cycles, the present study examined single nucleotide variations (SNVs) and copy number variations (CNVs) in HeLa and A549 cells exposed to wogonin-containing or acidic medium. In wogonin, both cell lines showed a mutation cycle during days 16–18. In acidic medium, both cell lines displayed multiple mutation cycles of different magnitudes. Genomic feature colocalization analysis suggests that CNVs tend to occur in expanded and unstable regions, and near promoters, histones, and non-coding transcription sites. Moreover, phenotypic variations in cell morphology occurred during the forward–reverse mutation cycles under both types of chemical treatments. In conclusion, chemical stresses imposed by wogonin or acidity promoted cyclic forward–reverse mutations in both HeLa and A549 cells to different extents.

Introduction

Cancer development has been investigated extensively at the molecular and genetic levels to better understand the dynamics of its progression, resistance to therapy, and biomarkers for risk stratification [1, 2]. While human malignancies are thought to originate mostly from a single cell, most tumors show startling heterogeneity in genomic structure, cell morphology, proliferation rate, and metastatic potential by the time of diagnosis [3,4,5]. Genetic markers, including loss of heterozygosity (LOH), gain of heterozygosity (GOH), and copy number variation (CNV), have delineated some of the mutations that play important roles in the emergence and proliferation of invasive tumors [6,7,8,9,10], and studies have revealed the important role of heterogeneity within tumors that would lead to the appearance of subclones with a fitness advantage in malignancies ranging from hematopoietic cancers to different types of solid tumors [11,12,13,14], resulting in the basic hallmarks of cancer: sustained proliferative signaling, evasion of growth suppressors, resistance to cell death, replicative immortality, induction of angiogenesis, and activation of invasiveness and metastasis [15]. Morphologically, the evolution from hyperplasia or dysplasia to neoplasia and malignant neoplasia has long been recognized, although the rates of progression or regression could seldom be ascertained [16]. In recent years, the process of cancer cell evolution has been examined extensively in terms of genomic changes, exemplified by the twenty-four frequently mutated genes found in colon and rectal cancers, including APC, TP53, SMAD4, PIK3CA and KRAS [17, 18]. Moreover, breast cancer-like stem cells (CSC) have been found to switch between fully tumorigenic states and other tumor stages, which may signal some form of reorganization within the cell [19, 20].

Our group has examined the question of whether the allelic SNVs occurring in various cancers might undergo systematic transitions between the genomes of normal blood cells (B-stage), morphologically largely ‘nontumor’ cells near the tumor (N-stage), ‘paratumor’ cells bordering the tumor (P-stage), and the tumor cells themselves (T-stage) in lung, liver, stomach, metastatic brain tumors and odontogenic myxoma. It was found that when a homozygous site in the B-stage genome was converted to a heterozygous site through a GOH mutation, the mutation was often followed by an LOH mutation at the same base position such that the pre-mutation homozygous status occurring in B-stage was restored. Likewise, when a heterozygous site in the B-stage genome was converted to a homozygous residue through an LOH mutation, the LOH was often reversed by a subsequent GOH mutation, restoring the pre-mutation heterozygous status [21, 22]. These forward–reverse mutation cycles of SNVs and CNVs could span the B–N–P stages or the N–P–T stages. Such forward–reverse mutations were observed at population level, i.e., emergence-disappearance of certain mutations in a population of cells under directional selection, rather than direct observation of mutation-backmutation in the same cell. Furthermore, forwardreverse mutation cycles were also observed in cultured HeLa cells spanning different consecutive time frames of cell growth following a change in culture medium. A plausible explanation for the importance of such SNV mutational cycles in tumor development could be that widespread forward mutations enable cells to acquire a range of random mutations, including those cancerous mutations that are essential to the hallmarks of malignancy. However, the numerous passenger mutations which accompany the appearance of the essential malignant mutations would reduce the vigor of cell growth and replication unless many of them are restored to their pre-mutational allelic states, which have been optimized by centuries of evolutionary selection. Therefore, the forward–reverse mutation cycles could contribute to the assembly of an adequate number of cancerous genes to bring about tumorigenesis. Moreover, the fact that already malignant HeLa cells also displayed such cycles suggests that these cycles assist not only in tumorigenesis but also in other cellular alterations that expedite vigorous cancer cell properties such as invasiveness and drug resistance. Accordingly, in the present study, the forward–reverse mutations elicited by treatment with wogonin and acidic pH were investigated in HeLa and A549 human cancer cells with respect to both DNA and cell morphology in order to assess whether there might be any observable coordination between the genetic and cell-structure consequences of the mutations.

Methods

Cell culture

The HeLa cervical adenocarcinoma and A549 lung adenocarcinoma human cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and cultivated in neutral Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (Gemini Bio-Products, Sacramento, CA) under a humidified atmosphere of 5% CO₂ at 37 °C. Neutral DMEM growth medium was buffered by 15 mM 3-(N-morpholino)propanesulfonic acid.HCl (MOPS) at pH 7.4, and acidic DMEM growth medium was buffered by 15 mM 2-(N-morpholino)ethanesulfonic acid.HCl (MES) at pH 6.4. The medium was changed every 2–3 days, and cells were collected for sequencing simultaneously with the medium change.

Wogonin treatment

Wogonin was obtained from Wako Pure Chemical Industries Ltd, Osaka. Stock solution was prepared at 250 mM in DMSO and diluted to different concentrations with DMSO. Cells were incubated with 0 or 50 uM wogonin.

Cell morphologies

Cells were grown on 6-well plates overnight in neutral DMEM prior to incubation with either wogonin in neutral DMEM, or with acidic DMEM for different lengths of time. Cell images were captured using the Nikon TS 100 live cell imaging observatory at 100× or 200× magnification. Cell length and circularity were measured using the ImageJ software, and the cells were classified into four morphologies based on cell length and circularity. The number of cells with each morphology was calculated for the two cell lines under acidic pH or wogonin treatment.

DNA extraction and AluScan sequencing

The cells were lysed with 500 μL tris–HCl (TES) buffer, homogenized for 10 min, and mixed with 250 μL alkaline phenol and 245 μL chloroform. The mixture was incubated for 10 min at room temperature and centrifuged; and the aqueous phase was added to an equal volume of isopropanol and 300 μL of sodium tris–EDTA (TE) buffer. After storage at − 20 °C overnight and centrifugation, the pellet was washed with 1 mL 75% ethanol, and resuspended in TE buffer. Samples were stored at  − 20 °C for further analysis. PCR using Alu retrotransposon consensus sequences-based primers was carried out as described [21,22,23]. A mixture of the four primers AluY278T18, AluY66H21, L12A/8, and R12A/267 was employed to produce the amplicons: DNA was denatured at 95 °C for 5 min, followed by 35 cycles of 30 s each at 95 °C, 30 s at 54 °C, 5 min at 71 °C, and finally another 5 min at 71 °C. Amplicons were purified with ethanol precipitation, and ≥ 3 μg purified products per sample were employed for Illumina GAII library construction and sequencing at Beijing Genomics Institute (Shenzhen, China).

SNV and CNV calling

SNV calling for the paired-end sequencing reads on the Illumina platform was performed for the AluScan sequences. Bioinformatics analysis including alignment, sorting, recalibration, realignment, and removal of duplicates using BWA (Burrows-Wheeler Aligner, version 0.6.1), SAMtools (Sequence Alignment/Map, version 0.1.18) and GATK (Genome Analysis Tool-Kit, version 3.5) to identify SNVs according to the standard framework as described [24]. The “UnifiedGenotyper” module of GATK was employed for genotyping the SNVs. Only genomic sequence regions with sufficient coverage, i.e., read depth > 8, were included in the analysis. The following parameters were used to filter SNVs of different genotypes: major allele frequency ≥ 95% for the “MM” loci; major allele frequency between 30 and 70% and QD ≥ 4 for the “Mm” or “mn” loci, and minor allele frequency ≥ 95% and QD ≥ 20 for the “mm” loci. Different QD values were used for “Mm” or “mn” and “mm” loci because, for homozygous variants, the number of reads contributing to the quality value was twice that of heterozygous variants. Consequently, the QD would be larger at the same depth. QD values of ≥ 4 for the 'Mm' or 'mn' loci and ≥ 20 for the 'mm' loci were used to ensure that strand bias, estimated using Fisher's exact test (FS), had an FS value ≤ 20 for both heterozygous ‘Mm’ or ‘mn’ loci and homozygous 'MM' or 'mm' loci [21]. CNVs were called by AluSc'nCNV2 based on 10 kb to 500 kb window sizes on autosomes; CNVs which appear in over 25% sample in the same window size would be regarded as recurrent CNV [25].

Overlapped between CNV and genomic features

The density or intensity of various genomic elements overlapping with the CNVs was estimated as described [26]. The density or intensity of each of the genomic features that overlapped with CNV was expressed relative to the genome average. The average score of the genomic feature within each CNV region was computed by the binnedAverage function in R. The weighted average score for CNV regions was calculated by summing the product of the scores and widths of CNV regions, and then dividing by the total width of all CNV regions. This value was then compared to the average score of the genomic feature across the entire genome to determine the relative enrichment or depletion of the genomic feature in CNV regions.

Statistical analysis

To assess the statistical significance of the SNV rate between the forward or reverse days and the other days, permutation tests with 10,000 permutations were performed. The resulting p-value quantifies the probability that the observed difference could occur by random chance. Chi-square tests were performed to determine the statistical significance of CNV phase distributions between the forward or reverse days and the other days, as well as to assess the differences in genomic zone distributions between sequences within different window sizes.

Results

Single nucleotide variations under chemical treatment

HeLa and A549 cell lines were treated with wogonin or acidic pH over three weeks and analyzed for SNVs. In either instance, using the HeLa or A549 genome in the preceding time frame as control, duplex bases on the control genome were recorded as MM, mm, or Mm loci, where M and m referred to a major or minor allele on the human reference genome hg19. On this basis, the SNVs of mutated residues from MM/mm to Mm were referred to as GOH mutations, whereas the SNVs of mutated residues from Mm to MM/mm were referred to as LOH mutations. Upon tracking the GOH and LOH occurrences over the three weeks, the HeLa cells were found to engage in a forward–reverse mutation cycle during treatment with wogonin (Fig. 1a, left panel), generating 250 GOH mutations from its 711,547 MM residues recorded on Day 14 via the G1-mutation step; all of these GOHs back mutated to MM via the L1-step, and none mutated to mm via the L2-step through an LOH on Day 16. As a result, the LOHs formed from the 250 GOHs were totally biased in favor of restoring the pre-mutated control MM loci (Fig. 2a). Likewise, there were 21 GOH mutations from its 196 mm residues recorded on Day 14; 20 of them back mutated to mm via the L4-step, and none mutated to MM via the L3-step on Day 16, which again favored restoring the pre-mutated control mm loci relative to any novel MM loci. Therefore, there was a strong propensity for the DNA residues mutated during the forward stage of the cycle to restore the original residues in the reverse stage, as highlighted by yellow triangles in the figure. The A549 cells under wogonin treatment also exhibited comparable mutations (Fig. 1a, right panel; Fig. 2b). On the other hand, when the HeLa and A549 cells were exposed to acidic pH, forward–reverse mutation cycles were observed on Days 1–3, 10–14, and 16–18 for HeLa cells, and on Days 9–11 and 30–32 for A549 cells (Fig. 1b). The chromosomal positions of the individual SNVs observed among the forward–reverse mutations are shown in Fig. 3. In accordance with Fig. 1, the forward mutations of MM and mm residues contained more GOHs, whereas their reversals contained more LOHs in Fig. 2. Moreover, when the forward and reverse SNV mutations are plotted in Fig. 4, where the Total Forward SNVs and Total Reverse SNVs are profiled to reveal the signatures of human cancers [23], the mutations were enriched in the C to T (blue) or T to C (pink) bars while the GOH in the forward stage is slightly more than the LOH in the reverse stage.

Fig. 1

A. GOH and LOH mutations of control MM, mm or Mm residues in HeLa (left diagram) or A549 cells (right diagram) under wogonin treatment. B. GOH and LOH mutations of control MM, mm or Mm residues in HeLa or A549 cells under acidic treatment. In each instance, GOHs outnumbered LOHs during the forward stage (red arrow), whereas LOHs outnumbered GOHs during the reverse stage (blue arrow). (***P < 0.005; **P < 0.01; *P < 0.05)

Fig. 2

Mutations of MM, mm, and Mm residues during forward and reverse stages. The MM, mm, and Mm residues of HeLa (A) and A549 (B) cells under wogonin treatment from day14 to day18 were estimated. G1 to G6 or L1 to L6 represent gains or losses of heterozygosity respectively. LOHs or GOHs showing large lineage effects were highlighted by yellow triangles. The red and blue highlighted dates were indicative of the forward and reverse stages respectively. The figures under each of the G1-G6 mutation stage represent the mutations observed from that stage

Fig. 3

Chromosomal position of SNV mutations during the cyclic forward and reverse mutations. A. Mutations in HeLa cells under wogonin treatment. B. Mutations in A549 cells under wogonin treatment. The red and blue arrows above each vertical panel show the dates of the forward and reverse mutations respectively, and the small color-coded rectangles are indicative of the major homozygous (MM), heterozygous (Mm), and minor homozygous (mm) sites respectively

Fig. 4

Profiles of SNVs in the forward and reverse stages. Top row: the total SNVs displayed by MM residues of HeLa and A549 cells under wogonin and acidic treatments during the forward and reverse stages; middle and bottom rows: GOH mutations and LOH mutations are shown in terms of their mutated triplets among the 16 × 6 kinds of triplets, with the second base of each triplet representing the mutated base, respectively. The pink columns highlight the intensively mutated T > C triplets, and the blue columns highlight the intensively mutated C > T triplets. There were more C to T and T to C GOHs in the forward stage than the reverse stage, but more C to T and T to C LOHs in the reverse stage than the forward stage

Copy number variations under chemical treatment

Within each of the eight vertical panels, when plotted according to the replicating period of the DNA, i.e., in G1b, S1, S2, S3, S4, or G2 phase (Fig. 5), the HeLa and A549 cell lines under wogonin treatment showed more CN-losses during the forward stage on Day 16 and more CN-gains during the reverse stage on Day 18. A majority of the CNVs occurring in the forward–reverse cycle involved DNA in the late S3-G2 replication phases. The phase distribution of each day in accordance with SNV of HeLa and A549 cells under wogonin or acidic pH is shown in Supplementary Fig. 1. This was also the case when the HeLa and A549 cells were exposed to acidic pH (Figs. S1c and S1d). The human genome has been classified into three types of sequence zones: Genic, Proximal, and Distal, based on the pairwise co-localizations of forty-two genomic features [26]. Figure 6a shows the relative distribution of CNVs from 10 to 500 kb windows: there were comparable ratios of Genic zone sequence: Proximal zone sequence: Distal zone sequence in the DNA fragments devoid of CNV and found in the windows of DNAs of varied sizes (upper panel). In contrast, among the DNA fragments with CNV, the smaller 10–50 kb fragments contained proportionately more Genic zone sequence compared to Proximal zone sequence or Distal zone sequence, while in the larger 100–500 kb fragments, the proportion of Proximal zone sequence or Distal zone sequence was higher (lower panel).

Fig. 5

Distribution of the observed CNVs among the DNA-replication phases G1b to G2. The L/G ratios indicated the ratio between CNVL and CNVG on different dates from the start of wogonin treatment of the HeLa or A549 cells. The total number of base pairs mutated included are shown on a scale of 0–4 on the y-axis. The open and hatched color bars represent CNVG and CNVL respectively, while the red and blue shaded dates represent the forward and reverse stages. Both type of cells yielded higher L/G ratios in Day 16 of the forward stage relative to the reverse stage. (***P < 0.005; **P < 0.01; *P < 0.05)

Fig. 6

Association of CNVs in genomic feature under wogonin treatment. A. Relative association of CNVs with three types of genomic sequence zones. Upper panel shows abundance of genomic zones in sequences without CNVs, while lower panel shows sequences with CNVs, in fourteen different window sizes (10–500 kb). The abundance of sequences belonging to the three types of genomic zones, Genic (blue), Proximal (green), Distal (red) zones measure in numbers of base pairs, showing prominent increment of genic zone sequences in small windows of CNVs. Amounts of base pairs belong to the three types of genomic sequence zones in the human reference genome hg19 are shown on the right-hand side of panel A. (***P < 0.005; **P < 0.01; *P < 0.05). B. Relative enrichment of various genomic features in CNVs occurred during forward stage (Upper panel) and reverse stage (Lower panel). Genomic features are labeled in their abbreviations (refer to Table S1 for full names) and arranged in four groups based on their relative enrichment in the three types of sequence zones (Genic, Proximal, Distal) of human genome and their use as genetic makers (Marker). CNVs were captured with fourteen different window sizes arranging from 10 to 500 kb, including copy number gains or copy number losses. Relative enrichment of genomic features at different window sizes were measured against genome averages and quantified in log-2 of fold change as illustrated in red-blue thermal scale. Red color illustrates enrichment while blue color illustrates deficiency of genomic features captured in CNVs. C. Distributions of 10 kb (in red) or 500-kb (in blue) CNVs and representative histone modification sites H3k27me3 (in green) and H3k79me2 (in orange) in 10-kb windows on the region spanning from 120,000 to 121,000 kb on chromosome 1. The red dashed lines represent the genome averages of H3k27me3 (1.64) and H3k79me2 (2.46), respectively

In addition, analysis of the association between genomic features and CNVs from the forward and reverse stages in 10-kb to 500-kb sized windows [27] revealed that the enrichment of CNVGs or CNVLs in H3K79me2 was reduced relative to other histone binding sites. Among the observed changes in genomic features affected by wogonin (Fig. 6b), the evolutionarily CpG rich regions (CpGe), segmental duplications (SDP), single-nucleotide polymorphism mapped to multiple locations (SNPM), and cluster (where different kinds of genetic-variant hotspots overlap) were also found to be enriched in CNVs; notably, these enrichments were more evident in CNVLs rather than CNVGs in the CNVs, which occurred in more than 25% of samples (Figure S2). The enrichment of genomic features in CNVs under acidic pH treatment was shown in Figure S3, which was similar to the enrichment profile of wogonin treatment. The results based on recurrent CNVs from wogonin or acidic treatment are shown in Fig. S4.

Cell morphology changes under chemical treatment

During cancer development, cells typically undergo both morphological and genomic alterations. Upon exposure of HeLa and A549 cells to wogonin or acidic pH, the morphologies of both cell lines exhibited evident variations over time. For HeLa cells, the sum of cells with types 1 and 2 morphologies increased in some time-frames, while the sum of cells with types 3 and 4 morphologies decreased during the same time-frames, and vice versa. These cell type changes in combination with the SNV findings in Fig. 1 and the CNV findings in Fig. 5 suggest that the types 1 and 2 morphologies were correlated with GOH increases and CN-losses pertaining to the MM and mm residues, whereas the types 3 and 4 morphologies were correlated with their LOH changes. For A549 cells, the sum of types 1 and 2 morphologies likewise trended to vary in the opposite direction to those of types 3 and 4 morphologies (Fig. 7). Therefore, the types 1 and 2 morphologies in HeLa and A549 cells were enhanced under conditions suboptimal for cell growth, while the types 3 and 4 morphologies were enhanced under conditions optimal for cell growth.

Fig. 7

Diversity of cell morphology. A. Different types of cell morphology of HeLa and A549 cells under wogonin treatment. B. Different types of cell morphology of HeLa and A549 cells under acidic treatment. The red arrows above the charts for HeLa and A549 cells mark the typical dates for the forward phases, while the blue arrows mark the typical dates for the reverse phases, of the cyclic SNV mutations

Discussion

Influence of chemical treatment on forward–reverse mutation cycles

Previous studies have observed forward–reverse mutations in both normal cultured HeLa cell line and cancer patients [21, 22]. The present study observed the occurrence of forward and reverse mutations under the influence of chemical treatments. The forward–reverse mutation profiles described here were experimentally observed at the population level rather than the cellular level, which should have resulted from directional selection on a prove of cancer cells with diverse mutational characteristics. The result of such population level analysis could not imply any pair of forward–reverse mutations actually occurred in the genome of the same cancer cell. Under the influence of wogonin, forward mutations were observed on day 16, followed by reverse mutations on day 18 in both cell lines, with a relatively high mutation rate. When exposed to acidic pH, both cell lines exhibited multiple and earlier occurrences of the forward–reverse cycles, but with a relatively lower mutation rate than with wogonin. These suggest that acidic pH at pH 6.4 could pose greater stress on these cells compared to 50 mM wogonin. The disparity in forward–reverse cycles observed across different chemical treatments, yet the consistency across varied cell lines, suggests that the impact of chemical treatments on forward–reverse mutations is agent-specific. The interdependence between successive mutations resulting in a linear selection of mutations in the forward–reverse mutation and the CG to TG rich GOH followed by TG to CG rich LOH is consistent with previous study [21]. These indomitable forward–reverse mutations persist not only in neutral medium but also under chemical agents, highlighting its potential clinical significance.

Contrast between forward and reverse part of the mutation cycles

During the forward stage, GOH occurs in SNV and CNVL occurs in CNV, and when reverse stage, LOH appears in SNV and CNVG appears in CNV. Meanwhile, the positions where forward reverse mutations occur in SNV remain consistent. In CNV, the feature association results shown in Fig. 6b indicate that the positions of CNVs where forward–reverse mutations occur are slightly shifted but generally remain the same. The cells were directionally selected to have smaller genome sizes, as illustrated by fewer CNVGs on the reverse day compared to CNVLs on the forward day, and more CNVLs on the reverse day compared to CNVGs on the forward day (Fig. 5). The cell morphology also showed differences during the forward–reverse cycle under both wogonin and acidic treatments. In the forward stage, cells predominantly exhibited types 1 and 2 morphologies, which are relatively small and round. In the reverse stage, the number of cells with types 3 and 4 morphologies, which are elongated and larger, increased.

Colocalization of genomic features with CNVs in the mutation cycles

We analyzed the association between CNVs appearing in forward or reverse stages and the genomic features. Most features within the Genic zone exhibit distinct CNV association patterns between window sizes, with predominantly CNVs enriched with features in 10-100 kb windows and fewer CNVs in 150-500 kb windows. The association of CNVs and features remains consistent across different window sizes in the Proximal zone, while those in the Distal zone display a reversal of the pattern observed in the Genic zone. It is notable that there are smaller CNVs in the Genic zone sequences as appeared in 10–50 kb windows. However, there are more CNVs in lager windows in the Proximal zones and even more so in the Distal zones. This result indicated that Genic zone sequences are less tolerable to lager size CNVs. The initiation of forward mutations from CNVL, coupled with the high association between CNVs and Proximal zone features, may suggest that cancer genome mutations start by targeting regions with low gene density, potentially indicating a self-reducing mechanism at the beginning of the forward–reverse mutation cycle. However, this does not mean that gene-rich regions are not targeted; on the contrary, it reflects the critical importance of these regions, which may make them less tolerant to the effects of mutations. Moreover, regions known for genomic instability, such as SDP and SNPM, appear high CNV association, underscoring their link to genetic instability. Interestingly, the CNVs displayed colocalizations with the MIR and L2 retrotransposons, but not with the Alu and SVA in 100–500 kb windows or L1 in 10–50 kb windows (Figure S2 and S4). In conclusion, these findings highlight the complex interplay between genomic features and mutations, offering insights into the underlying mechanisms driving forward–reverse mutations with various genomic contexts.

Closing remarks

The present study demonstrated that the forward–reverse mutation cycles evident from SNVs and CNVs, along with cell morphology changes, could be differentially affected by chemical insults, including the anti-cancer agent wogonin and acidic pH mimicking cancer micro-environment. Under both wogonin and acidic treatments, the forward stage of the mutation cycles was dominated by GOH and CNVL, in contrast the reverse stage showed more LOH and CNVG. The CNV regions were enriched with genomic sequences associated with instability, such as SDP. The present study demonstrated chemical treatment specific accelerations of forward–reverse mutation cycles, rendering increased somatic genome diversity to cancer cells, which in turn would likely provide malignant cancer cells with added vigor. The observed effects of chemical treatment on forward–reverse mutation cycles should be taken into account in a broader range of cell biology studies and further evaluated for potential applications in anti-cancer drug development and precision chemotherapy.