Abstract
Background/Aim: Dynamic DNA sequences (i.e. sequences capable of forming hairpins, G-quadruplexes, i-motifs, and triple helices) can cause replication stress and associated mutations. One example of such a sequence occurs in the RACK7 gene in human DNA. Since this sequence forms i-motif structures at neutral pH that cause replication stress and result in spontaneous deletions in prostate cancer cells, our initial aim was to determine its potential utility as a biomarker of prostate cancer. Materials and Methods: We cloned and sequenced the region in RACK7 where i-motif deletions often occur in DNA obtained from eight individuals. Expressed prostatic secretions were obtained from three individuals with a positive biopsy for prostate cancer and two with individuals with a negative biopsy for prostate cancer. Peripheral blood specimens were obtained from two control healthy bone marrow donors and a marrow specimen was obtained from a third healthy marrow donor. Follow-up computer searches of the genomes of 74 mammalian species available at the NCBI ftp site or frequencies of 6 dynamic sequences known to produce mutations or replication stress using a program written in Mathematica were subsequently performed. Results: Deletions were found in RACK7 in specimens from both older normal adults, as well as specimens from older patients with cancer, but not in the youngest normal adult. The deletions appeared to show a weak trend to increasing frequency with patient age. This suggested that endogenous mutations associated with dynamic sequences might accumulate during aging and might serve as biomarkers of biological age rather than direct biomarkers of cancer. To test that hypothesis, we asked whether or not the genomic frequencies of several dynamic sequences known to produce replication stress or mutations in human DNA were inversely correlated with maximum lifespan in mammals. Conclusion: Our results confirm this correlation for six dynamic sequences in 74 mammalian genomes studied, thereby suggesting that spontaneously induced replication stress and mutations linked to dynamic sequence frequency may limit lifespan by limiting genome stability.
Repair of double-strand breaks (DSBs) at stalled replication forks generally involves multiple mechanisms in eukaryotes, several of which elicit homologous recombination where short clonable fragments can be produced during fork restart (1). In previous work (2) we used the ligation mediated polymerase chain reaction (LMPCR) to clone spontaneously produced DNA fragments from isolated genomic DNA and then mapped the isolated sequences to the human genome in a screen for DSB hot spots. Sixty-two percent of the sequences isolated from a high passage prostate cancer cell line (PC3) mapped to within 5 kb of dynamic DNA sequence motifs. Dynamic sequences are known to be capable of forming non-B DNA structures like the G-quadruplex, i-motif, triplex and the foldback or the hairpin (Figure 1). Somewhat surprisingly, thirty-three percent of the sequences isolated from a primary prostate stromal cell line (IPS19I) were mapped to within 5 kb of a dynamic sequence motif. Since the primary stromal cell line was established from a normal individual the data suggested that either cell culture promotes double strand breaks at sites of dynamic DNA structure or that such breaks are common in apparently normal somatic cells.
Structures formed by dynamic sequence motifs. Triplex: This structure forms in duplex DNA. It is generally formed in AT rich regions and requires bases to pair in triplets (e.g., T:A:T). C:G:C+ triplets can also occur in triplex termed H-DNA; however, this generally requires low pH needed to protonate N3 of cytosine. The fourth strand is not base paired. i-Motif: This is a four-stranded structure that can form on the C-rich strand in regions of GC skew. It contains C:C+ base pairs in parallel strands of DNA. In this structure C:C+ base pairs are intercalated so as to link sets of two strands of parallel DNA in a four-stranded structure. The term i-motif or intercalated motif describes its unique base pairing property. Like H-DNA: these structures generally form at low pH; however, a significant number of sequences have recently been identified that form at neutral pH. G-quadruplex: This is a four-stranded structure that can form on the G-rich strand in regions of GC skew. In this structure four G residues are Hoogsteen base-paired in a planar structure. Stacking interactions supplemented by ions chelated between base planes favor the formation of these structures and they are typically highly stable under physiological conditions. Hairpin: Hairpin or foldback structures can occur on either strand in self complementary regions in duplex DNA. They form duplexes that contain Watson-Crick base pairs. This figure was modified from reference (42).
In subsequent work (3), we showed that deletions within the RACK7 gene were the result of i-motif formation at physiological pH by a (TCCC)9 element that was detected in the screen of the PC3 cell line for dynamic sequences linked to double strand breaks. Functional RACK7 recruits the histone demethylase KDM5C to down regulate RACK7 targeted enhancers and super enhancers by modulating H3K4 methylation levels (4). Since these active enhancers and super enhancers support tumorigenesis, RACK7 functions as a tumor suppressor gene (4). This has been demonstrated with RACK7 knockouts that were shown to produce elevated levels of H3K4 methylation and increased transcription from regions around RACK7 targeted enhancers (4). Consistent with a role in tumorigenesis, RACK7 has been shown to undergo breakage and reunion to form fusion proteins in some breast cancers (5) and leukemias (6), while high frequency mutations are observed in certain colon cancers (7). Given its role in tumorigenesis, deletions at the (TCCC)9 locus in RACK7 might be expected to be confined to tumor tissue. This is consistent with the absence of RACK7 clones in the screen of the prostate stromal cell line HSP-19I (2). However, in this report we noted that these deletions were present in normal human tissue, as well as cancer specimens and cell lines.
Much of the data on mutagenesis associated with dynamic sequences suggests that mutations are often associated with a breakdown in repair systems associated with disease (8-12). However, there is also a significant body of evidence suggesting that the probability of a somatic mutation at a dynamic DNA sequence depends on mere presence of the sequence and its intrinsic tendency for non-B structure formation (13-15). This suggests that mutation frequency would depend on the number of replication and transcription events experienced at dynamic DNA sites in a given tissue, as well as the efficiency repair. Hence, it is expected that the number of replication events and the potential for DNA damage experienced at sites of dynamic sequence will be linked to their frequency in a given genome. In short, spontaneous mutations associated with dynamic DNA sequences might contribute to the DNA damage that accumulates with age. In support of this possibility Nothobranchius fuzeri, the vertebrate with the shortest known lifespan in captivity has been shown to contain an unusually high tandem repeat frequency (16). To test that hypothesis, we asked whether or not the frequency of several dynamic sequences known to produce replication stress or mutations in human DNA is inversely correlated with maximum lifespan in mammalian genomes. Our results confirm this inverse correlation for six sequences capable of hairpin, i-motif, G-quadruplex and triple-helix formation in 74 mammalian genomes studied.
Materials and Methods
Patient information. Expressed prostatic secretion specimens were obtained from five patients prior to biopsy for prostate cancer under IRB approved protocols as previously described (17, 18). Peripherial blood was drawn from two healthy marrow donors and marrow was isolated from one healthy donor. Additional patient information is given in Table I and Table II.
Prostate biopsy patient data.
Marrow donor patient data.
B-Cell isolation. B-Cells were isolated from peripheral blood specimens from two healthy bone marrow donors as CD19+ cells using CD19 Microbeads (Miltenyl Biotech, Gaithersburg, MD, USA) using the method described by the manufacturer.
DNA isolation. DNA was isolated from each specimen using the QIAamp® Blood Mini kit (Qiagen Sciences, Frederick, MD, USA) using the instructions supplied by the manufacturer.
PCR amplification, cloning and sequencing of RACK7 region from Patient Specimens. PCR amplification and the cloning of PCR products from the RACK7 gene in Patient Specimen DNA was carried out as previously described (2). The resulting plasmids were isolated also as previously described (19). Plasmids with inserts of the appropriate size were sequenced at the City of Hope DNA Sequencing Lab or by Functional Biosciences Inc. (Madison, WI, USA) using the Sanger Sequencing method and M13 Reverse Primers.
DNA sequence alignment software. Sequences were aligned against the human reference sequence using MAFFT (20) and T-Coffee (21) online alignment packages as previously described (2).
Genomes and genome searches. All mammalian genomes were downloaded from the NCBI FTP site (https://ftp.ncbi.nlm.nih.gov/genomes). Each genome was the latest available at the time of the study. Exact genome identifiers for each genome along with its common name, systematic name, order, and family are given in Table III. Genomes were loaded into a Mathematica (Wolfram Research, Champaign IL, USA) program as a string of uppercase letters which was searched for the count of each sequence string, the length of the genome and the number instances of the letter “N” for the uncalled bases in the genome using the StringLength and StringCount subroutines in Mathematica. The Mathematica code for a representative genome search is given in Figure 2.
Mammalian genome information.
Representative Mathematica program code. In this example, the genome of the Asian Elephant was loaded as upper-case letters into Mathematica and searched using StringCount for each of the six step repeating elements studied and the number instances of uncalled bases “N” in the genome. StringLength was used to determine the number of base-pairs in the genome.
Results
RACK7 deletions are present in both normal tissues and tissues from cancer patients. In previous work, the IPS-19I cell line, a low passage primary culture established from a normal individual, was found to have normal karyotype (2). However, when we sequenced the region containing the (TCCC)9 i-motif forming element in RACK7 we noted that deletions similar to those observed in PC3, a high passage prostate cancer cell line (2, 3). Since the deletions in this apparently normal cell line may also have occurred during the establishment of the cell culture, we turned to human specimens.
Representative sequences from five normal individuals (Figure 3) with ages ranging from 22 to 74 showed RACK7 deletions centered in and around the (TCCC)9 element resembling those that were seen in the high passage PC3 cell line previously studied (2, 3). TCC to TTC or TAC mutations were also detected downstream within the (TCCC)9 region. While sequences flanking the region of extreme GC skew matched the reference sequence (GRCh38.p12 Primary Assembly), none of the observed sequences matched the reference sequence within the region of GC skew. Deletions were also present in specimens from patients with a diagnosis of cancer (Figure 4). In previous work with the PC3 cell line (TCCC)n i-motifs were found to form most efficiently when n≥7 (3). DNA from cancer patient specimens showed that deletions occurring in (TCCC)n elements with different lengths that were found in other genes also followed this same pattern with deletions formed when (TCCC)n element lengths were n≥7 (Figure 5).
RACK7 sequences obtained from patients without a diagnosis of cancer. (A) DNA from the bone marrow of a healthy 22-year-old bone marrow donor. (B) DNA from peripheral blood B-Cells from a healthy 41-year-old bone marrow donor. (C) DNA from peripheral blood B-cells from a healthy 60-year-old bone marrow donor. (D) DNA from a 69-year-old prostate patient with a negative biopsy for prostate cancer with high-grade prostatic intraepithelial neoplasia (HGPIN). (E) DNA from a 74-year-old prostate patient with a negative biopsy for prostate cancer with proliferative inflammatory atrophy (PIA). Sequences from the GC-Skew region of each cloned representative are shown for clarity. Regions flanking the region of GC-skew match the reference sequence (*) given in GRCh38.p12. However, in the region of the concatemer [(CCTG)8-CC-(TCCC)9-(TTCC)9], none of the sequences from these patients matches the reference sequence (*) for RACK7 from Human chromosome 20, GRCh38.p12 primary assembly.
RACK7 sequences obtained from cancer patients. (A) DNA from a 53-year-old prostate cancer patient with a Gleason Score 7 prostate cancer. (B) DNA from a 68-year-old prostate patient with a Gleason Score 8 prostate cancer. (C) DNA from a 68-year-old prostate patient with a Gleason Score 8 prostate cancer. Sequences from the GC-Skew region of each cloned representative are shown for clarity. Regions flanking the region of GC-skew match the reference sequence given in GRCh38.p12. However, none of the sequences from these patients match the reference sequence for RACK7 from Human chromosome 20, GRCh38.p12 primary assembly in the region of the concatemer [(CCTG)n-CC-(TCCC)n-(TTCC)n]. Although deletions characterize the vast majority of mutations at RACK7, the Patient in panel C had a short insertion in the region as well.
Length dependent deletions at (TCCC)n in cancer patient DNA. DNA from a 68-year-old prostate patient with a Gleason Score 8 prostate cancer shows deletions at (TCCC)n elements where n≥7. The BCR region of chromosome 22 (n=7) and the PLA2G4C gene from chromosome 19 (n=15), as well as the RACK7 gene of chromosome 20 (n=9) all show deletions in this patient DNA. Note that he (TCCC)5 region of RACK7 does not show deletions.
Among the three normal bone marrow donor specimens, both the 22-year-old and the 41-year-old profiles could be interpreted as allelic differences in those patients as opposed to deletions, however the profile for the specimens from the 60-year-old donor and the 69-year-old and 74-year-old cancer negative biopsy patients clearly show that deletions are present. On balance, the widespread occurrence of these spontaneous deletions in DNA in specimens from older healthy individuals and older patients with or without a diagnosis of cancer (Figure 4 and Figure 5), suggests that deletions may be a biomarker of biological age rather than a direct biomarker of cancer. Consistent with this possibility, the normal prostate biopsy from one of the 69-year-old prostate patient with a negative biopsy for prostate cancer (Figure 3E) showed proliferative inflammatory atrophy (PIA) which in many ways resembles the senescence associated secretory phenotype (SASP) seen to accumulate with age in the cells of normal individuals (22). SASP cells are the result of the partial failure of apoptosis in heavily damaged senescent cells (22).
Genomic frequencies of dynamic sequences that can produce mutations or replication stress. As noted in the introduction, our previous work suggests the sequences that are capable of i-motif formation can produce replication stress and promote the formation of spontaneous deletions. As such they fall into a group of dynamic sequences that are capable of forming hairpin, triple helix, and G-quadruplex non-B DNA structures, each of which has been shown to induce DNA damage in the form of deletions and expansions, in human DNA. Since these sequences are present in all mammalian genomes, the strong prediction that they are linked to species specific rates of ageing is that their frequency would be inversely related to measures of lifespan. We tested this prediction by examining the frequencies of several sequences known to be capable of non-B structure formation beginning with the (TCCC)n sequence associated with deletions in human patient specimens (Figure 6).
Frequencies of (TCCC)6 and (TGG)6 elements in mammalian genomes. The total count of (TCCC)6 elements and the sequenced genome size were determined using a program written in Mathematica. (A) Frequencies determined as the ratio of the total (TCCC)6 count to the sequenced genome size are plotted vs. the Maximum lifespan of each of the 74 mammalian genomes in the study. Many of the mammals in the full data set were from the same phylogenetic family. (B) This graph confines the (TCCC)6 data to a single member of each of 46 mammalian families. (C) Frequencies determined as the ratio of the total (TGG)6 count to the sequenced genome size are plotted vs. the Maximum lifespan of each mammal in the study. (D) This graph confines the (TGG)6 data to a single member of each of 46 mammalian families. Common names for several of the mammals studied are given at their respective lifespans above panels A and B.
As can be seen from Figure 6A, mammals with lifespans shorter than about 10 years can have much higher frequencies of the (TCCC)6 element than those with lifespans greater than about 10 years. In short, longer-lived mammals maintain these sequences at lower frequencies. This general finding remains if we confine the analysis to a smaller data set that contains only a single member of each of 46 of the known mammalian families (Figure 6B).
The frequency data for the (TGG)6 element shows this same pattern when the full data set is examined (Figure 6C) and the pattern is largely repeated in the data set for 46 mammalian families (Figure 6D). Here again, mammals with lifespans less than 10 years can have higher frequencies of the element than those with longer lifespans. Single-stranded sequences containing iterations of this element have been shown to be capable of forming a G-Quadruplex in which the two deoxyguanosine residues participate in G-tetrads in a two-step G-quadruplex with the thymidine paired with themselves joining parallel strands in the quadruplex (23). This structure was also found to be capable of blocking DNA synthesis in vitro (23). Little is known about potential functions of this element, but expansion mutations beyond the generally observed length of (TGG)7-12 to (TGG)>60 have been observed in human genomic DNA (24) presumably caused by structure-induced slippage.
Both (GT)10 and (CG)10 can form self-complementary hairpins or foldbacks, and under conditions of high salt or negative supercoiling their Watson-Crick paired duplexes can form left-handed Z-DNA (25, 26). Here again, the frequency data for the (GT)10 element shows this same pattern when the full data 74-member dataset is examined (Figure 7A) and the pattern is largely repeated in the data set for 46 mammalian families (Figure 7B). However, the frequencies are much higher than those of (TCCC)6 and (TGG)6 over the whole range of lifespans. Further, mammals with lifespans less than about 30 years can have higher frequencies of the element than those with longer lifespans. Similar results were obtained for the (CG)10 element (Figure 7C and D), except that mammals with lifespans less than about 10 years have higher frequencies of the element than those with longer lifespans. Further, frequencies observed for (CG)10 were very much lower overall than any of the sequences studied suggesting that it may be more mutagenic than the other sequences studied and that the sequence participates in required functions at some locations.
Frequencies of (GT)10 and (CG)10 elements in mammalian genomes. The total count of (GT)10 and (CG)10 elements and the sequenced genome size were determined using a program written in Mathematica. (A) Frequencies determined as the ratio of the total (GT)10 count to the sequenced genome size are plotted vs. the Maximum lifespan of each mammal in the study. Many of the mammals in the full data set were from the same phylogenetic family. (B) The graph confines the (GT)10 data to a single member of each of 46 mammalian families. (C) Frequencies determined as the ratio of the total (CG)10 count to the sequenced genome size are plotted vs. the Maximum lifespan of each mammal in the study. (D) This graph confines the (CG)10 data to a single member of each of 46 mammalian families.
The (GA)10 repetitive sequence presents another form of structurally dynamic sequence. (GA)n repeats, that are known to form triplex structures (27). Like many other dynamic sequences mutations at these sequences have been associated with human disease (28, 29). Here, mammals with lifespans less than about 30 years have elevated levels of (GA)10 repetitive sequences (Figure 8A and B).
Frequencies of (GA)10 and (GGA)6 elements in mammalian genomes. The total count of (GA)10 and (GGA)6 elements and the sequenced genome size were determined using a program written in Mathematica. (A) Frequencies determined as the ratio of the total (GA)10 count to the sequenced genome size are plotted vs. the Maximum lifespan of each mammal in the study. Many of the mammals in the full data set were from the same phylogenetic family. (B) This graph confines the (GA)10 data to a single member of each of 46 mammalian families. (C) Frequencies determined as the ratio of the total (GGA)6 count to the sequenced genome size are plotted vs. the Maximum lifespan of each mammal in the study. (D) The graph confines the (GGA)6 data to a single member of each of 46 mammalian families.
Triple helix formation has also been reported for the (GGA)n/(TCC)n sequence targeted by (GGA)n oligodeoxynucleotides when n=4 (30). However, the (AGG)20 sequence studied by Usdin (23) blocks DNA synthesis in vitro in the presence of K+ suggesting that it can form a G-quadruplex like structure similar to that suggested for (TGG)n (23). Since the (AGG)20 sequence contains 19 copies of the GGA sequence, we can infer that the (GGA)19 can block DNA synthesis by forming a similar structure in the presence of potassium ion. As noted in the introduction, blocks of this kind are potentially mutagenic. Structure induced replication stress for (GGA)10 elements is consistent with the data in Figure 8C and Figure 8D where it is seen that mammals with maximum life-spans less than about 30 years have much higher genomic frequencies of these sequences than mammals with longer maximum lifespans.
Discussion
It is important to recognize that it is difficult to make a clear phylogenetic argument about the evolution of the differences observed in the data given that biomarkers of longevity can change within a single species due to differences environmental factors affecting mortality (31). In this regard, we noted that the reported maximum lifespans of the Dingo and domestic dogs differ significantly (Table III). This form of relatively rapid alteration in life span within a species would be consistent with the relatively high meiotic mutation rates for step mutations in repeated elements measured in human DNA (32, 33). Even so, the data presented here suggests that the intrinsic somatic mutability of dynamic DNA sequences may provide a form of genome plasticity that may contribute to the evolution of differences in lifespan.
It is of particular interest that the (TCCC)9 element in RACK7 and the (TCCC)15 element in Phospholipase A2 GroupIVC (PLA2G4C) are both susceptible to i-motif induced deletions and preserved in the long-lived human genome (3). Both of these dynamic sequences are confined to introns in humans where the effects of deletions on gene expression could be diminished. In this regard, the accumulation of extensive deletions in introns like that found at PLA2G4C (Figure 5) could have deleterious effects on chromosome pairing in aging humans.
Our data do not address the placement of the repeated sequences in the various species studied here, but in certain cases they will reside in sequences with essential functions. For example, i-motif formation within the BCL2 promoter has been implicated in the control of transcription at that locus (34). Another well characterized example is found at the promoter of the mouse c-Ki-ras protooncogene. There, the region containing a (TCCC)7 element has been shown to be sensitive to S1 nuclease (35), suggesting that is adopts a non-B structure. Further, deletion of a (TCCC)7 element impairs transcription at that locus (35). Mutable but essential sequences of this type may be retained in long lived mammals by antagonistic pleiotropy (36). From this point of view, the presence of a dynamic sequence in a tumor suppressor or essential gene in a metabolic pathway would not result in significant tumor promotion or metabolic disfunction until after sexual maturity had been reached in a long-lived mammal so long as the total number of such sequences was low and normal repair functionality was maintained.
One caveat in interpreting the data is the known tendency of next generation sequencing assembly methods to underestimate the frequency of repetitive sequences (37, 38). Many genomes available from the NCBI FTP site might suffer from this problem. In the most often cited example of this problem, two human genomes assembled after next generation sequencing (39) were found to lack 16% of the repetitive elements present in the human genome reference sequence (38). Those sequences were published with a scaffold N50 values of 44300 (39), while 80% of the genomes studied here reported primary assembly N50 values 3618479 or more than 81 times greater than the data in (39). Given the unlikely possibility that the underestimation of repetitive sequence frequencies would systematically affect only genomes from longer lived organisms, the data suggests that short lived mammals are expected to accumulate DNA damage at higher rates than long lived mammals due to the inborn differences in total frequencies of dynamic DNA sequences. This suggestion is consistent with the general similarity in the efficiency of repair systems in mammals since those similar systems would be tasked with repairing quantitatively more damage in short lived mammals than in long-lived mammals. This contribution to aging is supported, in the particular cases of the G-quadruplex and i-motif forming sequences where non-homologous end joining is suggested as essential to the formation of the observed deletions (3, 10) by data showing that deficiency of DNA ligase 4 (Lig4) causes progressive loss of marrow stem cells in mouse (40).
Study limitations. First, the limited sample size of the human data we present is not adequate for a conclusion that mutations at dynamic sequences accumulate with age or that they predispose to cancer. We include that data only to clarify the reasoning that led to the genome searches which provide stronger evidence for the original hypothesis described in (41).
Second, in addition to the caveats noted in the discussion above, it is important to point out that different counting algorithms may yield different total counts for the sequences tested. StringCount in Mathematica scores blocks of each string in long tandem repeats by changing register by an entire block after each block is scored this giving a count of the total number of adjacent blocks in a long repeating unit. BLAST on the other hand estimates the number of blocks by moving the register by one base pair yielding a larger number that gives something akin to the number of ways the block can fit into a long repeating unit. We feel that qualitatively the trends in the data will be preserved by various counting algorithms.
Third, it must be noted that data in support of the mutagenetic potential of several of the repeated elements we studied in mammalian genomes relies on in vitro replication studies.
Conclusion
Our data on the occurrence of i-motif linked RACK7 deletions in human specimens shows that they are not a specific biomarker of cancer since the deletions are found in patients without a diagnosis of cancer and in healthy bone marrow donors. Nevertheless, our genomic search results demonstrate an inverse correlation between the frequencies of six dynamic sequences and maximum lifespan in 74 mammalian genomes studied. Each of the sequences studied, including the (TCCC)n element of RACK7 is capable of mutagenesis and/or replication stress. These results suggest that the inborn frequency of dynamic sequences may set species-specific rates at which mammals undergo genomic decay and aging. As such, they offer further support for the proposal (41) that “one means by which longer cancer free lifespans may have been achieved is by the elimination of troublesome DNA sequence motifs”.
Acknowledgements
This work was supported by the Biotechnology and Biological Sciences Research Council (BB/L02229X/1), by a grant 5R01-CA102521 to S.S.S. from the U.S. National Cancer Institute of the National Institutes of Health. Research reported in this publication also included work performed in the Integrative Genomics and Bioinformatics Core and the Cytogenetics Core supported by the National Cancer Institute of the National Institutes of Health under award number P30CA033572. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
Conflicts of Interest
The Authors declare that they have no conflicts of interest.
Authors’ Contributions
M. M. conducted experiments; F.P., N.C, and Z. W. helped with the design of experiments; IRB protocols were written by S.S. and G.M. S.S. designed and conducted experiments and searches and wrote the paper with the help of Z.W and N.C.
- Received January 14, 2024.
- Revision received March 2, 2024.
- Accepted March 7, 2024.
- Copyright © 2024 The Author(s). Published by the International Institute of Anticancer Research.
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.