The NCBI Genome Decoration page provided chromosome annotation software [46].

DNA flanking sequences at breakpoints were downloaded primarily from the GRCh38/hg38 version of the University of California, Santa Cruz Genome Browser as FASTA files and copied directly into BLAST. Results were checked against breakpoints in 101 triple-negative breast cancers from a population-based study [32]. The University of California, Santa Cruz Genome Browser’s Liftover function interconverted different versions of genome coordinates into GRCh38/hg38 coordinates.

This study presents analyses of publicly available data without recruiting additional human or animal subjects. Because this study is a secondary analysis, it is exempt from institutional review board and ethics approval. The data are in the public domain and are available for independent research and analysis [56]. It is not necessary to obtain permission to reuse public data. The original informed consent allows secondary analysis without additional permission.

EBV-mediated cancers such as NPC have defects in DNA repair and inflammatory pathways, resembling hereditary breast and ovarian cancers. To further characterize this resemblance, breakpoints in 70 NPC genomes were compared to breakpoints in 139 breast cancer genomes from high-risk women (BRCA1/BRCA2 mutation, familial concentration, or young age).

The distances from all breast cancer breakpoints to the nearest NPC breakpoints across the entire length of chromosome 1 produced results with so many points that they were difficult to interpret (Multimedia Appendix 3). Different laboratories collected these breakpoint data over many years. To allow for some variations, the data were grouped into 5000-bp increments (2×10^–5 relative error). As shown in Figure 1 and Multimedia Appendix 2, breast cancer breakpoints were most often clustered within 5000 bp of NPC breakpoints, but many breakpoints agreed much more closely. A total of 20 breast cancer breakpoints on chromosome 1 were within 500 bp of an NPC breakpoint, and several chromosomes had breast cancer and NPC breakpoints in essentially the same positions. As represented by Mann-Whitney U test results (Multimedia Appendix 2), breast cancer and NPC breakpoint distributions were statistically the same (P>.05) for chromosomes 6, 7, 10, 13, 14, 15, 22, and X, but different on chromosome 1 and other chromosomes (P<.05).

In contrast, liver cancer breakpoints at hepatitis B virus integration sites [57] differed from those in breast cancer or NPC (Figure 1). No breaks in 114 liver cancers on chromosome 1 were within 5000 bp of breaks in any NPC; only one break on chromosome 6 in 61 liver cancers fit this window. According to a meta-analysis, the chance that breakpoints on chromosomes 1, 2, 6, and 8 were not within 5000 bp in liver cancer versus NPC was 4.4 (95% CI 1.9‐10). NPC and liver cancer did not have the same breakpoint distributions (P<.001).

The above results revealed that breast cancer breakpoints in high-risk women were clustered near those in the EBV-associated cancer, NPC, on every chromosome. The next step was to decide whether these similarities depended on mutations in the breast cancer susceptibility genes, BRCA1 or BRCA2, by comparisons to sporadic breast cancers. The sporadic breast cancer group comprised 74 women, aged ≥70 years, with normal BRCA genes and no other known inherited, cancer-associated mutations [31]. Like breakpoints from high-risk women, many sporadic breast cancer breakpoints clustered around those in NPC (Figure 1). Breakpoints in these sporadic breast cancers clustered at chromosomal locations similar to breast cancers from high-risk women, although the frequencies and distributions sometimes differed significantly. The patients with sporadic breast cancer were older than the high-risk women, arguing against age as responsible for similarity to NPC breakpoints.

Ovarian cancer data enabled an additional test for EBV involvement in breast cancer because, like breast cancer, BRCA1 or BRCA2 mutations can also predispose patients to ovarian cancer [58]. Chromosome breakpoints in 25 mixed adenosquamous ovarian cancers were compared to breakpoints in NPCs. The results depicted in Figure 1 emulated breast cancer comparisons. Nearly half (12/25, 48%) the ovarian cancer cases had likely hereditary BRCA mutations. The remaining sporadic ovarian cancers gave the same results as the complete set but with less data. As in breast cancer, ovarian cancer breakpoint distributions clustered around NPC breakpoints, even without a hereditary BRCA1 or BRCA2 gene mutation driver.

EBV drives lymphomas as well as NPCs. Based on epidemiologic research results, FA-BRCA pathways protect against lymphomas [59,60]. If EBV is genuinely associated with breast cancer breakpoints, then breakpoint positions in EBV-mediated lymphomas should also resemble those of breast and ovarian cancers. Because MYC gene rearrangements are characteristic of EBV-associated lymphomas, the first test of this idea was to survey virus-like sequences surrounding the MYC gene locus on the human reference genome. Figure 2 shows that MYC resides in a literal forest of retrovirus sequences (eg, human immunodeficiency virus type 1 [HIV1], feline leukemia virus, porcine endogenous retrovirus, and human endogenous retrovirus [HERV]) interspersed with EBV-like sequences.

The concentration of virus sequences around MYC on chromosome 8 prompted the addition of the EBV-associated lymphoma DLBCL to breakpoint comparisons. As shown in Figure 2, the results revealed that hundreds of breast cancer and NPC breakpoints congregated around breakpoint positions in 88 DLBCLs [22]. This agreement was consistent with other similarities between breast cancers and these EBV-associated cancers, including deficits in FA-BRCA pathway–mediated DNA repair by homologous recombination [61] and NF-κB activation [11,62-64].

EBV is also a proven driver of at least one subset of BLs, typically those with MYC translocations. BL subsets can have mutations that impair homologous recombination [65], so results in Figure 2 revealed many breast cancer breakpoint positions near corresponding BL breakpoints. An older dataset from BLs [21] had translocation breakpoints in the virus sequence–rich area near the MYC locus, agreeing with about 140 breast cancer breakpoints. Four different NPC breakpoints produced over 100 matches to BL translocation breakpoints, beginning at 8250 bp apart. An unpaired, 2-tailed t test did not support a statistically significant difference between BL and NPC breakpoints in this area (P=.69).

Further tests were conducted to determine whether the functions of genes near clustered breakpoints supported a relationship between breast cancers and EBV-related cancers (GC [41], BL, and NPC). As illustrated in Figure 3, breast cancer breakpoints on chromosomes 6, 8, 11, and 17 aggregated near positions where breakpoints occurred in EBV-associated cancers. Many aggregated breakpoints were in the same areas as genes that control inflammation, antiviral defenses, apoptosis, intermediate filaments, epigenetic and chromatin regulation, estrogen receptor activity, mitotic structures, and mitotic controls (Table S1 in Multimedia Appendix 4). Breast cancer breakpoints that clustered around EBV-associated cancer breakpoints were especially numerous on chromosome 17. One of these clusters marked in Figure 3 included the HER2 amplicon and the topoisomerase 2a gene, with BRCA1 and SMARCE1 genes nearby. SMARCE1 encodes a part of a chromatin regulation complex. Chromosome 17 breakpoints near CNTROB and CTC1 genes connect EBV to centriole and telomere malfunctions during mitosis (Table S1 in Multimedia Appendix 4). Rearrangements near breakpoints may cause over- or underexpression of nearby genes (Table S2 in Multimedia Appendix 4). Many additional correlations were also likely revealed in Figure 3 but were not investigated further.

Results in this section show that breast cancer breakpoints clustered around breakpoints in additional EBV-associated cancers, where they affect critical functions needed to prevent breast cancer. Once these functions are compromised, cancer can occur without the continuing presence of EBV.

In preceding sections, breast and ovarian cancer breakpoints were found to distribute most frequently near characteristic sets of breakpoints associated with EBV-related cancers. The virus first attaches its EBNA1 protein to human DNA in the nucleus. Then, circular EBV episomes dock to this attached EBNA1 anchor. To test whether the initial EBNA1 attachment sites were related to breast cancer chromosome breakpoints, breast cancer breakpoints were compared to genes near EBV-docking sites. EBV-positive BL cells providing the data had up to 1569 EBV-docking sites on all chromosomes identified by 4C-chromatin capture experiments [47]. As shown in Figure 4A, the largest numbers of breast cancer breakpoints on most chromosomes clustered around the genes [47] nearest to genes at EBV-docking sites. In support of these comparisons, graphical estimation of virus-tethering sites on chromosome 2 from chromatin capture data for these EBV-positive cells also agreed with breast cancer breakpoints (Figure 4A). In an unrelated study [48], EBV-docking sites on chromosome 11 near known EBV anchor sites at the FAM-D and FAM-B genes were found near groups of breast cancer breakpoints, but imperfect palindrome sequences [66] were more distant (Figure 4B). This finding independently supports the idea that EBV-docking sites are near breast cancer breakpoints. Results in this section raise the possibility that EBV directly contributes to breast cancer chromosome breakpoints and fragmentation.

To further test whether EBV itself has some role in breaking chromosomes or altering their structures, human chromosomes were compared to all known viruses. As shown in Figure 5A, the results showed that nearly every breast cancer likely had undergone breakages near EBV-like sequences. Chromosome 8 alone had 59,566 significant (>200) viral homology scores. Based on data from 128 patients with breast cancers and 43,491 unique breakpoints, breakpoints in 123 (96.1%) out of 128 breast cancers were within 10,000 bp of a virus sequence. In 106 patients, the virus was an EBV tumor variant (HKHD40 or HKNPC60) with 3086 matching human sequences. According to the Fisher exact test, chromosome 8 breakpoints and EBV variant sequence matches were not independent (P<.001).

Many areas on other chromosomes also had 97% human-virus identity over nearly 2500 bp. It is implausible that this much similarity comes from EBV DNA being carried over into the human reference genome. Viral homology occurred with only a small, select portion of viral DNA [68]. Viral homologies were determined for a human genome in a homozygous karyotype, haploid cell line (46,XX) hydatidiform mole derived only from the paternal chromosomes in an X-bearing sperm cell after fertilization [69]. Results still showed extensive homology between the mole and EBV variants HKHD40 and HKNPC60 (Figure 5B).

HKHD40 and HKNPC60 variant sequences kept appearing in comparisons to human sequences, so these variants were tested against other herpesviruses to determine whether they were unusual. Hundreds of human gamma herpesvirus 4 variants were almost identical to HKHD40 and HKNPC60 over at least 2000 bp (Figure 5C). The matching sets of viruses included many high-risk herpesvirus isolates from NPCs [67]. Based on this information, HKHD40 and HKNPC60 strongly resembled other herpesvirus isolates, including many that confer high risks for NPC [10]. These results show that humans have interacted extensively with EBV; the results are not due to EBV impurities in the human reference genome, and the human genome has had close relationships with oncogenic EBV forms.

The evidence thus far supports a central hypothesis that EBV disables tumor suppressor mechanisms in breast cancer and can then disappear. This absence of viral particles is a significant experimental obstacle to testing this hypothesis. Unlike retroviruses, EBV and its variants do not have integrase enzymes, so EBV has no conventional way to insert itself into the human genome. EBV rarely integrates, with only one or two copies in BL cell lines [70].

BLAST analysis found about 65,000 areas of strong homology (E<1×10^–10) between the human reference genome and EBV. Because 65,000 is far more than realistic EBV integration events, it suggested the possibility that some EBV sequences were fragments created by a human version of the bacterial CRISPR (clustered regularly interspaced short palindromic repeats) system. As shown previously in Figure 3, breast cancers have breakpoints that cluster around breakpoints in EBV-associated cancers and involve MHC genes.

MHC genes are encoded on chromosome 6p21.3 in a region that becomes a candidate for such a human CRISPR version. Variants of human leukocyte antigens (HLAs) in the MHC are strong risk factors for NPC infections [71] because HLAs are required to break down and display fragments of some antigens to the immune system. A total of 13 breast cancers listed on the COSMIC website had a deletion near this HLA region. About 23% of breast cancers had mutations directly affecting HLA class I or II genes. Many more breast cancers had indirect connections because they had damage to multiple genes that interact with HLAs or were otherwise essential for immunity. The MHC region also holds NFKB1L1, a negative regulator of the NPC overexpressed gene hallmark, NF-κB. The 139 breast cancers from high-risk women had 284 breakpoints at chromosome 6p21.3. Breakpoints in the 70 NPC cancers also clustered there, with 40 breakpoints within the 27,865,296-34,017,013 segment on chromosome 6. Variability in the inactivation of MHC genes reflects the extreme diversity of this region.

In general, the bacterial CRISPR/Cas system loosely resembles the human piRNA system, so the distribution of piRNAs was graphed. As shown in Figure 6A, hundreds of piRNA sequences cluster near the MHC region (at ~29.7‐33.3 megabases). The piRNA system is known to inactivate virus-derived transposons (related to HERVs) by methylating or cleaving them. The distribution of piRNA fragments was then compared to the distribution of viral DNA fragments in the MHC region of chromosome 6. Figure 6B-F reveals striking similarities in how remnants of exogenous and endogenous viruses distribute relative to piRNAs. Remnants of both virus types were homologous to the same human sequence, and both types were interspaced between piRNA sequences, sometimes right next to each other. Most of these sandwiches were at a regular interval or a multiple of a regular interval.

This interspaced arrangement looked so much like CRISPR that it raised the question of whether piRNA defense mechanisms have inactivated some EBV variants in addition to their canonical role with endogenous viruses. Long stretches of endogenous transposon-like DNA sequences routinely matched exogenous viruses. As shown in Figure 6C and E, the same human DNA interval had homology both to endogenous transposons (HERV) and exogenous viral sequences (EBV variants, stealth virus 1, chikungunya virus, BeAn 58058 virus, human papillomavirus [HPV] 16, HIV1, and HERV). This result shows that the piRNA system can store the same piece of DNA to protect DNA against these different viruses.

Chromosome 6p21.3 also contains an EBV infection marker [72]. The marker was examined in 1538 breast cancers using existing methylation data [34]. As indicated in Figure 7, promoter methylation differed significantly from normal controls in the segment shown (30,523,984‐33,216,811 on chromosome 6). Hypermethylation occurred on STK19, a MHC class III gene for RNA surveillance [73,74]. Hypermethylation also occurred on a gene for preventing tumors (TNFB) [75] and a gene for responding to antigen-antibody complexes (C2). Polymorphisms in HLA-DMB antigen and SAPCD1, another class III MHC gene [76], at chromosome 6p21.3 had links to Kaposi sarcoma [77]. Human herpesvirus 8 (Kaposi sarcoma virus) is a Kaposi sarcoma driver and is closely related to EBV.

These results reveal that EBV has been attacking human DNA during evolution. There is a piRNA defense mechanism for human DNA near critical immune system genes, but both EBV-associated cancers and breast cancers inactivate some of the genes that guard piRNA defenses. The histocompatibility antigen gene region of chromosome 6 can be extensively fragmented in EBV-associated and breast cancers. MHC genes have the largest number of polymorphic forms in the human genome. This variation creates differences in viral susceptibility and inactivation. Even though most people are infected, not everyone will get an EBV-related disease or cancer.

The next question was whether EBV or other virus-like sequences in the human genome cause multiple rearrangements and clustered hypermutations (chromothripsis). As shown in Figure 8A, many positions on chromosome 6 where chromothripsis occurs [35] congregated around virus sequence start positions. A total of 1090 genome coordinates described chromothripsis fragments with copy number ≥3. These coordinates were unlikely to be random since they did not follow a normal distribution (P<.001). By simple linear regression analysis (R²=0.93), many viral sequence coordinates strongly correlated with chromothripsis positions. Figure 8B shows that as you move further away from a chromothripsis breakpoint, the frequency of breast cancer homology (score>500) to viruses decreases. This result implicates viral sequences as preferred sites where breast cancer chromosomes begin to fall apart. The equation shown mathematically describes the relationship between chromothripsis frequency and distance from viral sequences, and the constant in the equation suggests a baseline level of breakpoints.

These results suggest that homologous virus sequences at multiple positions could confuse DNA repairs already compromised by EBV in breast cancer and contribute to chromothripsis and clustered rearrangements.

The last question was whether EBV contributes to breast cancer metastasis. According to Yates et al [29], relapsed and metastatic breast cancer tumors keep their tumor-driver gene mutations and continue acquiring new ones. Late mutations in JAK-STAT and SWI-SNF signaling pathways drive established breast cancers into metastasis.

NPC often loses type-1 interferon genes (IFNA1, IFNA2, IFNA8, and IFNE) and nearby MTAP (32%-34% [11]) by homozygous deletions at chromosome 9p21.3. Interferons initiate canonical JAK-STAT signaling by binding to cell surface receptors that then activate internal Janus kinases (JAKs). The activated JAKs phosphorylate cytoplasmic STAT (signal transducer and activator of transcription) proteins, which travel to the cell nucleus to activate interferon-responsive genes. The percentages of breast cancers on the COSMIC website with mutations in a “JAK” or “STAT” isoform or transcript variant were calculated: 7.8% had a JAK mutation and 36.7% had a STAT mutation. Deletions of interferon genes in NPC also facilitate viral replication and block interferon from activating JAK-STAT signaling. Breast cancers (Multimedia Appendix 2) have 65 breakpoints strictly within this interferon-MTAP region (21,579,478‐20,503,534 on chromosome 9), not counting longer fragments that include the interval. As shown in Figure 9, breast cancer breakpoints align well with EBV-associated cancer breakpoints near the large cluster of interferon genes on chromosome 9.

Mutations in EBV-associated cancers show that Yates metastasis driver gene damage accompanies EBV infection. SWI-SNF (switch/sucrose non-fermentable) is a complex that repositions nucleosomes and supports genome stability [79]. SWI-SNF addresses obstacles to replication sensed by the FA-BRCA pathway [79,80]. Referring back to Figure 3, clustered breast cancer breakpoints on chromosome 17 around EBV breakpoints affect the SWI-SNF component SMARCE1. In addition, breast cancers can disable SWI-SNF by targeting ARID genes [29]. ARID1A is a COSMIC top-20 most frequently mutated gene in breast cancer. Like breast cancer, NPC has multiple recurrent aberrations in ARID1A genes. As shown in Figure 9, ARID1A lies near a hot spot where multiple breast cancer breakpoints approximately aligned with breakage points in EBV-associated cancers. The loss of ARID1A activates Annexin A1, which aligned closely with a region targeted by EBV-associated cancers on chromosome 9. A chromosome-1 site at about 150 million bp had a nearby histone-rich gene region. Histones are chromatin structures that SWI-SNF dynamically remodels to regulate access to genetic information. Histones can profoundly affect metastasis [78]. Figure 9 also reveals many additional alignments between breakpoints in breast and EBV-associated cancers that were not investigated further.

NPC often inactivates SWI-SNF components BAP1 and PBRM1 within a frequently damaged 3p21.3 gene cluster [11] at 52,400,000‐53,000,000 on chromosome 3. Analyses of breast cancers found 18 breakpoints within this short interval. DLBCL, another EBV-linked cancer, also had recurrent alterations in components of SWI-SNF complexes [81].

The Warburg effect (oxidative glycolysis) [68] favors metastasis. The Warburg effect occurs in NPC because pyruvate dehydrogenase (PDHB) genes on chromosome 3p are deleted or rearranged in almost all cases. Similar changes to chromosome 3p were found in breast cancers, which also undergo the Warburg effect [68]. This Warburg metabolic switch favors metastasis because it mitigates oxidative stress on cancer cells. Large amounts of lactate accumulate in the absence of PDHB to acidify the tumor microenvironment and interfere with the destruction of metastatic cells [82].

This section’s results show that EBV may push breast cancer into metastasis by interfering with JAK-STAT and SWI-SNF signaling pathways to facilitate viral replication while making the microenvironment more favorable to tumor growth.

This study finds that EBV contributes to breast cancer by disabling safeguards against tumors. Cancer then occurs because the safeguards remain disabled even if the virus is cleared. Multiple independent analyses identified residual genetic and epigenetic damage in cancer genomes and formed the basis of the model in Figure 10. Breakpoints in breast cancers in high-risk women, sporadic breast cancers, and even ovarian cancers cluster around breakpoints in known EBV-related cancers, including NPC, BL, DLBCL, and GC. Some genes clustered near breakpoints in these diverse EBV-associated cancers are critical to preventing breast cancers. Some breast cancer breakpoints are near genes at EBV-docking sites. Varying numbers of DNA breaks occur within the highly polymorphic forms of MHC region genes on chromosome 6. This damage adds to susceptible polymorphisms and immunodeficiencies to help explain why not everyone develops EBV-related cancers. Near the MHC region on chromosome 6, piRNA sequences are regularly interspaced between viral DNA sequences. The sandwiched arrangements are presumptive evidence of past infection and probably represent a DNA defense mechanism. These defenses fail when chromosome 6 breaks apart near start points of the large number of repetitive viral sequences in the human genome. The viral sequences confuse repairs already damaged by EBV, and bursts of mutation occur where scrambled fragments ligate. EBV disables the most reliable restoration of broken chromosomes back to their native forms, so repairs form structures with multiple centromeres. These structures undergo additional rounds of fragmentation during cell division. The process continually forms new cancer driver mutations and allows cancer to come back after successful therapy (Figure 10). An EBV methylation signature on chromosome 6 was far more abundant in 1538 breast cancers than in normal controls. Finally, EBV facilitates its own replication by damaging JAK-STAT and SWI-SNF signaling pathways, which pushes breast cancer into metastasis, while virus-associated changes on chromosome 3p interfere with the destruction of metastatic cells. Models [8,98] of EBV-infected human mammary cell cultures transplanted into immunosuppressed mice and EBV loss from NPC cells are consistent with these results.

The study herein has current and future clinical implications in addressing cancers and chronic diseases. An early childhood vaccine against EBV may reduce the incidence of breast cancer on a global scale. If this vaccine even approaches the effectiveness of the HPV vaccine for cervical cancer, then the reduction of breast cancer incidence would be substantial. In breast cancer cases where active infection can be demonstrated, immunotherapy or antivirals can be considered. The results also heighten concern about hidden dangers from viral infections. EBV infection leaves behind persistent genome abnormalities (“long EBV”) linked to breast cancer. Not everyone develops an EBV-related cancer even though almost everyone is infected, suggesting risk assessment should include MHC polymorphisms. MHC genes have abundant connections to both EBV infection [99] and breast cancer [100-102].

The strategy of using bioinformatics to identify markers of “long EBV” may well work for other cancers, multiple sclerosis [103], and other chronic diseases that are currently unexplained. Testing for persistent viral damage in genomes from biopsies is a new method for screening for breast cancer risk. The results may inform further prevention and treatment decisions. Cancer drug therapy has focused on finding and destroying cancer-driver gene products. The drugs are initially effective, sometimes for long periods, but then stop working. The cycles represented in Figure 10 are an occult, underlying process that can now be evaluated. Cancer treatment generates new clones that do not exist in the original population [104]. The underlying genome damage and EBV scars continually produce new cancer-driver mutations. Some antigens targeted by successful therapy for hematologic malignancies [105], such as DLBCL, may also be effective for breast cancers. The idea that breast cancers and hematologic malignancies can have similar breakpoints and translocation fusions suggests that there may be many more susceptible targets and that there are options to overcome resistance or tolerance [106]. The findings may further stimulate research into other EBV-associated diseases and cancers, leading to better and broader understanding.

Estrogen has been thought to generate the initial chromosome breakpoints leading to translocations in human breast cancer. However, young boys with BL do not produce estrogen from ovaries, yet Figure 3 shows that their malignant B-cells have many breakpoints [68,107] that approximately match breast cancer breakpoints. Normally, aromatase catalyzes the rate-limiting step in estrogen production [108], and aromatase acting on androgens is the primary source of most estrogens in breast tissue [109]. EBV-infected cells lose control of aromatase activity [108]. An EBV-mediated increase in aromatase activity explains why locations of breakpoints (Multimedia Appendix 5) are relatively independent of estrogen receptor status in breast cancer [68] and resemble locations in lymphoid cells (Figures 1-4 and 9). Transcription in response to artificially high estrogen levels created by EBV then induces topoisomerase-mediated DNA breaks. Double-strand break repair genes remove topoisomerase from these complexes, but damage to this process leaves pathological enzyme complexes still bound at a DNA breakpoint [110-112]. As shown in Figure 3, topoisomerase itself may be damaged. In either case, EBV-mediated deregulation of estrogen production, topoisomerase activity, and deaminases then collaborate to cause chromosome breaks and drive breast cancer.

Breast cancer chromosome breakpoints cluster around genes near EBV-binding sites (Figure 4), further suggesting that EBV participates in causing the breaks. The breaks lead to pathogenic chromosome rearrangements because EBV-induced damage forces restoration into error-prone methods by suppressing FA-BRCA pathway intermediates [14,15]. Repairs using the FA-BRCA pathway [113] need chromatin access, which requires the SMC5/6 cohesin complex [114,115]. In one scenario shown in Figure 10, SMC5/6 interacts with a crucial pathway intermediate, the FANCD2-FANCI heterodimer (“D2-I”) [17,116]. EBV variants deplete SMC5/6, preventing FA-BRCA–mediated DNA repairs and leading to chromosomes with too many centromeres. When mitosis pulls apart multicentromere chromosome structures, the forces shatter the chromosome and induce mutation storms [35]. EBV thus threatens a sprawling, interconnected repair system, including the BRCA pathway, FA proteins, an SMC5/6 scaffold, JAK-STAT signaling, and the SWI-SNF chromatin remodeling complex (Figure 10).

Of course, other environmental, genetic, or lifestyle factors also participate in breast cancer development, but EBV infection exacerbates their effects. Genome deficits in EBV-associated cancers and breast cancers interfere with restoring chromosomes from damage due to natural processes and exogenous mutagens. Some of this damage requires repair pathways that are subject to EBV interference.

Evidence underlying the model in Figure 10 has independent support from the literature. For example, viral load is a marker for the extent of cell-free DNA fragmentation [117]. EBV-mediated transformation routinely generates abnormal karyotypes [118]. The binding of EBNA1 sequence variants increases NPC risk and drives EBV lytic gene expression [119,120], which requires EBV-encoded nucleases [121-123]. Other herpesviruses related to EBV share the ability to fragment DNA and subvert DNA repair pathways [124-126]. EBV facilitates its own replication by interfering with signaling pathways that prevent metastasis [29,127-130]. Independent literature supports EBV participation in metastasis and the results shown in Figure 9. NPC has the highest metastatic rate among all head and neck cancers, and the levels of circulating EBV markers are highly predictive [10]. Finding EBV in lymph nodes of patients with NPC or primary cancer at an unknown site helps detect metastasis [131]. NPC patients with ≥500 copies of EBV per mL plasma had significantly higher rates of liver metastasis than patients with lower EBV levels [132]. EBV-infected B-cells and breast cancer cells both have amplified centrosomes (Figure 10), the mitosis-organizing centers that exert structural control over cell division. The EBV protein thymidine kinase takes up residence in the centrosome [133], and another EBV protein, BNRF1, initiates centrosome amplification in infected B-cells [134]. Overduplication of centrosomes confuses chromatid attachments to spindle fibers during mitosis. Chromosomes do not distribute properly into daughter cells, creating mistakes when the genome replicates [134,135]. Neither centrosome amplification nor chromosome fragmentation (chromothripsis) requires large numbers of viral particles or active infection.

Further bioinformatic tests may still add significant additional information. EBV activation brings massive changes to host chromatin methylation and structure [47,51,136]. Breast cancers have hundreds of these changes [34]. Results here further implicate epigenetic effects, so EBV effects on breast cancer epigenetics should be explored in more detail. EBV is implicated in cancers in multiple additional organs, and the methods developed here may help clarify its potential contributions. Predictions based on virus-human interaction structural biology may also be helpful. The ultimate direct test will be whether childhood recipients of an anti-EBV vaccine have reduced breast cancer incidence. If it even approaches the reduction of cervical cancer achieved by the HPV vaccine (up to 94%), a childhood EBV vaccine could effectively prevent many cases of breast cancer.

EBV itself creates a limitation because the virus can disappear after causing pathogenic genome damage that allows breast cancer to develop. This transitory virus presence forces the use of bioinformatics to look for persistent genome damage EBV leaves behind. EBV disappearance questions whether a group of cancers with EBV connections also contains “sporadic” cancers typed as EBV negative. The EBV-negative forms may have merely lost the criteria used to identify EBV infection, but EBV-related genome damage may still remain. Another limitation is that compared to breast cancers, known EBV-linked cancers such as GC, BL, and NPC are less common, so genome sequence data are also less common.

In summary, early childhood immunizations against inactivated EBV or selected EBV gene products may significantly reduce the incidence of breast, ovarian, and other cancers, and potentially unexplained chronic diseases. EBV variants lead to DNA breaks, mitotic abnormalities, and the loss of safeguards that protect against breast cancer and its metastasis. Breast cancer breakpoints cluster around breakpoints in EBV cancers, disrupting genes essential to prevent viral infection and breast cancers. A CRISPR-like region on chromosome 6 sequesters some of the thousands of pieces of EBV sequences in the human genome. The same area of chromosome 6 undergoes variable damage in breast cancer, contributing to the reason not everyone with EBV infection develops cancer. In susceptible people, EBV infection leaves behind pathogenic cancer-associated genome abnormalities (“long EBV”). Clinical implications include improvements in evaluating the chances that cancer will return, increased use of immunotherapy for patients with breast cancer that have active infection, and greater urgency in developing an effective EBV vaccine,.