HPV, Barrett's Dysplasia, and Esophageal Adenocarcinoma
HPV, Barrett's Dysplasia, and Esophageal Adenocarcinoma
This study is the first to find that HPV positivity as detected by PCR was significantly higher in patients with BD and EAC compared with controls and those with Barrett's metaplasia. Most of the HPV that was amenable to genotyping was of the HR type, i.e. HPV 16 and 18, which are the predominant strains found in cervical neoplasia, oropharyngeal malignancy, and anal cancer. Transcriptionally active HPV, as estimated by p16INK4A (a surrogate marker for HR HPV oncoprotein activity) IHC, and E6/E7 mRNA expression showed a similar pattern. p16INK4A positivity was greatest in BD and EAC and much less in control and BE subjects. The strength of the association with more severe GERD spectrum disease was greatest when HPV DNA, p16INK4A, and E6/E7 mRNA were all positive compared with all being negative. Also, both the combination of E6/E7 mRNA positivity with HPV DNA and p16INK4A with HPV DNA were independently but less strongly associated with more severe GERD spectrum disease. Furthermore, we have demonstrated that HPV infections in the esophagus were focal. In both dysplastic/endoscopically resected EAC and nondysplastic patients, HR-HPV was detected more commonly at the transformation zone (SCJ) compared with the lesion in BE or corresponding position in squamous epithelium as we previously hypothesized. Interestingly, in hospital and GERD controls as well as Barrett's nondysplastic individuals, HPV infections were largely biologically irrelevant at the time of biopsy, as evidenced by minimal p16INK4A positivity (5.9%) and absence of E6/E7 mRNA expression. Conversely, in patients with BD and EAC, 44% of patients had p16INK4A staining and in 30% of patients, E6/E7 oncoprotein expression was detected. Again, in BD tissue, this transcriptional activity was more concentrated in the SCJ. We hypothesize that progressive acid damage to the esophagus increases the likelihood of mucosal breaks and allows the virus to enter the basal layer of the transformation/transition zone, i.e., the junction between the squamous and the columnar lining. However, the number of cases observed was insufficient to test this hypothesis. HPV affinity is to junctional tissue, as basal cells are particularly accessible at the squamocolumnar transformation zone and especially susceptible to this viral infection. It is known that transforming HPV infections in different epithelia (cervix, anus, and oropharynx) cause cancers mainly at the transformation zone. A discrete population of SCJ cells of the cervix with a unique morphology and gene expression profile has been implicated in the pathogenesis of cervical cancer. Intriguingly, it has also been reported that BE is derived from a discrete population of embryonic cells residing at the SCJ. Moreover, it has been demonstrated that a single cell of origin (common progenitor) could differentiate into squamous or columnar epithelium, thus exhibiting esophageal plasticity. Furthermore, using a transgenic mouse model, it has been demonstrated that BD and BE can be initiated at the SCJ.
Our results indicate a much higher HPV prevalence of 18% in controls compared with our prior expected prevalence of 9% in esophageal tissue. When we commenced this study in 2008, there were no reliable HPV prevalence data available for the Australian population. A guesstimate prevalence of 9% was obtained from the meta-analysis by de Sanjose et al. on worldwide HPV prevalence in cervical tissue of women with normal cytology. A recent study reported HPV DNA prevalence of any serotype of 41.5% and HPV 16/18 of 12.8% in cervical tissue of 2,152 Australian women aged between 17 and 40 years attending a community-based health service for routine Pap screening. In another publication, the HPV genoprevalence in women aged 18–24 years (n=202) attending family planning clinics in Sydney, Melbourne, and Perth during the pre-vaccine period of 2005–2007 has been reported to be 21.3% for HPV 16 and 8.4% for HPV 18. Hence, the above viral prevalence rates in Australian women are comparable to our control group prevalence of HR-HPV. However, none of these studies are directly relevant, as none involved observations of esophageal HPV prevalence.
In a retrospective study from Mexico, HPV DNA, mostly low-risk types, was detected by in-house PCR in 96% of paraffin-embedded tissue biopsies of BE (without endoscopic or histopathological classification) and 88% of undefined esophageal cancers. Nevertheless, Rai et al. reported that HPV infection in fresh BE epithelium only (1/73; no control, dysplastic, or adenocarcinoma tissue was assessed) is an infrequent event; data similar to a recent publication on nondysplastic BE from the United States (0/39 BE, 0/88 controls). Iyer et al. retrospectively investigated the prevalence of HPV in formalin-fixed samples, i.e., 84 BE (only 6 were dysplastic), 36 EAC, and 29 GEJ (rather than SCJ) specimens. DNA PCR, p16INK4A IHC, and chromogenic in situ hybridization results were not statistically different between the groups (BE 28%, EAC 31%, and controls 21%).
Some of these studies used tissue not conforming to standard endoscopic and histological definition of BE. All used a less sensitive nonnested PCR method, failed to address the transformation zone (SCJ rather than the GEJ), and most lacked clarity (or absence) of controls. Our unpublished preliminary data indicate that viral copy numbers in esophageal lesions may be low as compared with cervical neoplasia. Thus, highly sensitive detection methods, preferably using fresh frozen tissue, are required. Moreover, racial and geographic variations in the prevalence of HPV may partially account for the varying results between the studies. In this regard, the oral HPV prevalence of all serotypes in a large population-based US study involving 5,779 participants was 7% and that of HPV 16 was 1%. There has been no similar-sized study conducted in Australia, although one very small investigation by Sedlak-Weinstein et al. presented at the 47 Annual Scientific Meeting of the Australia and New Zealand Division of the International Association for Dental Research 2007 demonstrated the presence of HPV in the oral cavity of over 50% of healthy control subjects (n=31), the majority of whom were of the high-risk genotype. An earlier study performed on an adult Australian Caucasian population involving 60 subjects revealed that HPV DNA was found in the oral cavity of 60% of individuals and the predominant genotype was HPV 16. These studies may explain the higher background detection rate in our control subjects as opposed to that in the study of El-Serag et al.. In one respect, our findings are similar to most of the above studies on viral detection rates in esophageal epithelium, i.e., we have not established any meaningful association of HPV with Barrett's metaplasia or controls.
In our well-designed, prospective study addressing the squamous–metaplasia–dysplasia–adenocarcinoma sequence, we used the combination of nested PCR HPV DNA detection, p16INK4A IHC, and RT–PCR analysis of E6/E7 oncoprotein expression in tissue samples, which are all widely accepted methods to detect tumor-associated HPV infections. In fact, assessment of E6/E7 mRNA expression in fresh frozen tissue, as utilized in our study, is considered the gold standard for establishing the presence of biologically active HPV and would identify the truly viral-associated lesions. Although P16INK4A immunoreactivity is considered an excellent surrogate marker of biologically active HPV infection, it is not specific for HR-HPV. In fact, it is a better marker of retinoblastoma protein (pRb) inactivation than HR-HPV infection. Thus, pRb loss without HPV infection (e.g., mutational inactivation of pRb), other unknown molecular abnormalities of pRb/p16 cell cycle regulation, and a false-negative HPV DNA detection can all give rise to discrepancies.
There were a number of technical limitations to this study. First, the hybrid nature of the study population limits the accuracy of absolute population prevalence rates. This arose because progression from BE to BD is relatively uncommon and the numbers of patients with BD in the secondary-care setting where the study commenced were inadequate for statistical purposes. It had not been anticipated that the main differences in the GERD spectrum disease would be between the BE and BD patients. A decision was made to increase the numbers of patients with dysplasia during the course of the study by applying the same recruitment criteria within a tertiary-referral center with a much higher number of patients who had more severe disease. Although there is no obvious reason that the rate of HPV infection would be higher in the tertiary-care population other than if there was a causal association with BD, such a chance association cannot be ruled out. Second, the relatively small numbers of BD and EAC cases meant that subgroup analysis was inadequate to clearly differentiate the effects of different combinations of HPV DNA, p16INK4A, and E6/E7 mRNA positivity on disease severity. Third, the focal nature of HPV DNA positivity (the SCJ and lesion samples were often discordant in different patients) leaves open the possibility that there is circumferential and longitudinal discordance in the patients. We were unable to assess HPV status between pairs of biopsies from the same location as each of the biopsies was analyzed for different measures of HPV activity. It would have been reassuring to know whether there was consistent agreement between pairs of biopsies from the same location, which would have required multiple tissue specimens from the same site. This is exemplified in the dysplastic group where the majority had unifocal disease and the biopsies may have been obtained from nondysplastic areas. Microdissection of tissue specimens may have increased the yield of dysplastic cells for subsequent analysis. Nevertheless, we obtained targeted biopsies using chromoendoscopy in cases of BE, BD, and endoscopically resected EAC, which would have reduced the risk of obtaining nondysplastic tissue. Fourth, all HPV DNA–positive samples detected by nested PCR for the L1 gene and a proportion of viral DNA–negative specimens were subjected to E6/E7 mRNA analysis. It was assumed that all HPV DNA–negative tissue would be negative for E6/E7 mRNA and this was confirmed by the analysis of 56 HPV DNA–negative subjects. Rarely, disruption of the GP5+/6+ primer region due to a breakpoint in the L1 region during HPV integration may have produced a false-negative DNA test. This may have resulted in an underestimate of the E6/E7 mRNA oncogene expression in the tissue samples. Nevertheless, this discrepancy is very low and has been estimated to be between 2 and 3% in cervical neoplasia.
This study does not prove causality. It was not designed to do so, but was intended as the first in a sequence of studies that would achieve that goal. This investigation was cross-sectional, intended to demonstrate, as it has, an association between the presence of HPV DNA and the severity of GERD spectrum disease. This study suggests that HPV is associated with the transition from BE to dysplastic disease/adenocarcinoma. This is not the same as demonstrating that in an individual patient, HPV infection precedes neoplastic transformation. A longitudinal study demonstrating the temporal relationship between change in HPV positivity and histological disease progression would increase confidence in the validity of the association. The logistics of a longitudinal study would be difficult (large patient numbers followed up for a long duration, and knowing the right timing to observe and obtain biopsy samples from the patients), and it would not be undertaken without this initial cross-sectional study to refine the observations to be made. Even such a longitudinal demonstration would not prove that HPV infection caused the malignant change rather than the neoplastic alteration permitting the development of a transforming infection. It would necessitate intervention trials to demonstrate causation, and to explore treatment options including vaccination, although this presupposes that any preventative treatment would be effective. Studies showing an increasing viral count and greater HPV integration into the host genome with increasing disease severity would give further credence to a causal link. The demonstration of HPV replication in cell lines derived from BE and EAC would assist, assuming what happens in such in vitro studies is reflected commonly in real patients. In addition, providing evidence for the oncogenic properties of E6/E7 in BE cell lines and that EAC develops from such stable transduction would be useful in establishing a causal relationship. Furthermore, animal models may show in vivo feasibility.
Despite these limitations, given the strong association of transcriptionally active HR-HPV with BD and EAC, but not BE and controls, it is tempting to speculate that the virus may potentially be involved in esophageal carcinogenesis. These data provide robust justification for further detailed longitudinal, interventional, and molecular studies in this important area.
Discussion
This study is the first to find that HPV positivity as detected by PCR was significantly higher in patients with BD and EAC compared with controls and those with Barrett's metaplasia. Most of the HPV that was amenable to genotyping was of the HR type, i.e. HPV 16 and 18, which are the predominant strains found in cervical neoplasia, oropharyngeal malignancy, and anal cancer. Transcriptionally active HPV, as estimated by p16INK4A (a surrogate marker for HR HPV oncoprotein activity) IHC, and E6/E7 mRNA expression showed a similar pattern. p16INK4A positivity was greatest in BD and EAC and much less in control and BE subjects. The strength of the association with more severe GERD spectrum disease was greatest when HPV DNA, p16INK4A, and E6/E7 mRNA were all positive compared with all being negative. Also, both the combination of E6/E7 mRNA positivity with HPV DNA and p16INK4A with HPV DNA were independently but less strongly associated with more severe GERD spectrum disease. Furthermore, we have demonstrated that HPV infections in the esophagus were focal. In both dysplastic/endoscopically resected EAC and nondysplastic patients, HR-HPV was detected more commonly at the transformation zone (SCJ) compared with the lesion in BE or corresponding position in squamous epithelium as we previously hypothesized. Interestingly, in hospital and GERD controls as well as Barrett's nondysplastic individuals, HPV infections were largely biologically irrelevant at the time of biopsy, as evidenced by minimal p16INK4A positivity (5.9%) and absence of E6/E7 mRNA expression. Conversely, in patients with BD and EAC, 44% of patients had p16INK4A staining and in 30% of patients, E6/E7 oncoprotein expression was detected. Again, in BD tissue, this transcriptional activity was more concentrated in the SCJ. We hypothesize that progressive acid damage to the esophagus increases the likelihood of mucosal breaks and allows the virus to enter the basal layer of the transformation/transition zone, i.e., the junction between the squamous and the columnar lining. However, the number of cases observed was insufficient to test this hypothesis. HPV affinity is to junctional tissue, as basal cells are particularly accessible at the squamocolumnar transformation zone and especially susceptible to this viral infection. It is known that transforming HPV infections in different epithelia (cervix, anus, and oropharynx) cause cancers mainly at the transformation zone. A discrete population of SCJ cells of the cervix with a unique morphology and gene expression profile has been implicated in the pathogenesis of cervical cancer. Intriguingly, it has also been reported that BE is derived from a discrete population of embryonic cells residing at the SCJ. Moreover, it has been demonstrated that a single cell of origin (common progenitor) could differentiate into squamous or columnar epithelium, thus exhibiting esophageal plasticity. Furthermore, using a transgenic mouse model, it has been demonstrated that BD and BE can be initiated at the SCJ.
Our results indicate a much higher HPV prevalence of 18% in controls compared with our prior expected prevalence of 9% in esophageal tissue. When we commenced this study in 2008, there were no reliable HPV prevalence data available for the Australian population. A guesstimate prevalence of 9% was obtained from the meta-analysis by de Sanjose et al. on worldwide HPV prevalence in cervical tissue of women with normal cytology. A recent study reported HPV DNA prevalence of any serotype of 41.5% and HPV 16/18 of 12.8% in cervical tissue of 2,152 Australian women aged between 17 and 40 years attending a community-based health service for routine Pap screening. In another publication, the HPV genoprevalence in women aged 18–24 years (n=202) attending family planning clinics in Sydney, Melbourne, and Perth during the pre-vaccine period of 2005–2007 has been reported to be 21.3% for HPV 16 and 8.4% for HPV 18. Hence, the above viral prevalence rates in Australian women are comparable to our control group prevalence of HR-HPV. However, none of these studies are directly relevant, as none involved observations of esophageal HPV prevalence.
In a retrospective study from Mexico, HPV DNA, mostly low-risk types, was detected by in-house PCR in 96% of paraffin-embedded tissue biopsies of BE (without endoscopic or histopathological classification) and 88% of undefined esophageal cancers. Nevertheless, Rai et al. reported that HPV infection in fresh BE epithelium only (1/73; no control, dysplastic, or adenocarcinoma tissue was assessed) is an infrequent event; data similar to a recent publication on nondysplastic BE from the United States (0/39 BE, 0/88 controls). Iyer et al. retrospectively investigated the prevalence of HPV in formalin-fixed samples, i.e., 84 BE (only 6 were dysplastic), 36 EAC, and 29 GEJ (rather than SCJ) specimens. DNA PCR, p16INK4A IHC, and chromogenic in situ hybridization results were not statistically different between the groups (BE 28%, EAC 31%, and controls 21%).
Some of these studies used tissue not conforming to standard endoscopic and histological definition of BE. All used a less sensitive nonnested PCR method, failed to address the transformation zone (SCJ rather than the GEJ), and most lacked clarity (or absence) of controls. Our unpublished preliminary data indicate that viral copy numbers in esophageal lesions may be low as compared with cervical neoplasia. Thus, highly sensitive detection methods, preferably using fresh frozen tissue, are required. Moreover, racial and geographic variations in the prevalence of HPV may partially account for the varying results between the studies. In this regard, the oral HPV prevalence of all serotypes in a large population-based US study involving 5,779 participants was 7% and that of HPV 16 was 1%. There has been no similar-sized study conducted in Australia, although one very small investigation by Sedlak-Weinstein et al. presented at the 47 Annual Scientific Meeting of the Australia and New Zealand Division of the International Association for Dental Research 2007 demonstrated the presence of HPV in the oral cavity of over 50% of healthy control subjects (n=31), the majority of whom were of the high-risk genotype. An earlier study performed on an adult Australian Caucasian population involving 60 subjects revealed that HPV DNA was found in the oral cavity of 60% of individuals and the predominant genotype was HPV 16. These studies may explain the higher background detection rate in our control subjects as opposed to that in the study of El-Serag et al.. In one respect, our findings are similar to most of the above studies on viral detection rates in esophageal epithelium, i.e., we have not established any meaningful association of HPV with Barrett's metaplasia or controls.
In our well-designed, prospective study addressing the squamous–metaplasia–dysplasia–adenocarcinoma sequence, we used the combination of nested PCR HPV DNA detection, p16INK4A IHC, and RT–PCR analysis of E6/E7 oncoprotein expression in tissue samples, which are all widely accepted methods to detect tumor-associated HPV infections. In fact, assessment of E6/E7 mRNA expression in fresh frozen tissue, as utilized in our study, is considered the gold standard for establishing the presence of biologically active HPV and would identify the truly viral-associated lesions. Although P16INK4A immunoreactivity is considered an excellent surrogate marker of biologically active HPV infection, it is not specific for HR-HPV. In fact, it is a better marker of retinoblastoma protein (pRb) inactivation than HR-HPV infection. Thus, pRb loss without HPV infection (e.g., mutational inactivation of pRb), other unknown molecular abnormalities of pRb/p16 cell cycle regulation, and a false-negative HPV DNA detection can all give rise to discrepancies.
There were a number of technical limitations to this study. First, the hybrid nature of the study population limits the accuracy of absolute population prevalence rates. This arose because progression from BE to BD is relatively uncommon and the numbers of patients with BD in the secondary-care setting where the study commenced were inadequate for statistical purposes. It had not been anticipated that the main differences in the GERD spectrum disease would be between the BE and BD patients. A decision was made to increase the numbers of patients with dysplasia during the course of the study by applying the same recruitment criteria within a tertiary-referral center with a much higher number of patients who had more severe disease. Although there is no obvious reason that the rate of HPV infection would be higher in the tertiary-care population other than if there was a causal association with BD, such a chance association cannot be ruled out. Second, the relatively small numbers of BD and EAC cases meant that subgroup analysis was inadequate to clearly differentiate the effects of different combinations of HPV DNA, p16INK4A, and E6/E7 mRNA positivity on disease severity. Third, the focal nature of HPV DNA positivity (the SCJ and lesion samples were often discordant in different patients) leaves open the possibility that there is circumferential and longitudinal discordance in the patients. We were unable to assess HPV status between pairs of biopsies from the same location as each of the biopsies was analyzed for different measures of HPV activity. It would have been reassuring to know whether there was consistent agreement between pairs of biopsies from the same location, which would have required multiple tissue specimens from the same site. This is exemplified in the dysplastic group where the majority had unifocal disease and the biopsies may have been obtained from nondysplastic areas. Microdissection of tissue specimens may have increased the yield of dysplastic cells for subsequent analysis. Nevertheless, we obtained targeted biopsies using chromoendoscopy in cases of BE, BD, and endoscopically resected EAC, which would have reduced the risk of obtaining nondysplastic tissue. Fourth, all HPV DNA–positive samples detected by nested PCR for the L1 gene and a proportion of viral DNA–negative specimens were subjected to E6/E7 mRNA analysis. It was assumed that all HPV DNA–negative tissue would be negative for E6/E7 mRNA and this was confirmed by the analysis of 56 HPV DNA–negative subjects. Rarely, disruption of the GP5+/6+ primer region due to a breakpoint in the L1 region during HPV integration may have produced a false-negative DNA test. This may have resulted in an underestimate of the E6/E7 mRNA oncogene expression in the tissue samples. Nevertheless, this discrepancy is very low and has been estimated to be between 2 and 3% in cervical neoplasia.
This study does not prove causality. It was not designed to do so, but was intended as the first in a sequence of studies that would achieve that goal. This investigation was cross-sectional, intended to demonstrate, as it has, an association between the presence of HPV DNA and the severity of GERD spectrum disease. This study suggests that HPV is associated with the transition from BE to dysplastic disease/adenocarcinoma. This is not the same as demonstrating that in an individual patient, HPV infection precedes neoplastic transformation. A longitudinal study demonstrating the temporal relationship between change in HPV positivity and histological disease progression would increase confidence in the validity of the association. The logistics of a longitudinal study would be difficult (large patient numbers followed up for a long duration, and knowing the right timing to observe and obtain biopsy samples from the patients), and it would not be undertaken without this initial cross-sectional study to refine the observations to be made. Even such a longitudinal demonstration would not prove that HPV infection caused the malignant change rather than the neoplastic alteration permitting the development of a transforming infection. It would necessitate intervention trials to demonstrate causation, and to explore treatment options including vaccination, although this presupposes that any preventative treatment would be effective. Studies showing an increasing viral count and greater HPV integration into the host genome with increasing disease severity would give further credence to a causal link. The demonstration of HPV replication in cell lines derived from BE and EAC would assist, assuming what happens in such in vitro studies is reflected commonly in real patients. In addition, providing evidence for the oncogenic properties of E6/E7 in BE cell lines and that EAC develops from such stable transduction would be useful in establishing a causal relationship. Furthermore, animal models may show in vivo feasibility.
Despite these limitations, given the strong association of transcriptionally active HR-HPV with BD and EAC, but not BE and controls, it is tempting to speculate that the virus may potentially be involved in esophageal carcinogenesis. These data provide robust justification for further detailed longitudinal, interventional, and molecular studies in this important area.
Source...