Low-Positive Antibody Titer against Helicobacter pylori Cytotoxin-Associated Gene A (CagA) May Predict Future Gastric Cancer Better Than Simple Seropositivity against H. pylori CagA or against H. pylori

Introduction

About half of the world population has been estimated to be infected with Helicobacter pylori. The IARC designated H. pylori as a human class I carcinogen in 1994 (1). This was evidenced by several meta-analyses of retrospective and prospective case-control studies on gastric cancer (2-5). Prospective studies showed that people infected with H. pylori had a 2.5- to 3-fold increased risk for gastric cancer in the future (2, 3). Although people without infection with H. pylori showed negligible risk for noncardia gastric cancer (6, 7), only a part of infected subjects developed gastric cancer. Many factors other than infection with H. pylori seem to control the development of gastric cancer [i.e., strains of H. pylori with different virulence, host genetic factors, and lifestyle factors, such as smoking, salt intake, and other dietary history (8, 9)]. To screen high-risk groups for gastric cancer among H. pylori–infected subjects and understand the host-pathogen interaction, we need more new biomarkers relating to future cancer risk in H. pylori–infected subjects.

Cytotoxin-associated gene A (CagA) product determines the virulence of H. pylori (10, 11). CagA protein is injected into epithelial cells on colonization in the stomach, and it perturbs cellular signaling events in the injected epithelial cells by recruiting and binding with cellular phosphatase SHP-2 (12) and activates transcriptional factors (13, 14). Consequently, the CagA-positive strain induces more severe inflammatory reaction than the CagA-negative strain does. A meta-analysis of 16 studies showed that infection with CagA-positive strains increased gastric cancer risk over the risk associated with H. pylori infection alone (15). However, it has not been elucidated whether the magnitude of immune response to CagA is associated with risk for gastric cancer. Thus far, only one study investigated the association between CagA antibody titer and gastric cancer in a prospective way (16) and another study investigated the issue in a retrospective way (17). Interestingly, a low titer of CagA IgG antibody was associated with high risk for gastric cancer in both studies, although it was not clear whether the observed phenomenon was reproducible in other prospective and retrospective studies. In the present study, we have investigated firstly whether a low antibody titer against CagA is an independent risk factor for future noncardia gastric cancer by a nested case-control study in the Adult Health Study cohort of atomic bomb survivors.

Materials and Methods

Cohorts

The Adult Health Study cohort was established in 1958 as the clinical subcohort of the Life Span Study cohort in Hiroshima and Nagasaki, which consisted of 20,000 age-, gender-, and city-matched proximal and distal survivors and control residents not present in the cities at the time of bombings (18). Subjects have been checked biennially in our outpatient clinics in Hiroshima and Nagasaki.

Cases and Controls

From 1969, serum samples have been stored systematically in either frozen or freeze-dried condition on each occasion of visiting outpatient clinics. Cancer incidence cases were detected from the Hiroshima Tumor and Tissue registry and Nagasaki Cancer Registries. There were 719 gastric cancer cases among Adult Health Study participants diagnosed from January 1970 to December 2001 who visited our outpatient clinics before diagnosis. Second primary and third primary gastric cancer cases and cases with less than two stored samples were omitted. Consequently, 423 primary cases remained. Two cases did not store serum samples sufficient enough for the present experiment, 64 cases did not store serum samples 1 to 5 years (mean, 2.3 years) before diagnosis, and 1 case could not be matched to three controls as mentioned below. Consequently, 356 cases were finally selected. Three controls per each case were selected from cohort members matched on gender, city, age (±5 years), year of serum storage (±1 year), and type of storage method (frozen or freeze dried) and countermatched on dose groups. Some cohort members were selected as a control for more than one case at different time points in a nested case-control study design as discussed elsewhere (19). Table 1 shows the demographic features of subjects with intestinal-type and diffuse-type noncardia gastric cancer and cardia gastric cancer.

Radiation Dose

We took the radiation dose as the γ dose plus 10 times the neutron dose using the stomach doses calculated by dosimetry system DS02 (20). After comparing the tertiles of dose in cases to those in the full Adult Health Study, we decided on 50 and 750 mGy as the cutpoints for the three nonzero dose categories. For each case, we took one control from each of the three dose categories not occupied by the case. We matched cases with unknown dose to one control with zero dose and two controls with unknown dose.

Serologic Tests

A serologic test for H. pylori was done in duplicate using an enzyme immunoassay kit (AutoAce H. pyloriG, Alfresa Pharma). According to the manufacturer's instruction, 100-fold diluted serum was measured and <16.5 units/mL was considered as negative. A serologic test for CagA was done in duplicate using an ELISA kit (CagA IgG EIA WELL, Radim), in which recombinant CagA protein was used as an antigen. Sensitivity and specificity of the kit were given by the manufacturer as 93.7% and 100%, respectively. According to the manufacturer's recommendation, 300-fold diluted serum was measured and <15 relative units/mL was considered as negative.

Chronic Atrophic Gastritis

Grade II chronic atrophic gastritis (CAG) was diagnosed by the criteria of Miki et al. (21) using pepsinogen measurements (i.e., pepsinogen I of <50 μg/L and pepsinogen I/pepsinogen II ratio of <3.0). We tested other cutoff value of pepsinogen I (i.e., 30 or 70 μg/L); however, it did not alter our conclusion about the risk of CAG on noncardia gastric cancer.

Smoking and Other Clinical and Epidemiologic Information

Smoking and other dietary information was obtained from three sources: Adult Health Study interview in 1963 to 1965, Mail survey in 1968, and Mail survey in 1978 to 1980. The most recent data on smoking before gastric cancer diagnosis or the equivalent time for matched controls were used. Smoking habit was divided into “never,” “stopped,” and “current smoking.”

Ethical Consideration

This study (Radiation Effects Research Foundation Research Protocol 2-04) was reviewed and approved by the Radiation Effects Research Foundation Research Protocol Review Committee on December 12, 2003 and by the Ethical Committee on February 26, 2004.

Statistical Analyses

The basic risk model used in the conditional logistic regression iswhere RR is the relative risk, D is radiation dose in gray, and the x_is are the other risk factor variables. Each RR estimate shown here for an individual risk factor is adjusted for the other risk factors in the model as shown in the table in which it is included. RR estimates for the various combinations of H. pylori IgG and CagA IgG are with respect to the reference category: seronegative for H. pylori IgG. Samples seronegative for H. pylori IgG but seropositive for CagA IgG were included in this reference category because there were not enough cases or not enough controls to evaluate them separately; furthermore, there are few enough such results to be consistent with estimated false-positive error rates for CagA IgG. We refer to the portion in parentheses as the “linear term” and the remainder as the “log-linear term.” We fitted regressions using Epicure software (HiroSoft International Corp.). The linear term for radiation was included in all models as radiation effects on cancer generally fit to a linear nonthreshold model.

Results

Distribution of IgG Antibody Titers against H. pylori and CagA

The histograms of the logarithms of H. pylori IgG results, shown separately for controls and cases in Fig. 1A , seem bimodal, with one peak below the manufacturer's recommended cutoff of 16.5 (log = 2.8; Fig. 1A, vertical line) and one well above the cutoff. The results above the cutoff could not be fitted well enough with a single normal distribution to give acceptable results with a χ² goodness-of-fit test, but two distributions sufficed, as shown by the dotted and dashed curves, and the mixture proportion for them was very similar in cases and controls.

Among CagA IgG results, a notable feature was that the distribution of the logarithms of CagA results in H. pylori IgG-seropositive samples for cancer cases showed a distinct area of high frequency in the region just above the cutoff value of 15 (log = 2.7; Fig. 1B, solid vertical line), which is not present in the distribution of results for controls. A third normal distribution just above the cutoff value gave a significant improvement of fit as shown in Fig. 1B (bottom).

We investigated the occurrence of samples that were seronegative for H. pylori IgG but seropositive for CagA IgG titer. The proportion of samples in this category was actually greater in controls (71 of 869, 8%) than in cases (12 of 319, 4%; P = 0.007, Fisher's exact test) and did not differ by histologic classification of intestinal versus diffuse type. These proportions seem small enough to be consistent with the expected error rates of standard diagnostic tests for the two assays involved.

To determine a cutoff value for distinguishing a low-titer and a high-titer region within positive results, we did several types of analyses. Some common features in the appearance of the distributions suggest that good results might be obtained with fewer fitted distributions by using either a transformation other than the logarithm or a distribution other than the normal. However, the appearance of a separate distribution in the low-titer region of the CagA IgG test results for cases seems quite robust. Work in analyzing these distributions is ongoing and additional details will be presented in a separate report. Cutoff values were evaluated for integers on the interval from 20 to 30, and it was found that, whereas the fraction of cases included as “low-titer CagA” increases from about 8% to 20% on this interval, the estimated RR of “low-titer CagA” in all noncardia cases using the model shown in Table 2 is about 3.2 to 3.4 near the ends of the interval and is almost constant at about 3.7 to 3.8 at values from 23 to 26. The values reported in this work are based on a cutoff of 23 (logarithm 3.1), which maximized the estimated RR.

The Analysis of Risk Factors for Future Noncardia Gastric Cancer

In the preliminary analyses, we investigated the association of noncardia gastric cancer with age at the time of bombing, body mass index, education, alcohol, and other dietary habits besides variables in Table 2 and found none of them being a significant risk factor. Thus, we omitted these variables from the final model. Among seropositive subjects for H. pylori IgG, those with low IgG antibody titer against CagA showed higher and more significant risk (RR, 3.9; P < 0.001) for future noncardia gastric cancer than those with CagA IgG seronegative (RR, 2.2; P = 0.0052) or high IgG antibody titer against CagA (RR, 2.0; P = 0.0022). We evaluated a linear contrast between low and high IgG antibody titer, which confirmed a statistically significant difference between their RR estimates. CAG and smoking were also independent risk factors (Table 2). There was an interaction between radiation dose and smoking, and radiation showed significant risk only for noncurrent smokers. Likewise, there was a marginally significant interaction between radiation dose and CAG in all noncardia gastric cancer (P = 0.0635) and diffuse-type gastric cancer (P = 0.0503) in a separate analysis. We also investigated interactions between the following variables, but none of them was significant: radiation dose versus H. pylori IgG seropositivity, radiation dose versus CagA, and gender versus H. pylori IgG seropositivity.

The Risk Factors for Noncardia Gastric Cancer of Diffuse and Intestinal Types

For intestinal-type gastric cancer, H. pylori IgG(+)/CagA_low, representing 15% of intestinal-type cases, showed significant risk (RR, 9.9; P < 0.001), whereas H. pylori IgG(+)/CagA_high, representing 60% of intestinal-type gastric cancer, did not (RR, 1.4; P = 0.26; Table 3 ). In contrast, both H. pylori IgG(+)/CagA_low (RR, 4.8; P = 0.0013) and H. pylori IgG(+)/CagA_high (RR, 3.2; P = 0.0027) were significant risk factors for diffuse type. CAG showed similar levels of risk for intestinal and diffuse type (RR, 2.2 and 2.4, respectively). Current smoker was a significant risk factor only for diffuse type (RR, 4.6; P < 0.001), and the risk of current smoker for diffuse type was significantly different from that for intestinal type (P = 0.0372). Radiation dose showed significant risk for diffuse type among nonsmokers (RR, 4.0; P = 0.0457) but not for diffuse type among nonsmokers or intestinal type regardless of smoking status. However, radiation risks between diffuse and intestinal types were not significantly different.

Discussion

In the present study, we have confirmed Nomura et al.'s report (16) that the lowest quartile of antibody titer to CagA is a risk factor for intestinal-type gastric cancer in a prospective study and extended the observation. First, the distribution of logarithmic converted CagA titer can be divided into high and low in cases, and low antibody titer, representing 15% of gastric cancer cases, is a risk factor for both intestinal-type and diffuse-type gastric cancer. Second, it is true even after including CAG and smoking status in a model. Thus, CagA might play a role other than simply inducing CAG in the process of carcinogenesis.

One strong point of this cohort-based case-control study is the usage of stored sera 1 to 5 years (mean, 2.3 years) before diagnosis. By using stored sera, we can evaluate the host-pathogen interaction before the interaction is deeply affected by the natural disappearance of H. pylori colonization due to atrophic gastritis or by immune suppression associated with advanced cancer. The second strong point of this study is gastric cancer cases being restricted to first primary cases. By eliminating second primary or third primary gastric cancer, we can avoid the effect of antitumor drugs or radiation used for former cancer on the progression of gastric cancer.

One weak point of this study is the usage of old stored samples. Along with the storage years, the biological activity of protein or peptide tends to decay or erode. In the present study, we wish to minimize this erosion effect by two maneuvers [i.e., the utilization of pepsinogen I/pepsinogen II ratio and the selection of controls adjusted for both the type of serum storage (either frozen or freeze dried) and the year of storage (within 1 year)]. Although storage years eroded pepsinogen and antibodies, cases were matched with controls on the time of blood donation, so sample aging would affect cases and controls equally. Thus, we believe that the usage of old stored samples did not introduce a large bias in the present study.

We have to be cautious when we interpret the biological meaning of low IgG antibody titer against CagA. It is well known that H. pylori IgG antibody titer tends to reduce faster than CagA IgG antibody (22) or even becomes false negative along with the progression of CAG. This is because the loss of gastric epithelial cells is associated with the reduced antigenic load of H. pylori (7, 15). However, the distribution of H. pylori IgG antibody was not biphasic in cancer subjects 2.3 years before diagnosis and almost the same as control subjects (Fig. 1A), indicating that low IgG titer against CagA was not primarily due to the reduced load of H. pylori in the stomach. In a separate analysis, we investigated whether CAG had progressed faster in low CagA antibody group than in high CagA antibody group by two pepsinogen I and pepsinogen II measurements 1 to 5 years and 8 to 15 years before diagnosis in 1,027 cases and controls. It was found that the progression of CAG during these intervals did not differ between two groups (data not shown). Likewise, the low antibody titer against CagA was unlikely to be due to the general immune-compromised state in cancer subjects 2.3 years before diagnosis, as antibody response to H. pylori was not suppressed (Fig. 1). Thus, the low antibody response is unique to CagA protein.

There are at least two plausible scenarios for low IgG titer against CagA. The first scenario is immune response gene control. It is known that certain HLA polymorphisms are associated with either resistance or susceptibility to H. pylori–associated pathology (23). The second scenario is that low IgG antibody titer against CagA is caused by the reduced transcription of the CagA gene in H. pylori–infected subjects. It is known that CagA transcription is controlled by gastric acidity (23). CagA expression is expected to be low in those who are hypochlorhydria either by genetic reason (24-26) or by the progression of atrophic gastritis. Thus, low CagA IgG antibody titer might be a surrogate marker of prolonged hypochlorhydria. These two scenarios are testable by investigating genetic polymorphisms in HLA and proinflammatory cytokines in the future. Whatever the scenario is, low CagA IgG titer is an interesting biomarker that serves as a clue to further understand the host-pathogen interaction.

Radiation effect on gastric cancer has been estimated by the Life Span Study, in which both mortality and incidence were used as end points: the RRs of radiation dose (1 Gy) were 1.20 for mortality (27) and 1.32 for incidence (28), respectively. However, in these studies, radiation effect was not analyzed on gastric cancer of different histologic types with other important risk factors, such as H. pylori or smoking. Estimated RR of radiation dose is 1.4 if smoking is not adjusted for (data not shown), and this estimated value is not so different from the above-mentioned RRs. In this study, it is shown firstly that radiation increased the risk for noncardia gastric cancer of diffuse type but not for intestinal type after adjusting for H. pylori, CagA, and CAG. Secondly, radiation risk for noncardia gastric cancer was negligible in smokers. On the face of it, this result may be unexpected, but it has some similarity to recent findings on the combined effects of radiation and smoking on lung cancer, where radiation risk was negligible for people with daily smoking dose of more than 16 cigarettes (29). In recent years, it has become known that radiation can cause nontargeted or bystander effects and can also induce delayed effects (30-32). The presence of interaction between atomic bomb irradiation in 1945 and smoking decades later may support the nontargeted and delayed effect of radiation. However, we have to be cautious not to overinterpret these results, as there must be unexpected confounding factors not included in the current analysis.