Proc. Natl. Acad. Sci. USA Vol. 76, No. 5, pp. 2438-2441, May 1979 Medical Sciences

Serum iron levels and response to hepatitis B virus (renal dialysis/aspartate aminotransferase/cancer/hepatitis carrier/antibody against hepatitis B)

CRAIG FELTON, EDWARD D. LUSTBADER, CYNTHIA MERTEN, AND BARUCH S. BLUMBERG The Institute for Cancer Research, The Fox Chase Cancer Center, Philadelphia, Pennsylvania 19111

Contributed by Baruch S. Blumberg, February 12,1979

ABSTRACT Response to hepatitis B virus (HBV) infection [HBV surface antigen (HBsAg) and antibody to HBsAg (antiHBs)], serum iron, total iron-binding capacity, hematological status (erythrocytes, Hb, and hematocrit), and evidence of liver damage (serum glutamic pyruvic transaminase; aspartate aminotransferase, L-aspartate:2-oxoglutarate aminotransferase, EC 2.6.1.1) were determined for 201 patients on chronic renal dialysis. Four factors-serum iron level, transaminase level, sex, and HBV response [i.e., infected-HBsAg(+) (HBsAg positive), anti-H1Bs(+) (anti-HBs positive), or no response]-were analyzed simultaneously to test the hypothesis that serum iron is higher in those with HBsAg in their serum than in those without HBsAg, independent of the transaminase level. Four independent, statistically significant two-factor interactions were identified. (i) Serum iron is higher in those HBsAg(+) (ii) Serum iron is higher in those with increased transaminase. (iii) Transaminase is higher in those HBsAg(+). (iv) Males are more likely to be HbsAg(+) and females are more likely to be antiHBs(+) Also, those who are HBsAg(+) have significantly higher percent iron saturation (serum iron/total iron mdig capacity). That is, the hypothesis was supported by the findings. Several additional biological hypotheses are suggested, including a possible role of increased iron levels in susceptibility and response to HBV infection and the possible relationship between higher iron levels and the likelihood of HBV infection progressing to primary hepatocellular carcinoma. In addition, further tests of the initial hypothesis in nonhospitalized populations with endemic HBV infection are proposed.

The role of iron in susceptibility and response to infectious disease (primarily bacterial but possibly viral) has been discussed for some years and has recently been summarized (1). Serum iron becomes elevated in hepatitis, presumably released from hepatocytes as a result of cellular damage. We found that males with Down syndrome who had hepatitis B surface antigen (HBsAg) in their blood had significantly higher levels of serum iron than those who did not have HBsAg (2). However, the serum iron level was not correlated with the level of serum glutamic pyruvic transaminase (SGPT) (aspartate aminotransferase; L-aspartate:2-oxoglutarate aminotransferase, EC 2.6.1.1), which is considered to be an indication of liver cell damage. Therefore, it appeared that the higher serum iron in the HBsAg(+) (HBsAg positive) individuals was not due only to the release of cellular iron which accompanies liver damage. Based on these observations, we decided to study a second group of individuals at high risk of hepatitis B virus (HBV) infection. Specifically, we tested the hypothesis that the serum iron level is higher in patients on chronic renal dialysis who have HBsAg in their blood than those who have antibody to HBsAg (antiHBs) or those who lack both HBsAg and anti-HBs (no evidence of infection). There are additional reasons for learning more about the relationship of iron to hepatitis infection. Tumor cells may require iron for effective growth and spread (3). Infection with

HBV is significantly associated with chronic liver disease and primary cancer of the liver in Asia, Africa, and other regions where these are common diseases. The relation may be etiologic, and we are currently studying strategies for the control of HBV infection. In connection with this, it is important to identify interactions that might contribute to the development of cancer. In the tropics, iron metabolism in humans is affected by a large number of environmental and pathologic conditions, including nutrition, methods of food preparation, hemolytic anemias, and bacterial and viral infection (3, 4). Further study of iron in relation to hepatitis B infection may contribute to an understanding of this complex problem.

METHODS Sera were collected in iron-free tubes over a 4-day period from patients in two commercial renal dialysis clinics operated by Biomedical Applications of Philadelphia. Serum iron, total iron-binding capacity, SGPT, HBsAg, and anti-HBs were determined at The Institute for Cancer Research. Serum iron and total iron-binding capacity were determined colorimetrically with Hyland Ferro-chek II kits (Travenol, Costa Mesa, CA), HBsAg by the Ausria II method (Abbott), and anti-HBs by direct hemagglutination with HBsAg-coated erythrocytes (Electronucleonics Laboratories, Bethesda, MD) (5). Sera were considered positive for anti-HBs if they agglutinated cells at a titer of 1/4 or greater. SGPT was determined by a modification of the method of Henry et al. (6). All other laboratory tests were performed by the commercial laboratory (Metpath, Philadelphia, PA) that routinely provides services for the dialysis clinic.

PATIENT POPULATION A total of 201 sera were collected. The distribution of patients by HBV response and sex is shown in Table 1. The patients included in this study were receiving dialysis three times each week. All of the patients routinely receive iron therapy, either Fe2SO4 (325 mg three times a day) or intravenous iron (Imferon). The response of the patient to hepatitis B infection is not considered in determining the iron therapy. Further, renal dialysis patients have a chronic anemia apparently unrelated to iron deficiency. RESULTS AND DISCUSSION Examination of the present data (Table 2) began with an analysis similar to that of the Down syndrome study.(2). Since only males were available in the Down syndrome study, just the male dialysis patients were used for this analysis. There was a significantly higher level of serum iron in male patients with HBsAg than in those without HBsAg. SGPT levels in HBsAg(+)

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U. S. C. §1734 solely to indicate this fact.

Abbreviations: HBV, hepatitis B virus; HBsAg, HBV surface antigen; anti-HBs, antibody to HBVsAg; HBsAg(+), HBsAg positive; antiHBs(+), anti-HBs positive; SGPT, serum glutamic pyruvic transaminase.

2438

Proc. Natl. Acad. Sci. USA 76 (1979)

Medical Sciences: Felton et al.

Table 3. Distribution of data according to HBV response, SGPT level, serum iron level, and sex

Table 1. Distribution of renal dialysis patients by sex and response to HBV infection Male Female Total HBsAg(+) Anti-HBs(+) No evidence of infection Total

27 36 50 113

12 40 36 88

2439

Low iron Male Female

39 76 86 201

Middle iron Male Female

High iron Male Female

Low SGPT (_ 15 international units/liter) 2 2 5 6 3 HBsAg(+) 11 11 14 6 Anti-HBs(+) 12 No evidence 9 10 9 6 of infection 19 High SGPT (>15 international units/liter) 1 11 1 1 0 HBsA&(+) 2 1 0 4 0 Anti-HBs(+) No evidence

patients were, as expected, higher than the levels in HBsAg(-) patients. The measurements of total iron-binding capacity, Hb, and hematocrit in patients with HBsAg were also compared with those of the patients without HBsAg, and no significant differences were found. These findings are consistent with those of the Down syndrome study. In this investigation more data was available than in the Down syndrome study because females were included and anti-HBs status was determined. The differences in serum iron found between those who were HBsAg(+) and the other HBV response groups occurred primarily in the males. Percent iron saturation was higher in HBsAg(+) patients than in all others, but the only statistically significant difference occurred in the males. The SGPT levels were significantly lower in patients with anti-HBs than in the HBsAg(+) and apparently uninfected patients. There were also significant differences between males and females in erythrocyte, Hb, and hematocrit levels (males had higher values). Among the HBV response groups the difference was significant only in anti-HBs(+) patients. Some interesting biological models relating to differences between males and females in response to hepatitis B infection could be made if these findings are supported in subsequent investigations. We then proceeded to search for interactions between the four variables: serum iron, sex, HBV response, and SGPT level. Iron levels were scored as low, middle, or high based on a division of the data into three groups containing approximately

of infection

1

3

4

3

3 11 3 4 3

8

9

Three of the individuals from Table 1 had incomplete results and are not included.

interactions of these four variables were then evaluated to see how much each contributed to the explanation of the data. The best model consisted of the iron and HBV, iron and SGPT, HBV and SGPT, and sex and HBV interactions. To determine whether any of these interactions were due to the effect of transfusions and intravenous iron therapy received by some patients, we removed all data from patients receiving either treatment within 1 month prior to iron and HBV testing and reanalyzed the remaining data. HBV response, SGPT level, and serum iron were considered, but not sex due to insufficient numbers. The interactions between iron and HBV and between iron and SGPT remained significant, while the interaction between HBV and SGPT was only marginally significant. The best model included the interactions between iron and SGPT and between iron and HBV and not the interaction between HBV and SGPT. The interactions included in these models are consistent with previous observations. London and Drew (7) have shown that, in renal dialysis patients (which include the subjects of this study), males are significantly more likely to become carriers of HBsAg and females to develop anti-HBs. The virus and SGPT interaction has been known for many years (see for example, ref. 8) and was one of the criteria for the establishment of the association between Australia antigen and hepatitis. An

equal numbers of individuals. The distribution of the data according to these four variables is shown in Table 3. The testing of a series of hypotheses indicated that the data could be explained by considering only two-factor interactions. (Details of this analysis are given in the Appendix.) Pairwise

Table 2. Mean values for groups of renal dialysis patients

SI, mg/dl HBsAg(+)

Male 1141

Female 214

Total 1451

TIBC, mg/dl Male Female Total 278

305

Anti-HBs(+) No evidence of infection Total

HBsAg(+) Anti-HBs(+) No evidence of infection Total

95*

107

102

10395 9 103 117 RBC, 106 cells/mm3 Male Female Total 2.6 2.2 2.5 2.6 * 2.3 2.5 2.4 2.5 **

2.3 2.3

2.3

Male

42.41

285

*# 0 ** go

**

304

284

281 286 Hb, g/dl Male Female 6.8 6.7 7.1 * * 6.3 283 288

6.7 6.9

*

6.4 6.4

29.6*

282

34.6 34.9

6.6

Differences were assessed by the Mann-Whitney test. % iron saturation blood cells. *, P = 0.05; **, P = 0.01; ***, P = 0.001.

46.5

43.6,

28.1n ** ~~~~~~~~~**Ig **I0 20.8

25.8

36.3

33.0*

11.5*

6.2

10.8

34.1

14.8 16.9

21.4 15.9

17.2

33.4 36.4 Hematocrit, % Male Female 22.4 21.8 23.5 * 20.8 21.9 22.6

=

Total

0

294

Total 6.8 6.7

Female

SGPT, international units/liter Male Female Total

% iron saturation

*

20.8 20.0

*'

Total 22.2 22.1 21.5

serum iron (SI)/total iron-binding capacity (TIBC). RBC, red

2440

Medical Sciences: Felton et al.

increase of serum iron in association with liver damage (i.e., elevation of SGPT) has been clinically recognized in patients with hepatitis and liver disease. The interaction between iron and HBV was shown in the Down syndrome study and confirmed in this study. However, it was not recognized until this analysis that all four interactions act independently and simultaneously and, in particular, that the presence of HBsAg by itself, independent of SGPT elevation, is associated with elevations of serum iron. Implications This study indicates that the association of elevated serum iron with the presence of HBsAg in the serum is due not only to the breakdown of hepatic cells, but also to the presence of HBsAg itself. It is remarkable that these differences can be detected in this patient population given the possible masking effects of iron therapy, the anemia of chronic dialysis, and the decrease of iron absorption due to the use of aluminum hydroxide (9). We now wish to determine if these same associations exist in other populations with a high frequency of HBV infection-for example, nonhospitalized individuals in Africa, Asia, or Oceania. If our findings are supported in these populations, then there are some interesting biological and pathological questions that can be raised. All living cells require iron. In humans, a method of defense after microbial infection is to decrease the available iron by decreasing intestinal absorption and increasing liver storage. Do HBsAg carriers have a different response when faced with this stress? Are the increased iron levels a consequence of the HBV infection, or are individuals with high iron levels (or a propensity to develop them) more likely to become HBsAg carriers? Recently some relationships between HBV and bacteria have been observed. Incubation of HBsAg with Pseudomonas results in the inability to detect HBsAg by immunological methods, although the appearance of the surface antigen is not markedly changed (ref. 10; L. K. Weng and W. T. London, unpublished data). Millman and McMichael (11) have identified a glycoprotein present in various animal sera that can bind to HBsAg and prevent reaction with anti-HBs. Could the bacteria or glycoprotein contribute to increased levels of iron by preventing the recognition of HBsAg by the immune system? Millman et al. (12) have reported that transferrin is one of the host proteins associated with HBsAg. Transferrin saturated with iron (as opposed to unsaturated transferrin) is selectively bound to liver cells (13). The transferrin is incorporated (by pinocytosis), and iron is released into the cell. Iron-laden transferrin on the surface antigen of HBV may facilitate the entry of the virus into the hepatocyte. It would be interesting to know if iron is detectable in appreciable amounts in purified HBsAg. Perhaps HBV interacts with normal erythropoietic mechanisms to alter utilization of iron. If there are additional stores of iron in HBsAg carriers, does that enhance their propensity to develop or foster the growth of primary hepatocellular carcinoma, and, if so, could this information be useful in prevention or treatment of this disease? APPENDIX ON STATISTICAL METHODS The second part of the data analysis presented here was inductive in that there was no predetermined order imposed on the variables. Hence, the tests of significance ought to be considered as conjectures or new hypotheses rather than conclusions. An independent set of data is necessary to verify the conjectures. Nonetheless, a systematic approach to identifying the variables or combinations of variables that seem to present the most satisfactory explanation of the data is useful in providing clues to further research. The data consist of the number of individuals having com-

Proc. Nati. Acad. Sci. USA 76 (1979)

plete results on the four variables-iron (I), sex (S), HBV response (V), and SGPT (T). I and V have three possible categories, while S and T have two. Hence, a 3 X 2 X 3 X 2 contingency table results (Table 3). Logarithmic linear models are used to estimate expected values for each bin of the table. Then these are compared by maximum likelihood procedures to the observed values (14). The models are hierarchal; that is, if the IV interaction is present in the model, then I and V individually are included. The inclusion of a variable or combination of variables in a model implies that the estimated expected values are constrained to have the same marginal totals as the observed values. The analysis consisted of a sequence of hypotheses. First, a series of general hypotheses were tested in order to determine the highest order of interaction among the variables (d.f. is the degrees of freedom). x2 d.f. p Hi: All bins equally likely 60.83 6 0.15 6.12 The nonsignificance of H3 and H4 implies that attention should be directed to the two-factor interactions. Marginal and partial associations were computed for all the two-factor effects in order to decide which of these seem to be necessary (15). Partial Marginal association association p x2 d.f. x2 p H5: No IS effect 2 2.33 >0.25 1.19 >0.5 H6: No IV effect 4 12.5 0.05 H7: No IT effect 2 26.52 0.05 H9: No ST effect 1 0.26 >0.5 0.12 >0.5 H10: No VT 2 11.57 0.5 1.75 >0.25 H15: No IS effect 2 9.7