High degree of correlation between HBsAg and IL23/IL17?
Posted by Zhang_plos on 14 Oct 2013 at 03:04 GMT
After reading the article carefully, I feel obliged to write something about the almost perfect correlation between HBsAg and IL17/IL23. In this paper, Dr. Qihong Wang and his colleagues asserted a critical role of HBV surface antigen in stimulating IL-23 in intrahepatic antigen presenting cells and thus causing liver damage. The authors first quantified intrahepatic IL-23 and related cytokine gene expression in healthy controls, chronic HBV and ACLF patients. An elevated IL-23 and IL-17 is observed in CHB and ACLF patients liver specimens. They further demonstrated a high degree of correlation between HBsAg level (above 10000 IU/ml) and IL-23, IL-17 mRNA. However, after inspecting the correlation plot between HBsAg and IL23/IL17 mRNA(Figure 5), (1)we found that there are only 4 dots with IL-23 over 30 and 4 dots with IL17 over 60 (see annotated figure 5). In contrast, the other plots consistently displayed 8 dots in IL-23 and 7 dots in IL-17. Does this mean that there are four specimens with high IL-17 expression and three specimens with high IL-23 expression while HBsAg level is below 10000? If it is the case, this indicates that around half of the IL-23/IL-17 high samples have low HBsAg? In addition, by naked eye counting, we found that there are around 42 data points in HBsAg-IL-23 and IL-17 correlation plot while other plots have around 50 points. This means that 4 of the ~8 HBsAg low patients have high IL-23 and 3 of the ~8 HBsAg low patients have high IL-17?This is at odds with the theory presented by the authors. (2)Indeed, there is no point excluding samples with low HBsAg (<10000IU/ml) from correlation analysis since a wide range of surface antigen level is needed to perform a truly rigorous test. (3)Furthermore, figure 5 used quantification data from CHB patients' samples while no data is presented in ACLF ones. As we see in Figure 1, ACLF patients have a even high expression of IL-23. Why these data are not presented? All these evidence enable me to speculate that the authors have introduced biased criteria in their statistical analyses in order to fit their theory.
Response:
Thanks for the comments on our paper. All the above comments are focused on the correlation between IL-23/IL-17 and HBsAg, thus we response collectively by dividing the comments into three parts (red color number).
(1) We have stated in the text that only data from patients with HBsAg>10000 IU/ml were included in panel 3 of Figure 5 but other Figures include those data with HBsAg> and < 10000 IU/ml, thus it is necessary that the dot number in panel 3 of Figure 5 is different from those in other Figures. Hope we have clarified this issue.
(2) The reason why we exclude the data with HBsAg<10000 IU/ml in panel 3 of Figure 5 is just due to the truth that we find that the relationship between IL-23/IL-17 and HBsAg is significantly correlated only when the concentration of HBsAg is above 104. Therefore, we have not artificially set a criterion to exclude certain data, instead, we include only the data with HBsAg concentration above 10000, just based on the truth. We think we have stated clearly this issue in the paper.
(3) The reader is right that we only conducted experiments using CHB samples in Figure 5 because we think that we had got enough data from CHB samples. More importantly, the main novelty of this paper is the mechanisms for how HBsAg inducing the antigen presenting cells to produce IL-23, so we put more efforts to conduct the related experiments instead of exploring more clinical correlations, which will lower the impact of this paper.

There are also some minor points that should be clarified.
1. In Table 1, the author showed that 108 CHB patients were included in the study, however we only found around 50 data points in Figure 5. It is also true for Figure3,4. Does figure 3-5 used the same samples or different samples are used? This should be clarified. Similarly, 15 data points were found in Figure 3-5 for ACLF patients while Table 1 showed "58".
Response
In this study, a total of 166 HBV-infected patients, including 108 patients with CHB and 58 patients with ACLF, were enrolled. We collected the periphery blood or liver tissues of these patients. However, these samples including tissues and blood were used for IFC, IHC, Western Blot and in vitro co-culture experiments, respectively. However, not all samples will be used for all these experiments because the liver tissue samples from hepatitis B patients were too small to be used for several purposes. Therefore, the dot number in each Figure is inevitably not in accordance with that in Table 1. Therefore, the concrete sample number should be the dot number in each Figure. Specifically, in correlation analysis, the same samples were used. To compare the IL-17 levels in periphery and local tissues, the samples from same patient were used. In other experiments, we used all the samples that were correlated with the study, including tissues and periphery blood. As the dots in each Figure have been shown clearly, thus we don’t think we should clarify this issue additionally.

2. Table 1 indicated that 62 healthy controls were included in this study. However, the authors did not specify how they obtained so many healthy liver samples and how they performed ethics procedures. In our experience, obtaining truly healthy liver biopsies are practically difficult and require rigorous ethics evaluation the authors obtained healthy liver samples by other means (e.g. normal live tissues from surgical operation etc), it should be explicitly stated.
Response
Thanks for your reminding. There is no a word in this article indicating that we obtained 62 liver samples from healthy controls. Information in the Table 1 just indicates the total volunteer numbers that contributed liver tissues or blood involved in this study. Actually, we perform this study by 15 liver tissues and 47 blood samples from healthy controls. These samples come from the liver transplant donor or paracancerous tissue. Moreover, this study was approved by the ethics committee of the Third Military Medical University, Chongqing, China.
Taken together, to alleviate concerns in the solidity of the data and clarify some important facts, I think the most obvious option is that the authors need to provide a full datasheet containing patients' basic clinical parameters and their gene expression level. Indeed, the correlation plot between HBsAg and IL-23/IL-17 is crucial in connecting in vivo data with ex vivo stimulation assays. We urge that the authors should release the basic underlying data for Figure 1,3,5. I hope that the editorial office can fully consider the request of a HBV researcher in order to clarify the basic facts on this article.
Response
We think we have clarified the concerns from the reader.