小鼠是研究人类疾病的重要模式动物之一，其基因组与人类基因组的相似性达到90%以上。一般认为，人类与小鼠在疾病发生的机理上基本相似，而两者在炎症响应基因表达谱方面也可以部分对应。

可是，在2013年初，美加两国多位科学家以“炎症和宿主损伤响应大型合作研究计划”（the Inflammation and Host Response to Injury, Large Scale Collaborative Research Program）的名义，在《美国国家科学院院报》（PNAS）上发表论文称，小鼠模型在基因组响应上极少模拟人类炎性疾病。

该文指出，尽管源自不同病因的急性炎症在人体中所导致的基因组响应十分相似，但小鼠模型中的基因组响应却与人体中的基因组响应很不相同。当人体中基因表达出现明显变化时，小鼠中的相应基因却呈随机性改变，相关系数仅为0-0.1，即完全不相关。因此，作者建议在人类炎性疾病的转化医学研究中应该更看重人体实验结果，而不是依据小鼠实验。

Genomic responses in mouse models poorly mimic human inflammatory diseases

Abstract

当这篇文章发表后，在科学界尤其是生物医学界引起巨大反响。《纽约时报》甚至一反常态地发表了足以耸人听闻的看法：“在错误的导向之下，宝贵的时间与无数的资金就这样被白白地浪费掉了”。更有甚者，学者们在使用小鼠疾病模型时提心吊胆，学生们则不知道没了小鼠他们如何完成毕业论文。相反，争取动物权益人士从此掌握了一个极具说服力的话柄。

然而，在本周刚刚出版的PNAS上，两位日本科学家发表最新论文，有力地反驳了美加科学家的观点，其论文题目更是针锋相对：小鼠模型在基因组响应上极大模拟人类炎性疾病。仅仅一字之差（poorly vs greatly），小鼠作为模式动物的命运几乎阴阳两隔，而造成这种“冰火两重天”的原因竟然仅仅是统计学方法的不同！

日本学者完全使用美加学者的转录组数据，但只是计算表达水平有显著差异的基因，结果求出人体数据与小鼠模型数据之间的相关系数在0.43-0.68之间，变化幅度为77%-93%，而且许多代谢途径的调节模式基本相同。因此，作者得出了完全相反的结论：小鼠可以用作人类炎性疾病模型动物！

Genomic responses in mouse models greatly mimic human inflammatory diseases

Abstract

The use of mice as animal models has long been considered essential in modern biomedical research, but the role of mouse models in research was challenged by a recent report that genomic responses in mouse models poorly mimic human inflammatory diseases. Here we reevaluated the same gene expression datasets used in the previous study by focusing on genes whose expression levels were significantly changed in both humans and mice. Contrary to the previous findings, the gene expression levels in the mouse models showed extraordinarily significant correlations with those of the human conditions (Spearman’s rank correlation coefficient: 0.43–0.68; genes changed in the same direction: 77–93%; P = 6.5 × 10⁻¹¹ to 1.2 × 10⁻³⁵). Moreover, meta-analysis of those datasets revealed a number of pathways/biogroups commonly regulated by multiple conditions in humans and mice. These findings demonstrate that gene expression patterns in mouse models closely recapitulate those in human inflammatory conditions and strongly argue for the utility of mice as animal models of human disorders.

实际上，在这篇反驳文章发表之前，美国本土就有科学家对“小鼠不适合作为人类炎性疾病模型”的说法提出质疑，其要点集中在单一品系小鼠与遗传多样性人体之间进行比较的合理性上。同时，也指出了其方法学上的不足：首先，采样时间不同。人体白细胞是在损伤后一年收集，而小鼠白细胞则在损伤后一周收集。未能使用平行信息，说基因组响应不相似就不准确。其次，探针敏感性不同，数据却同等对待。第三，计算错误。相关系数R可以是负数，相关系数的平方R2怎能是负数呢？最后，无重复验证。只使用芯片数据，未使用定量PCR与Western杂交数据。 

Concerns over interspecies transcriptional comparisons in mice and humans after trauma

This paper raises some legitimate issues regarding mouse vs. human differential and temporal responses at a gross whole blood level to biological perturbations occurring in other tissues. However, we are concerned that the failure to examine more than a single immune-polarized mouse strain, the lack of correction for differentially abundant cell types, and the use of data analysis approaches that did not consider these factors in an additive fashion strongly limit the conclusions that can be drawn from the data. We question the value of broad comparisons between genetically diverse patients and a single strain of mice. Male C57BL/6 inbred mice, which were used exclusively, have minimal genetic variation and are predisposed to Th1-mediated immune responses (2). Inclusion of additional mouse strains should have been considered in the study design to avoid pseudoreplication (3). Therefore, genetic background is the more appropriate unit of replication and not the individual patient or mouse.

Differences between human and mouse responses to traumas are likely exaggerated when time course data are used in aggregate rather than comparing biologically analogous time intervals and individual cell populations. Mouse and human leukocyte populations differ; thus, the majority of observed effects could be attributed to differential cell population margination or cell death rather than pathway-specific alterations of otherwise constant levels of different cell types. The authors apparently compared expression data from total blood leukocytes collected up to 1 y after injury in human patients vs. only 1 wk in murine models. Although it is certainly true that recovery rates between humans and mice differ, it is inaccurate to suggest that the genomic responses are dissimilar when parallel information is not used.

An additional issue is that data were analyzed without consideration of potential error resulting from variations in sensitivity among different probe sets for genes on the microarrays, which may have led to overestimation of expression divergence between species (4, 5). Further, although negative R values can indicate an inverse correlation, the coefficient of determination, R², cannot be negative (see figures 3 and 4 and table 1 in ref. 1). Finally, validation of expression levels using complementary methods (e.g., quantitative PCR or Western blotting) was not presented.

We believe these concerns must be addressed before broad conclusions can be drawn from the data. The current conclusions are especially troubling in light of the attention given to this study by the popular media. Although concerns exist regarding the utility of animal models in translating basic research to successful clinical treatments, we believe the current conclusions drawn from this manuscript are of questionable validity.

针对这个书面质疑，原作者一一进行了解答。对于单一小鼠品系的选择，考虑了三点因素：一是C57B/6是最常用的；二是其他品系对脂多糖有抗性；三是结果跟别人相似或一致。对于取样时间，认为小鼠基因恢复比人体疾病基因恢复明显不同，依时比较比单一比较更严格，其实他们进行了多种形式的比较，包括变化的方向性与程度、变化的响应时间与恢复时间、时间依存模式。关于技术及统计学问题，回答是不是进行原始数据比较，而是对变化数据进行比较，而对R2的解释已列在图表中。最后，原作者强调，他们并非否定所有小鼠模型，但只是认为小鼠作为炎性疾病模型不合适。

Reply to Osterburg et al.: To study human inflammatory diseases in humans

Osterburg et al. (1) raise questions from our recent publication (2). First, in our program, a single mouse strain, C57B/6, was selected because it has been the most commonly used in the field for decades. Furthermore, all strains of mice are remarkably resistant to LPS relative to humans. If anything, C57B/6 mice are less resistant and therefore potentially closer to the human response than many other mouse strains (3). Last, figure 4 and table 1 of ref. 2 show that our results are consistent with those of other independent mouse studies and not specific to strain, model, or investigator.

Osterburg et al. (1) question whether the choice of time intervals was appropriate. Gene recoveries in mouse models differ markedly compared with those in complex human diseases. In general, time course comparisons are much more rigorous than single time point or cross-sectional studies in capturing the similarities and differences in the gene changes between humans and mice (4). We performed the time course comparison and compared multiple characteristics of the response in humans and mice, including directionality and maximum magnitude of the changes (figures 1, 4, and S1 and table 1 of ref. 2), the response time and recovery time of the changes (figures 2B and S5 of ref. 2), and time course patterns (figures 2A, S6, and S7 of ref. 2). For both species, the gene response time occurred within the first 6–12 h (figure S5 of ref. 2).

Regarding the technical and statistical issues raised by Osterburg et al., instead of comparing the raw expression values from the microarrays between human and mouse, the comparisons were performed on the changes of expression values between disease conditions and controls within each species, where the same array platform was used for each of the species. The annotation of –R² was explained in the figure 3 legend and table 1 of ref. 2.

The myriad of ways that mice differ from humans including the different time intervals and the fact that mouse and human leukocyte cell populations differ raise the important question as to whether it is appropriate to try to adjust the model system to more closely compare similar features. To try to adjust or somehow correct for the mouse–human differences either in cell number, time course, or in any other way would introduce artifact. In our study, for example, had we adjusted the leukocyte populations, the genomics would not have reflected the in vivo condition.

Our article provides data for what most investigators already know from their experiences: current mouse models poorly reflect human inflammatory diseases. We are not damning all mouse models. Rather, we propose that the scientific community raise the bar to require model systems to more accurately reproduce the molecular features of human inflammatory disease and we should reprioritize our infrastructure, resources, tools, and methodologies to study human inflammatory diseases in humans.

个人感觉，从大的方面来说，仅仅依据转录组数据就得出小鼠不适合用作人类炎性疾病模型的结论是值得商榷的；从小的方面来说，基因芯片数据未得到定量PCR和Western杂交数据支持显然是不充分的。尽管这样的学术争论可能还会持续下去，但依据“唯一”数据得出“非凡”结论的做法肯定是不严谨的。

有意思的是，PNAS把两种完全相反的意见呈现出来让众人评判，毫不忌讳自己可能发表了至少一篇错误的论文。我认为这是良性的学术争论，科学的相对真理正是在这种理性思辨中逐渐逼近绝对真理的。

 蓝色的蒙古高原

（降央卓玛原唱，曾庆平翻唱）

出差在外，酒店录制，音效不佳，敬请海涵！