When a DNA stock sample from a JNK2^-/- mouse was sent to us from JAX, PCR analysis revealed that the mouse was homozygous for the intact WT allele of nicotinamide nucleotide transhydrogenase (Nnt^+/+) and not the homozygous mutant Nnt^-/- allele, which is unique to the C57BL/6J substrain of C57BL/6 mice (5) (Table 1 and Fig. S1). This finding was confirmed when PCR analysis was repeated with DNA from tails of several JNK2^-/- mice more recently (results not shown), establishing beyond doubt that the JNK2^-/- colony at JAX was definitely not on a C57BL/6J (Nnt^-/-) background. We next assessed the effect of mispairing wild-type controls for the JNK2^-/- mice by comparing them to C57BL/6NJ (Nnt^+/+) and C57BL/6J (Nnt^-/-) mice. The designations 6J and 6N in the nomenclature of C57BL/6 substrains refers to them being derived from C57BL/6 colonies which were isolated from the Jackson Laboratory (6J) and sent to the National Institutes of Health (6N) in 1951(5). As the first step, we compared the susceptibility of C57BL/6J (Nnt^-/-) and C57BL/6NJ (Nnt^+/+) substrains to AILI. C57BL/6J mice were less susceptible to AILI than the C57BL/6NJ substrain, as determined biochemically by the measurement of serum ALT activities (Fig. 1A), a biomarker of liver injury (15), and histochemically by reduced amounts of hepatic perivenous necrosis (Fig. 1B). Lastly and most importantly, we confirmed the results of our original study that JNK2^-/- mice from JAX were more susceptible than C57BL/6J WT mice to AILI (14) and also showed that this finding could be reversed when JNK2^-/- mice were paired with C57BL/6NJ (Nnt^+/+) WT mice (Fig. 1). Overall, these results support the findings by other researchers as discussed earlier that JNK2 has a pathologic role in AILI.

Our findings raised the possibility that mispairings of JNK2^-/- and C57BL/6J WT controls mice from JAX in other studies dealing with JNK2 signaling may have also led to inaccurate interpretations of data (see Supporting Information for such misparings). A case in point is the conflicting role of JNK2 in a T-cell model of liver injury induced by concanavalin A that can also be explained by these mismatches (Fig. S2).

The studies with JNK2^-/- mice prompted us to explore whether similar C57BL/6 background problems might also have occurred in investigations with other GEM. We picked an additional 79 GEM from JAX for our studies based upon the recommendation by JAX that C57BL/6J WT mice could be used as controls and the wide-spread use of these strains in studies dealing with innate and adaptive immune systems in physiology and pathology. When DNA samples from each of the GEM were genotyped, 26 of them were found to be either homozygous Nnt^+/+ (19 samples) or heterozygous Nnt^+/- (7 samples) (Table 1 and Fig. S1), while all others were homozygous Nnt^-/- (Table S1). Since 12 of the 26 in addition to the JNK2^-/- mice had not been backcrossed again to any C57BL/6 substrain as of April 2011 according to information from JAX websites for each of the GEM strains (Table 1), it is likely that these colonies remain Nnt^+/+ except possibly for the heterozygous Nnt^+/- strains. Similarly, among the 19 GEM strains that were found Nnt^+/+, the 6 that were backcrossed only one extra time to C57BL/6 substrain (Table 1) would be on Nnt mixed background. Together, this suggests that many strains are still incorrectly mispaired with C57BL/6J and therefore could lead to confounding results.

Although recent studies have not uncovered genetic differences among substrains of C57BL/6N (Nnt^+/+) mice (5, 8) it is possible that more complete genetic analyses in the future may reveal genetic diversity within these substrains that results in phenotypic differences among C57BL/6N substrains. If this were case, then mismatches in C57BL/6N GEM and WT controls may also be a source of inaccuracies in the literature. Another potential complicating factor is that genomic copy number variations may also lead to phenotypic differences within and between substrains of mice (16, 17). As it is known that genetic and phenotypic differences can also exist within strains of other mice that have diverged at various sites around the world, including the widely used Balb/c mouse (16, 18), it is possible that background mismatches of GEM with WT controls have occurred with these mice as well.

Our findings underscore the critical need for researchers to be extremely careful when designing experiments with GEM and WT controls on a C57BL/6 background, obtained from vendors and other researchers. Investigators can help avoid mispairing C57BL/6J (Nnt^-/-) and C57BL/6N (Nnt^+/+) substrains by Nnt genotyping mice prior to beginning experiments. This problem could also be alleviated if journals required authors to include in their manuscripts information concerning the source and C57BL/6 substrain background of GEM including number of times backcrossed as well as and relevant information about the WT controls used in their studies. Similarly, vendors of GEM should report similar information and as well as details of any further backcrossing of their GEM onto C57BL/6 WT mice on their websites. However, it always best to choose control WT mice that are either age-matched WT littermates of GEM or age-matched WT mice on to which the GEM were backcrossed (19). This approach would have been the optimal choice to show that JNK2^-/- mice are less susceptible than WT controls to AILI.

We have also found in two distinctive animal models of liver pathology, AILI and concanavalin A-induced liver injury, that C57BL/6N mice are more susceptible than C57BL/6J mice to liver injury (Figs.1 and S2, respectively). These results were surprising as mitochondrial oxidative stress plays a pathologic role in both models of liver injury (20, 21), and because the Nnt mutation in C57BL/6J mice encodes for an inactive form of mitochondrial Nnt that normally catalyzes the interconversion of NADH into NADPH (22), which has an important antioxidant role in regenerating reduced glutathione and thioredoxin during mitochondrial oxidative stress (22). Moreover, recent studies show that additional genetic differences exist between C57BL/6J (Nnt^-/-) and C57BL/6 (Nnt^+/+) substrains(5, 8). It is anticipated that mechanistic studies in the future that can unravel the susceptible differences of C57BL/6 substrains mice to AILI and concanavalin A-induced liver injury could lead to the identification of potential risk factors for not drug-induced liver injury, but also for liver injury caused by other pathologies.

1_si_001