Correction MEDICAL SCIENCES Correction for “Genomic responses in mouse models greatly mimic human inflammatory diseases, ” by Keizo Takao and Tsuyoshi Miyakawa, which appeared in issue 4, January 27, 2015, of Proc Natl Acad Sci USA (112:1167–1172; first published August 8, 2014; 10.1073/pnas.1401965111). The authors note the following corrections: Fig. 1 and its corresponding legend appeared incorrectly be- cause genes were inappropriately included for the analyses for two reasons. First, some genes from the Human Burn and Human Trauma dataset were inappropriately included due to data handling errors. Second, the data originally presented in Fig. 1 included all genes with an absolute fold change (FC) greater than 1.2 for both human and mouse conditions. However, the genes with jFCj > 2.0 in human conditions and jFCj > 1.2 in mouse conditions should have been used, as in Fig. 3. Genes that meet the same criteria are appropriately analyzed, and corrected data are now presented in Fig. 1. As a result of this change, on page 1167, left column, in the Abstract, lines 10–12, “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 )” should instead appear as “Spearman’s rank correlation coefficient: 0.48–0.68 in Fig. 1; significance of overlap: P = 6.5 × 10 -11 to 1.2 × 10 -35 in Fig. 2; genes changed in the same direction: 59.5–93.2% in Fig. 3.” Also as a result of this, the third paragraph on the left column of page 1168, starting with “We conducted a Kolmogorov- Smirnov test” should instead appear as: “We conducted a Kolmogorov–Smirnov test to check for the normality of the distribution of gene expression data in human burn conditions, which were compared with mouse models as a reference dataset. The assumption of normality was rejected either for the fold changes or for the log-twofold changes of the gene expression levels (P < 0.0001). Therefore, we mainly used nonparametric Spearman’s correlation coefficient (ρ) for the correlation anal- yses. The criteria for the selection of the genes of interest was absolute fold change >2.0 in human diseases and >1.2 in mouse conditions, and P < 0.05 in both conditions. The correlations of the gene changes as assessed by Spearman’s correlation co- efficient indicated that there were highly significant similarities in gene responses between each of the human conditions and those of the mouse models (Fig. 1; ρ = 0.48–0.68, P < 0.0001 for every comparison between human conditions and the corresponding mouse models). There were also highly significant correlations among different mouse models (Fig. 1; ρ = 0.23–0.84, P < 0.0001 for every comparison between a pair of mouse models).” Also as a result of this, on page 1171, right column, in the third full paragraph, lines 5 and 6 “In Fig. 1, genes meeting the criteria of P < 0.05 and fold change >1.2 are plotted in the graph” should instead appear as “In Fig. 1, genes meeting the criteria of absolute fold change >2.0 in human diseases and >1.2 in mouse models, and P < 0.05 in both conditions, are plotted in the graph. ” Fig. 2 also appeared incorrectly due to a copy and paste error. In addition, on page 1172, left column, first paragraph, lines 3 and 4, “genes with a P value of 0.05 of less and an absolute fold change of 1.2 or greater were used” should instead appear as “genes with a P value < 0.05 and an absolute fold change >1.2 were used.” Fig. 3 also appeared incorrectly due to a few errors in gene selection and in typing. As a result on page 1168, right column, second full paragraph, lines 9–11, “Fig. 3; human: fold change >2.0; mouse model: fold change >1.2; R = 0.26–0.51; P < 0.0001 for all comparisons; percentage: 48.1–86.2” should instead ap- pear as “Fig. 3; human: absolute fold change >2.0; mouse model: absolute fold change >1.2; R = 0.26–0.57; P < 0.0001 for all comparisons; percentage: 59.5–86.2.” Additionally, on page 1172, left column, first paragraph, lines 7 and 9, “P value of 0.05 or less” should instead appear as “P value < 0.05. ” Also as a result of this, Table S2 appeared incorrectly. Please see separate SI Correction. Fig. 4 also appeared incorrectly due to data input error. As a result, the paragraph on pages 1169–1170 starting with “Some of the pathways/biogroups with high overlap are shown” should instead appear as: “Some of the pathways/biogroups with high overlap are shown in Fig. 4 A–D. The significance of the overlap between each condition and pathways/biogroups is also shown in the zgraph. There was significant overlap between genes anno- tated in GO as ‘ innate immune response’ and the genes up-regulated in the mouse models of burn (Fig. 4A, P = 6.7 × 10 -24 ), sepsis (P = 4.6 × 10 -33 ), and infection (P = 1.3 × 10 -16 ), as well as in human burn (P = 4.8 × 10 -54 ), trauma (P = 4.1 × 10 -90 ), and sepsis conditions (P = 6.3 × 10 -21 ). Significant overlap was also detected between ‘genes involved in cytokine signaling in immune system (canonical pathways, Broad MSigDB)’ and genes up-regulated in the mouse models of sepsis (Fig. 4B, P = 4.0 × 10 -36 ), burn (P = 1.6 × 10 -11 ), and infection (P = 1.3 × 10 -8 ), as well as in human burn (P = 1.6 × 10 -32 ), trauma (P = 6.5 × 10 -52 ), and sepsis conditions (P = 3.2 × 10 -12 ). With regard to down-regulated pathways/biogroups, genes annotated ‘lymphocyte differentiation (GO)’ significantly overlapped with genes down- regulated in the mouse models of burn (Fig. 4C, P = 2.2 × 10 -8 ), trauma (P = 1.4 × 10 -15 ), sepsis (P = 4.2 × 10 -20 ), and infection (P = 1.3 × 10 -18 ), as well as in human burn (P = 1.7 × 10 -18 ), trauma (P = 5.6 × 10 -12 ), and sepsis conditions (P = 3.5 × 10 -30 ). There was also significant overlap between ‘genes involved in Translocation of ZAP-70 to immunological synapse (canonical pathways, Broad MSigDB)’ and genes down-regulated in all of the human disease conditions and mouse models of these conditions (Fig. 4D, human burn, P = 1.3 × 10 -11 ; human trauma, P = 0.0003; human sepsis, P = 8.7 × 10 -25 ; mouse burn, P = 0.0003; mouse trauma, P = 5.1 × 10 -10 ; mouse sepsis, P = 1.3 × 10 -7 ; and mouse infection, P = 1.8 × 10 -9 ).” Lastly, a portion of the Materials and Methods appeared in- correctly. The paragraph on the right column of page 1171 starting with “The datasets that we analyzed in the present study” should instead appear as: “The datasets that we analyzed in the present study were the same as those used in the study by Seok et al. (1) and are registered in NextBio. In Fig. 1, the following datasets were used for gene expression pattern analyses: ‘leukocytes of patients with severe burns on >20% of total body surface area vs. healthy controls’ from GSE37069 is referred to as ‘Human Burn’; ‘white blood cells of severe blunt trauma patients 28 d after injury vs. healthy subjects’ from GSE36809 is referred to as ‘Human Trauma’; ‘whole blood of sepsis patients with community-acquired www.pnas.org PNAS | March 10, 2015 | vol. 112 | no. 10 | E1163–E1167 CORRECTION Downloaded by guest on March 21, 2021 Downloaded by guest on March 21, 2021 Downloaded by guest on March 21, 2021 Downloaded by guest on March 21, 2021 Downloaded by guest on March 21, 2021 Downloaded by guest on March 21, 2021 Downloaded by guest on March 21, 2021 Downloaded by guest on March 21, 2021
Transcript
Correction
MEDICAL SCIENCESCorrection for “Genomic responses in mouse models greatlymimic human inflammatory diseases,” by Keizo Takao and TsuyoshiMiyakawa, which appeared in issue 4, January 27, 2015, of Proc NatlAcad Sci USA (112:1167–1172; first published August 8, 2014;10.1073/pnas.1401965111).The authors note the following corrections:Fig. 1 and its corresponding legend appeared incorrectly be-
cause genes were inappropriately included for the analyses for tworeasons. First, some genes from the Human Burn and HumanTrauma dataset were inappropriately included due to datahandling errors. Second, the data originally presented in Fig. 1included all genes with an absolute fold change (FC) greaterthan 1.2 for both human and mouse conditions. However, thegenes with jFCj > 2.0 in human conditions and jFCj > 1.2 inmouse conditions should have been used, as in Fig. 3. Genes thatmeet the same criteria are appropriately analyzed, and correcteddata are now presented in Fig. 1.As a result of this change, on page 1167, left column, in the
Abstract, lines 10–12, “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)” should instead appear as “Spearman’srank correlation coefficient: 0.48–0.68 in Fig. 1; significance ofoverlap: P = 6.5 × 10−11 to 1.2 × 10−35 in Fig. 2; genes changedin the same direction: 59.5–93.2% in Fig. 3.”Also as a result of this, the third paragraph on the left column
of page 1168, starting with “We conducted a Kolmogorov-Smirnov test” should instead appear as: “We conducted aKolmogorov–Smirnov test to check for the normality of thedistribution of gene expression data in human burn conditions,which were compared with mouse models as a reference dataset.The assumption of normality was rejected either for the foldchanges or for the log-twofold changes of the gene expressionlevels (P < 0.0001). Therefore, we mainly used nonparametricSpearman’s correlation coefficient (ρ) for the correlation anal-yses. The criteria for the selection of the genes of interest wasabsolute fold change >2.0 in human diseases and >1.2 in mouseconditions, and P < 0.05 in both conditions. The correlations ofthe gene changes as assessed by Spearman’s correlation co-efficient indicated that there were highly significant similaritiesin gene responses between each of the human conditions andthose of the mouse models (Fig. 1; ρ = 0.48–0.68, P < 0.0001 forevery comparison between human conditions and the correspondingmouse models). There were also highly significant correlationsamong different mouse models (Fig. 1; ρ = 0.23–0.84, P < 0.0001 forevery comparison between a pair of mouse models).”Also as a result of this, on page 1171, right column, in the third
full paragraph, lines 5 and 6 “In Fig. 1, genes meeting the criteria ofP < 0.05 and fold change >1.2 are plotted in the graph” shouldinstead appear as “In Fig. 1, genes meeting the criteria of absolutefold change >2.0 in human diseases and >1.2 in mouse models, andP < 0.05 in both conditions, are plotted in the graph.”Fig. 2 also appeared incorrectly due to a copy and paste error.
In addition, on page 1172, left column, first paragraph, lines 3and 4, “genes with a P value of 0.05 of less and an absolute foldchange of 1.2 or greater were used” should instead appear as“genes with a P value < 0.05 and an absolute fold change >1.2were used.”
Fig. 3 also appeared incorrectly due to a few errors in geneselection and in typing. As a result on page 1168, right column,second full paragraph, lines 9–11, “Fig. 3; human: fold change>2.0; mouse model: fold change >1.2; R = 0.26–0.51; P < 0.0001for all comparisons; percentage: 48.1–86.2” should instead ap-pear as “Fig. 3; human: absolute fold change >2.0; mouse model:absolute fold change >1.2; R = 0.26–0.57; P < 0.0001 for allcomparisons; percentage: 59.5–86.2.”Additionally, on page 1172, left column, first paragraph, lines
7 and 9, “P value of 0.05 or less” should instead appear as“P value < 0.05.”Also as a result of this, Table S2 appeared incorrectly. Please
see separate SI Correction.Fig. 4 also appeared incorrectly due to data input error. As
a result, the paragraph on pages 1169–1170 starting with “Someof the pathways/biogroups with high overlap are shown” shouldinstead appear as: “Some of the pathways/biogroups with highoverlap are shown in Fig. 4 A–D. The significance of the overlapbetween each condition and pathways/biogroups is also shown inthe zgraph. There was significant overlap between genes anno-tated in GO as ‘innate immune response’ and the genesup-regulated in the mouse models of burn (Fig. 4A, P = 6.7 ×10−24), sepsis (P = 4.6 × 10−33), and infection (P = 1.3 × 10−16),as well as in human burn (P = 4.8 × 10−54), trauma (P = 4.1 ×10−90), and sepsis conditions (P = 6.3 × 10−21). Significantoverlap was also detected between ‘genes involved in cytokinesignaling in immune system (canonical pathways, Broad MSigDB)’and genes up-regulated in the mouse models of sepsis (Fig. 4B, P =4.0 × 10−36), burn (P = 1.6 × 10−11), and infection (P = 1.3 × 10−8),as well as in human burn (P = 1.6 × 10−32), trauma (P = 6.5 ×10−52), and sepsis conditions (P = 3.2 × 10−12). With regard todown-regulated pathways/biogroups, genes annotated ‘lymphocytedifferentiation (GO)’ significantly overlapped with genes down-regulated in the mouse models of burn (Fig. 4C, P = 2.2 × 10−8),trauma (P = 1.4 × 10−15), sepsis (P = 4.2 × 10−20), and infection(P = 1.3 × 10−18), as well as in human burn (P = 1.7 × 10−18),trauma (P = 5.6 × 10−12), and sepsis conditions (P = 3.5 × 10−30).There was also significant overlap between ‘genes involved inTranslocation of ZAP-70 to immunological synapse (canonicalpathways, Broad MSigDB)’ and genes down-regulated in all of thehuman disease conditions and mouse models of these conditions(Fig. 4D, human burn, P = 1.3 × 10−11; human trauma, P = 0.0003;human sepsis, P = 8.7 × 10−25; mouse burn, P = 0.0003; mousetrauma, P = 5.1 × 10−10; mouse sepsis, P = 1.3 × 10−7; and mouseinfection, P = 1.8 × 10−9).”Lastly, a portion of the Materials and Methods appeared in-
correctly. The paragraph on the right column of page 1171starting with “The datasets that we analyzed in the present study”should instead appear as: “The datasets that we analyzed in thepresent study were the same as those used in the study by Seoket al. (1) and are registered in NextBio. In Fig. 1, the followingdatasets were used for gene expression pattern analyses: ‘leukocytesof patients with severe burns on >20% of total body surface area vs.healthy controls’ from GSE37069 is referred to as ‘Human Burn’;‘white blood cells of severe blunt trauma patients 28 d after injuryvs. healthy subjects’ from GSE36809 is referred to as ‘HumanTrauma’; ‘whole blood of sepsis patients with community-acquired
infection vs. healthy subjects’ from GSE28750 is referred to as‘Human Sepsis’; ‘WBC from blood at 7 d after burn injury vs.burn injury sham’ from GSE7404 is referred to as ‘Mouse Burn’;‘WBC from blood at 3 d after trauma hemorrhage vs. traumahemorrhage sham’ from GSE7404 is referred to as ‘MouseTrauma’; ‘Blood of C57BL6J mice 4 h after Staphylococcusaureus infection vs. uninfected’ from GSE19668 is referred to as‘Mouse Sepsis’; ‘Blood from 8-wk-old BALB-c mice 1 d after tail
vein injection - Candida albicans vs. saline control’ from GSE20524is referred to as ‘Mouse Infection.’ Datasets used in Figs. 2–4 werespecified in Table S1, Table S2, and in the Materials and Methodssubsection Comparison of Pathways/Biogroups Altered in the HumanDiseases and Mouse Models, respectively. Only genes with a P value<0.05 and an absolute fold change >1.2 were considered to bedifferentially expressed.”All corrected figures and their corrected legends appear below.
Fig. 1. Correlations of gene changes among human burns, trauma, sepsis, and the corresponding mouse models. Scatterplots and Spearman’s rank corre-lations (ρ) of the fold changes. The criteria for gene selection were as follows: absolute fold change > 2.0 in human diseases, absolute fold change > 1.2 inmouse models, P < 0.05 in both conditions. Vertical bar and horizontal bar for each panel represents fold change in right and upper panels, respectively.N represents the number of probes differentially expressed in both conditions of the comparison in each panel. Murine models were highly significantlycorrelated with human conditions with Spearman’s correlation coefficient (ρ = 0.48–0.68; P < 0.0001 for every comparison between human conditions andmouse models). The correlations between different mouse models were also significant (ρ = 0.23–0.84; P < 0.0001 for every comparison).
E1164 | www.pnas.org
Dow
nloa
ded
by g
uest
on
Mar
ch 2
1, 2
021
Human Burn vs. Mouse Burn (GSE7404)
Human Burn vs. Mouse Sepsis(GSE26472)
Human Burn vs. Mouse Infection(GSE20524)
Human Burn vs. Mouse Sepsis(GSE19668)
Human Burn vs. Mouse Trauma(GSE7404)
940 genesHuman
Overlap P value:
3.9 × 10-34
Overlap P value:
6.3 × 10-13
Overlap P value:
6.5 × 10-11
Overlap P value:
1.2 × 10-35
Overlap P value:
3.4 × 10-35
↑↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
Human↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
267 genes 397 genes852 genes
578 genes 212 genes 513 genes221 genes
564 genesHuman↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
353 genes 237 genes1044 genes
771 genesHuman↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
Human↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
773 genes 441 genes1266 genes
593 genes 482 genes 816 genes591 genes
BA
C
E
D
0
10
20
30
40
50
60
-log
(p-v
alue
)
0
5
10
15
20
-log
(p-v
alue
)0
10
20
30
40
50
60
-log
(p-v
alue
)
0
5
10
15
20
-log
(p-v
alue
)
0
10
20
30
40
-log
(p-v
alue
)
Fig. 2. Statistical comparison of the direction of the gene expression changes between human burns and mouse models. Vertical bar represents the sig-nificance of the overlap between gene sets. Genes whose expression levels were changed in human burns significantly overlapped with those in the conditionof mouse burn (A, overlap P value = 3.9 × 10−34), mouse trauma (B, 6.3 × 10−13), mouse sepsis from GSE19668 (C, 1.2 × 10−35), mouse sepsis from GSE26472(D, 6.5 × 10−11), and mouse infection (E, 3.4 × 10−35). Value is expressed as the –log 10 of the P value. Statistical significances regarding the directionality of thegene expression changes were derived from the nonparametric ranking method provided by the bioinformatics platform NextBio.
PNAS | March 10, 2015 | vol. 112 | no. 10 | E1165
CORR
ECTION
Dow
nloa
ded
by g
uest
on
Mar
ch 2
1, 2
021
40
50
60
70
80
90
100
0 0.2 0.4 0.6 0.8 1Gen
es c
hang
ed to
the
sam
e di
rect
ion
(%)
Correlation of fold changes of genes (R)
Human diseases (Human: |FC| > 2.0, Human: all, Seok et al.)
Mouse model (Human: |FC| > 4.0, Mouse: |FC| >1.2)
Mouse model (Human: |FC| >2.0, Mouse: |FC| >1.2)
Mouse model (Human: |FC| > 2.0, Mice: all, Seok et al.)
Fig. 3. Comparison of the genomic response to severe acute inflammation from different etiologies in human and murine models. The datasets that wereused in Seok et al. (1) and are registered in NextBio were reevaluated using the criteria of Seok et al. (1) to select the genes of interest. We chose genes withsignificant responses in both the human burn dataset and in the mouse dataset for comparison, whereas Seok et al. chose the 4,918 genes with significantresponses in the human datasets regardless of the significance of the changes in the genes in mice. The datasets used here are listed in Table S2. Shown arePearson’s correlations (R; x axis) and directionality (%; y axis) of the gene response from multiple published datasets in GEO compared with human burns.Note that R2 data taken from Seok et al. were recalculated to obtain the R values shown here for comparison (blue symbols). Most of the mouse modelsshowed a high directionality score (random chance is 50%), indicating that their gene expression patterns are similar to the that in human burns, as long asstringent criteria are applied to the selection of the genes of interest.
E1166 | www.pnas.org
Dow
nloa
ded
by g
uest
on
Mar
ch 2
1, 2
021
www.pnas.org/cgi/doi/10.1073/pnas.1502188112
activated suppressed
activated suppressed
activated suppressed
activated suppressed
Significance of overlaps with the Broad MSigDB biogroup (- log)
Significance of overlaps with the Broad MSigDB biogroup (- log)
Significance of overlaps with the GO biogroup (- log)
Significance of overlaps with the GO biogroup (- log)
Genes involved in Cytokine Signaling in Immune system
Innate immune response
Lymphocyte differentiation
Genes involved in Translocation of ZAP-70 to Immunological synapse
B
A
C
D
Mouse InfectionMouse Sepsis
Mouse TraumaMouse Burn
Human SepsisHuman Trauma
Human Burn
-10 0 10 20 30 40 50 60 70 80 90
Mouse InfectionMouse Sepsis
Mouse TraumaMouse Burn
Human SepsisHuman Trauma
Human Burn
-10 0 10 20 30 40 50 60
Mouse InfectionMouse Sepsis
Mouse TraumaMouse Burn
Human SepsisHuman Trauma
Human Burn
-10 -5 0 5 10 15 20 25 30
Mouse InfectionMouse Sepsis
Mouse TraumaMouse Burn
Human SepsisHuman Trauma
Human Burn
-5 0 5 10 15 20 25
Fig. 4. Representative pathways or biogroups shared by the human diseases and mouse models. (A–D) Shown are bar graphs of statistical significance(−log 10 P value) for the representative pathways with significant regulation in the human and murine models. (A) Genes annotated as “innate immuneresponse (GO)” significantly overlapped with genes up-regulated in the mouse models as well as in human diseases. (B) Significant overlap was also detectedbetween “genes involved in cytokine signaling in immune system (canonical pathways, Broad MSigDB)” and genes up-regulated in both the mouse modelsand human diseases. (C) Genes annotated “lymphocyte differentiation (GO)” and genes down-regulated in the mouse models and human diseases signif-icantly overlapped. (D) There was also significant overlap between “genes involved in translocation of ZAP-70 to immunological synapse (canonical pathways,Broad MSigDB).” As can be seen, in the pathways/biogroups shown here, the mouse models are comparable to the human conditions in the significance ofenrichment between each condition and pathways/biogroups. For a complete list of pathways/biogroups shared by the human diseases and mouse modelssee Dataset S1.
Genomic responses in mouse models greatly mimichuman inflammatory diseasesKeizo Takaoa,b and Tsuyoshi Miyakawaa,b,c,1
aSection of Behavior Patterns, Center for Genetic Analysis of Behavior, National Institute for Physiological Sciences, Okazaki, Aichi 444-8585, Japan;bCore Research for Evolutional Science and Technology, Japan Science and Technology Agency, Kawaguchi, Saitama 332-0012, Japan; and cDivision of SystemsMedical Science, Institute for Comprehensive Medical Science, Fujita Health University, Toyoake, Aichi 470-1192, Japan
Edited by Ruslan Medzhitov, Yale University School of Medicine, New Haven, CT, and approved June 11, 2014 (received for review January 31, 2014)
The use of mice as animal models has long been considered es-sential in modern biomedical research, but the role of mousemodels in research was challenged by a recent report that genomicresponses in mouse models poorly mimic human inflammatorydiseases. Here we reevaluated the same gene expression datasetsused in the previous study by focusing on genes whose expressionlevels were significantly changed in both humans and mice.Contrary to the previous findings, the gene expression levels inthe mouse models showed extraordinarily significant correlationswith those of the human conditions (Spearman’s rank correlationcoefficient: 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 ofthose datasets revealed a number of pathways/biogroups com-monly regulated by multiple conditions in humans and mice. Thesefindings demonstrate that gene expression patterns in mousemodels closely recapitulate those in human inflammatory condi-tions and strongly argue for the utility of mice as animal models ofhuman disorders.
The use of mice as animal models of human disorders has longbeen considered essential for elucidating the underlying
mechanisms of disease, as well as for translational research frombench to bedside. This notion was seriously challenged by the recentreport by Seok et al. (1) that genomic responses to different acuteinflammatory stressors are highly similar in humans but very poorlyreproduced in the corresponding mouse models. Seok et al. (1)investigated gene expression changes in individuals with trauma,burns, and endotoxemia and compared them with those in mousemodels of these conditions. Surprisingly, in their study there werefew correlations between the gene expression changes in the humanconditions and those in mouse models, whereas the gene expressionpatterns were similar between humans with trauma, burns, andendotoxemia. Based on their findings, Seok et al. (1) concluded thathigher priority should be focused on studying complex humanconditions directly rather than relying on mouse models to studyhuman inflammatory diseases. The study has drawn much attentionfrom researchers (2–10) and mass media (11–14) and has been citedmore than 360 times since its publication only 18 months ago. Alt-metric analysis of the article indicates that it is among the top 1% ofall publications tracked by the system (as of January 20, 2014). Mostof the reactions were an expression of general concern regarding theutility of mouse models for biomedical research.In the present study, we reevaluated the same gene expression
datasets analyzed in Seok et al. (1) using more conventionalstatistical methods. There are a few critical differences in theanalysis methods used between the previous study and ours.First, we focused on genes whose expression levels were signifi-cantly changed in both humans and mice. The previous studyanalyzed sets of genes that were significantly changed in thehuman conditions regardless of the significance of the changes inmouse models for comparison, which is not a conventionalmethod of comparing two gene expression datasets. Assumingthat mouse models would mimic only partial aspects of humandisorders, inclusion of genes that showed no significant response
to the stimulus would generally decrease the sensitivity to detectthe responses shared by the disorders and their models. For thisreason, we excluded such genes from our analysis. Second, wecompared each of the conditions in a single mouse study inde-pendently with the human reference conditions. Mouse studies,such as GSE7404 and GSE19668, included multiple conditions orgene sets. For example, GSE19668 contains multiple datasets,including those for two different mouse strains and multiple time-course data points after infection. Because humans and mice areexpected to be quite different, the optimal conditions/parametersthat most closely mimic human conditions should be rigorouslysearched and considered the best model when trying to establishany animal model of a human disorder. Therefore, among suchmultiple conditions in a mouse study we chose the gene set with thehighest similarity to the human reference condition and used thisset for further analyses. Third, we mainly used Spearman’s rankcorrelation coefficient (or Spearman’s ρ), instead of Pearson’s cor-relation coefficient (R) or correlation coefficient of determination(R2), because there is no reason to assume linearity and normaldistribution of fold changes or log-twofold changes of gene expres-sion levels. Fourth, we used a bioinformatics tool, NextBio (15),to conduct more sophisticated nonbiased statistical analyses of thesimilarity between gene sets. NextBio employs a normalized rankingapproach, which enables comparability across data from differentstudies, platforms, and analysis methods by removing dependence onabsolute values of fold change and minimizing some of the effects ofthe normalization methods used and platform effects. We also usedNextBio to conduct meta-analyses of the human and mouse gene setsto determine the pathways/biogroups that were commonly changedin the human and mouse gene sets.
Significance
The role of mouse models in biomedical research was recentlychallenged by a report that genomic responses in mousemodels poorly mimic human inflammatory diseases. Here wereevaluated the same gene expression datasets used in theprevious study by focusing on genes whose expression levelswere significantly changed in both humans and mice. Contraryto the previous findings, the gene expression patterns in themouse models showed extraordinarily significant correlationswith those of the human conditions. Moreover, many path-ways were commonly regulated by multiple conditions inhumans and mice. These findings demonstrate that gene ex-pression patterns in mouse models closely recapitulate those inhuman inflammatory conditions and strongly argue for theutility of mice as animal models of human disorders.
Author contributions: T.M. designed research; K.T. and T.M. performed research; K.T. andT.M. analyzed data; and K.T. and T.M. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: [email protected].
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1401965111/-/DCSupplemental.
ResultsCorrelation of Gene Changes to Trauma, Burns, and EndotoxemiaBetween Human Subjects and Murine Models. The fold changes ofgene expression between patients and healthy subjects for eachdataset of human burn (GSE37069), trauma (GSE36809), andsepsis (GSE28750) conditions and between treated and controlgroups for the murine models (GSE7404 for burn and trauma,GSE19668 for sepsis, and GSE20524 for infection) were obtainedfrom a bioinformatics platform, NextBio.Between any two datasets, the agreements of the gene fold
changes (Spearman’s rank correlation between the two datasetsfor Fig. 1 and Pearson’s correlation for Fig. 3) as well as the di-rectionality of the changes (P values derived from nonparametricranking method by NextBio for Fig. 2 and the percentages of geneschanged in the same direction between the two datasets for Fig. 3)were compared.We conducted a Kolmogorov–Smirnov test to check for the
normality of the distribution of gene expression data in humanburn conditions, which were compared with mouse models asa reference dataset. The assumption of normality was rejectedeither for the fold changes or for the log-twofold changes of thegene expression levels (P < 0.0001). Therefore, we mainly usednonparametric Spearman’s correlation coefficient (ρ) for thecorrelation analyses. The criteria for the selection of the genes ofinterest was fold change >1.2 and P < 0.05 within each conditionto be compared. The correlations of the gene changes as assessedby Spearman’s correlation coefficient indicated that there werehighly significant similarities in gene responses between each ofthe human conditions and those of the mouse models (Fig. 1; ρ =0.43–0.68, P < 0.0001 for every comparison between humanconditions and the corresponding mouse models). There were alsohighly significant correlations among different mouse models
(Fig. 1; ρ = 0.47–0.57, P < 0.0001 for every comparison betweena pair of mouse models).Nonparametric ranking analysis by NextBio rejected, with
extraordinarily high confidence, the hypothesis that mousemodels show only coincidental overlap of the directionality ofgene changes with those in human burn conditions [Fig. 2;overlap P value = 3.9 × 10−34, 6.3 × 10−13, 1.2 × 10−35, 6.5 × 10−11,and 3.4 × 10−35 for mouse models of burn, trauma, sepsis(GSE19668), sepsis (GSE26472), and infection, respectively],demonstrating that the directionality of the changes in mousemodels was highly similar to that in human burn conditions.Outputs of these analyses using NextBio are available fromthe URLs shown in Table S1.Although the data violate the assumption of a normal distri-
bution, we created a scatterplot for the responses to inflammationfrom different etiologies in humans and murine models (Fig. 3)using Pearson’s correlation coefficient R and percentages of geneschanged in the same direction to compare with figure 4 in Seoket al. (1). Contrary to the conclusion of Seok et al., there weresignificant correlations and similarities of the direction in the geneexpression changes between human burn conditions and mousemodels (Fig. 3; human: fold change >2.0; mouse model: foldchange >1.2; R = 0.26–0.51; P < 0.0001 for all comparisons;percentage: 48.1–86.2). With more stringent criteria for gene se-lection (human: fold change >4.0; mouse model: fold change
HumanBurn
HumanTrauma
HumanSepsis
MouseBurn
MouseTrauma
MouseSepsis
P < 0.0001 P < 0.0001 P < 0.0001P < 0.0001P < 0.0001
Fig. 1. Correlations of gene changes among human burns, trauma, sepsis,and the corresponding mouse models. Scatterplots and Spearman’s rankcorrelations (ρ) of the fold changes of the genes responsive to both con-ditions for each pair of interest (P < 0.05; fold change >1.2). Vertical bar andhorizontal bar for each panel represents fold change in right and upperpanels, respectively. Murine models were highly significantly correlated withhuman conditions with Spearman’s correlation coefficient (ρ = 0.43–0.68; P <0.0001 for every comparison between human conditions and mouse models).The correlations between different mouse models were also significant (ρ =0.47–0.57; P < 0.0001 for every comparison).
Human Burn vs. Mouse Burn (GSE7404)
Human Burn vs. Mouse Sepsis(GSE26472)
Human Burn vs. Mouse Infection(GSE20524)
Human Burn vs. Mouse Sepsis(GSE19668)
Human Burn vs. Mouse Trauma(GSE7404)
940 genesHuman
Overlap P value:
3.9 × 10-34
Overlap P value:
6.3 × 10-13
Overlap P value:
6.5 × 10-11
Overlap P value:
1.2 × 10-35
Overlap P value:
3.4 × 10-35
↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
Human↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
267 genes 397 genes854 genes
578 genes 212 genes 513 genes221 genes
564 genesHuman↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
353 genes 237 genes1044 genes
771 genesHuman↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
Human↑ Human↑Mouse↑ Mouse↑
Human↓ Human↓↓Mouse ↓Mouse
773 genes 441 genes1266 genes
593 genes 482 genes 816 genes591 genes
BA
C
E
D
0
10
20
30
40
50
60
-log
(p-v
alue
)
0
5
10
15
20
-log
(p-v
alue
)
0
10
20
30
40
50
60
-log
(p-v
alue
)
0
5
10
15
20
-log
(p-v
alue
)
0
10
20
30
40
-log
(p-v
alue
)
Fig. 2. Statistical comparison of the direction of the gene expressionchanges between human burns and mouse models. Vertical bar representsthe significance of the overlap between gene sets. Genes whose expressionlevels were changed in human burns significantly overlapped with those inthe condition of mouse burn (A, overlap P value = 3.9 × 10−34), mousetrauma (B, 6.3 × 10−13), mouse sepsis from GSE19668 (C, 1.2 × 10−35), mousesepsis from GSE26472 (D, 6.5 × 10−11), and mouse infection (E, 3.4 × 10−35).Value is expressed as the –log 10 of the P value. Statistical significancesregarding the directionality of the gene expression changes were derivedfrom the nonparametric ranking method provided by the bioinformaticsplatform NextBio.
1168 | www.pnas.org/cgi/doi/10.1073/pnas.1401965111 Takao and Miyakawa
>1.2), R values and the percentages of genes changed in the samedirection were generally greater [Fig. 3; R = 0.36–0.59; P < 0.0001for all comparisons except for mouse sepsis (P = 0.0162); percent-age: 74.4–93.2]. Detailed information on the analyses is availablein Table S2.
Comparison of the Significantly Regulated Pathways/Biogroups BetweenHuman Conditions and Mouse Models. Pathways/biogroups commonlyaltered across datasets of human diseases and mouse models weredetermined through a combination of rank-based enrichment sta-tistics and biomedical ontologies using NextBio. In the human burncondition, significant enrichment of up-regulated and down-regu-lated gene expression was identified in 875 and 197 pathways/bio-groups, respectively, from a total of 6,176 sets of pathways/biogroups registered in NextBio [1,454 sets from Gene Ontology(GO) database, 4722 sets from canonical pathways in BroadMSigDB]. As a selection criterion, a P value that was corrected bythe Bonferroni correction for multiple comparisons (P < 8.1 × 10−6=0.05/6,176) was used for this identification. Among the 875 or 197pathways/biogroups with significant enrichment of up-regulated ordown-regulated genes in the human burn condition, the pathways/biogroups that demonstrated significant enrichment in other hu-man and mouse conditions as well were determined with a P valuecorrected with the Bonferroni correction (P < 5.7 × 10−5 = 0.05/875or P < 2.6 × 10−4 = 0.05/197, respectively). Of the 875 pathways/biogroups with significant enrichment of up-regulated genes inhuman burn conditions, 313, 163, 345, and 114 also showed sig-nificant enrichment in mouse models of burn, trauma, sepsis, andinfection, respectively. Unexpectedly, only half of the pathways/biogroups up-regulated in human burn conditions were shared bythe other human disease conditions (trauma 52.1%, sepsis 53.4%).Significant enrichment of up-regulated genes in six pathways/bio-groups was detected across all conditions (three human conditionsand four mouse models). Similarly, among the 195 pathways/bio-groups with significant enrichment of down-regulated genes inhuman burn conditions, 85, 44, 60, and 130 also showed significantenrichment in mouse models of burn, trauma, sepsis, and infection,
respectively. Significant enrichment of up-regulated genes in fivepathways/biogroups was detected across all seven conditions.Pathways/biogroups that show enrichment with a P value less than0.6 in at least one of the conditions are listed in Dataset S1.Some of the pathways/biogroups with high overlap are shown
in Fig. 4 A–D. The significance of the overlap between eachcondition and pathways/biogroups is also shown in the graph.There was significant overlap between genes annotated in GOas “innate immune response” and the genes up-regulated in themouse models of burn (Fig. 4A, P = 6.3 × 10−24), sepsis (P = 4.6 ×10−33), and infection (P = 1.3 × 10−16), as well as in human burn
Fig. 3. Comparison of the genomic response to severe acute inflammationfrom different etiologies in human and murine models. The datasets thatwere used in Seok et al. (1) and are registered in NextBio were reevaluatedusing the criteria of Seok et al. (1) to select the genes of interest. We chosegenes with significant responses in both the human burn dataset and in themouse dataset for comparison, whereas Seok et al. chose the 4,918 geneswith significant responses in the human datasets regardless of the signifi-cance of the changes in the genes in mice. The datasets used here are listedin Table S2. Shown are Pearson’s correlations (R; x axis) and directionality (%;y axis) of the gene response from multiple published datasets in GEO com-pared with human burns. Note that R2 data taken from Seok et al. wererecalculated to obtain the R values shown here for comparison (blue sym-bols). Most of the mouse models showed a high directionality score (randomchance is 50%), indicating that their gene expression patterns are similar tothe that in human burns, as long as stringent criteria are applied to theselection of the genes of interest.
activated suppressed
activated suppressed
activated suppressed
activated suppressed
Significance of overlaps with the Broad MSigDB biogroup (- log)
Significance of overlaps with the Broad MSigDB biogroup (- log)
Significance of overlaps with the GO biogroup (- log)
Significance of overlaps with the GO biogroup (- log)
Genes involved in Cytokine Signaling in Immune system
Innate immune response
Lymphocyte differentiation
Genes involved in Translocation of ZAP-70 to Immunological synapse
Mouse InfectionMouse Sepsis
Mouse TraumaMouse Burn
Human SepsisHuman Trauma
Human Burn
-10 0 10 20 30 40 50 60
Mouse InfectionMouse Sepsis
Mouse TraumaMouse Burn
Human SepsisHuman Trauma
Human Burn
-10 -5 0 5 10 15 20 25 30
B
A
C
D
Mouse InfectionMouse Sepsis
Mouse TraumaMouse Burn
Human SepsisHuman Trauma
Human Burn
-10 0 10 20 30 40 50 60 70 80 90
Mouse InfectionMouse Sepsis
Mouse TraumaMouse Burn
Human SepsisHuman Trauma
Human Burn
-5 0 5 10 15 20 25
Fig. 4. Representative pathways or biogroups shared by the human dis-eases and mouse models. (A–D) Shown are bar graphs of statistical signifi-cance (−log 10 P value) for the representative pathways with significantregulation in the human and murine models. (A) Genes annotated as “in-nate immune response (GO)” significantly overlapped with genes up-regu-lated in the mouse models as well as in human diseases. (B) Significantoverlap was also detected between “genes involved in cytokine signaling inimmune system (canonical pathways, Broad MSigDB)” and genes up-regu-lated in both the mouse models and human diseases. (C) Genes annotated“lymphocyte differentiation (GO)” and genes down-regulated in the mousemodels and human diseases significantly overlapped. (D) There was alsosignificant overlap between “genes involved in translocation of ZAP-70 toimmunological synapse (canonical pathways, Broad MSigDB).” As can beseen, in the pathways/biogroups shown here, the mouse models are com-parable to the human conditions in the significance of enrichment betweeneach condition and pathways/biogroups. For a complete list of pathways/biogroups shared by the human diseases and mouse models see Dataset S1.
Takao and Miyakawa PNAS | January 27, 2015 | vol. 112 | no. 4 | 1169
(P = 5.0 × 10−51), trauma (P = 4.1 × 10−90), and sepsis conditions(P = 6.3 × 10−21). Significant overlap was also detected between“genes involved in cytokine signaling in immune system (canonicalpathways, Broad MSigDB)” and genes up-regulated in the mousemodels of sepsis (Fig. 4B, P = 4.0 × 10−36), burn (P = 1.6 × 10−11),and infection (P = 1.3 × 10−8), as well as in human burn (P = 5.5 ×10−28), trauma (P = 6.5 × 10−52), and sepsis conditions (P = 3.2 ×10−12). With regard to down-regulated pathways/biogroups, genesannotated “lymphocyte differentiation (GO)” significantly over-lapped with genes down-regulated in the mouse models of burn(Fig. 4C, P = 2.2 × 10−8), trauma (P = 1.4 × 10−15), sepsis (P = 4.2 ×10−20), and infection (P = 1.3 × 10−18), as well as in human burn(P = 1.7 × 10−18), trauma (P = 5.6 × 10−12), and sepsis conditions(P = 3.3 × 10−30). There was also significant overlap between “genesinvolved in Translocation of ZAP-70 to immunological synapse(canonical pathways, Broad MSigDB)” and genes down-regulatedin all of the human disease conditions and mouse models of theseconditions (Fig. 4D, human burn, P = 1.3 × 10−11; human trauma,P = 0.0003; human sepsis, 8.7 × 10−25; mouse burn, P = 0.0003;mouse trauma, P = 5.1 × 10−10; mouse sepsis, P = 1.3 × 10−7; andmouse infection, P = 1.8 × 10−9).
DiscussionContrary to the findings reported by Seok et al.(1), the findingsof the present study using the same datasets presented in theirstudy demonstrated that the expression levels of the genes inmouse models of human disease conditions were highly signifi-cantly correlated with those of the human conditions. The per-centage of genes that changed in the same direction betweenhuman diseases and mouse models was greater than chance leveland statistically highly significant (P = 6.5 × 10−11 to 1.2 × 10−35).Moreover, meta-analysis of those datasets revealed a number ofpathways/biogroups commonly and significantly altered by mul-tiple conditions in both humans and mice. Thus, our findingswere quite different from those reported by Seok et al. (1) usingvirtually identical datasets, resulting in almost completely oppo-site conclusions.How can such discrepancies be explained? As a research
group that uses mice as a main tool for studying human dis-orders, we applied commonly used methodologies to the analy-ses to effectively detect similarities between humans and mice. Itis generally well known and accepted among researchers usinganimal models that mouse models mimic only partial aspects ofhuman disorders. To identify the signals/signs of the similaritieswith higher sensitivity, genes that do not show significant re-sponses to the experimental manipulations must be excluded fromthe analyses, because those nonresponsive genes would simplyproduce noise in the correlation analysis. For example, 13,586 and3,116 genes are changed (P < 0.05 and fold change >1.2) in hu-man burn conditions and mouse models of infection, respectively,and 1,992 among them are commonly changed in both humansand mice. We calculated the correlation coefficients and percen-tages of genes that changed in the same direction as the 1,992overlapping genes. However, Seok et al. (1) used 4,981 genes forthe calculation, which included all 3,250 genes that were changed(false discovery rate <0.001 and fold change >2.0) in human burnconditions, and the 1,668 genes with significant changes only inother human conditions, regardless of their responses in mousemodels. With the calculation method used in the previous study,the biased inclusion of genes that showed a significant responseonly in the human conditions and not in mouse models greatlyreduces the value of the overall correlation coefficient/percentageof the genes that changed in the same direction and misses thepossible responses in the mouse model that recapitulate the hu-man condition. The genes shared by humans and mice (i.e., the1,992 genes in the example described above) are those that mainlyrepresent the aspects of human disorders and biological pathwaysmimicked by the mouse model. It is not surprising that Seok et al.(1) found little correlation in each pathway, considering that genes
showing no significant response in the mouse models were in-cluded in the calculation.To compare pathways/biogroups that are commonly changed
between human diseases and mouse models, Seok et al. (1) usedPearson’s correlation coefficient of determination R2, whichshould not be used for detecting similarities between differentspecies. The use of R2 is inadequate in this case for the followingreasons. First, Pearson’s correlation assumes a linear relation-ship between normally distributed data, and there is no reasonto assume a linear relationship among fold changes or log-two-fold changes of gene expression levels. In this case, the geneexpression level of each gene may follow a different function ofthe severity of inflammation. Some of the genes may have line-arity but others may not, and such differences among genescould be more apparent between different species. Second, thenormality of the distribution cannot be assumed regardingfold changes and log-twofold changes of gene expression levels.Indeed, in the present study, the assumption of a normaldistribution for the datasets was formally rejected by theKolmogorov–Smirnov test. Thus, the use of Pearson’s correla-tion in this analysis is not justified. In addition, R2 provides theimpression that the correlation is small when R is an in-termediate value. Supposing that R is 0.14–0.28, which is usuallyconsidered an intermediate or moderate correlation, R2 wouldbe 0.02–0.08, which seems small to those not familiar with theuse of R2. In the present study, we used Spearman’s rank cor-relation, a nonparametric statistical method that does not as-sume linearity or normal distribution of the data.Another problem with the study by Seok et al. (1) is that they
did not try to find the optimal condition in a mouse study thathas multiple conditions in a dataset and that time-course datawere used in aggregate. In GSE19668, for example, there areeight conditions that can be compared (2, 4, 6, or 12 h afterinfection in C57BL6J or AJ mice). The overlap P values with thehuman burn dataset vary from 8.7 × 10−8 (C57BL6J, 2 h afterinfection) to 1.2 × 10−35 (C57BL6J, 6 h after infection). In ouranalyses, we used the dataset that showed the highest similarity forfurther analyses. In general, it is preferable to use conditions thatmost highly mimic the human disorders when using animal mod-els. In fact, when researchers develop an animal model of a humancondition, they screen the models to determine which modelmost highly mimics the particular aspect of the human condi-tion being studied.We also used the sophisticated statistical methods provided by
NextBio to find potential similarities between human conditionsand mouse models. Expression profiles vary considerably fromstudy to study and from species to species. Different studies fordifferent species can yield different dynamic ranges, distributionsof fold changes, and P values that reflect the conditions used.To allow for interstudy comparisons, NextBio employs a non-parametric approach, in which ranks are assigned to each genesignature based on the magnitude of fold change. Ranks are thenfurther normalized to eliminate any bias owing to varying plat-form size. This normalized ranking approach enables compar-isons across data from different studies, platforms, and analysismethods by removing the dependence on absolute values of foldchange and minimizing some of the effects of the differences ofnormalization methods, platform, and species used. Use of thismethod demonstrated highly significant similarities betweenhuman and mouse conditions (P value = 6.5 × 10−11 to 1.2 × 10−35).Evaluating similarities across microarray studies is an importanttopic for contemporary biomedical research. Simple Pearson’scorrelations are widely used, but other sophisticated statisticalmeasures, including that of NextBio, have been proposed toquantify the similarity of any two microarray studies (16). Itwould be interesting to assess the same datasets with otheradvanced analytical measures.The study by Seok et al. (1) has been interpreted by some
researchers and mass media as robust and persuasive evidencethat there are no correlations or similarities between mousemodels and human conditions, provoking discussions of the
1170 | www.pnas.org/cgi/doi/10.1073/pnas.1401965111 Takao and Miyakawa
validity of animal models as a tool for studying human diseaseand for translational research (3, 8). Commentaries in responseto the previous study claim that the implications of the study bySeok et al. (1) may well go beyond mice and sepsis (10), and thatbetter definitions of clinical phenotypes, especially in heteroge-neous or overlapping conditions such as neurodevelopmentaland psychiatric diseases, as well as more emphasis on rigorouslydefining molecular alterations in human patients, is needed (8).It should be noted that, even for schizophrenia, which couldbe considered a uniquely human disorder (17), recent mousemodels have been developed using an approach similar to thoseused in the present study that are shown to closely recapitulatenot only the clinical or behavioral phenotypes but also the mo-lecular alterations in transcriptional and protein changes in thebrain (18–21).The New York Times published an article attempting to
translate the findings in the study by Seok et al. (1) for a generalaudience, noting, “But now, researchers report evidence that themouse model has been totally misleading for at least three majorkillers—sepsis, burns, and trauma. As a result, years and billionsof dollars have been wasted following false leads, they say” (11).In the present study, however, we provide strong evidence, byreanalysis of the same datasets that were used by Seok et al. (1),that mouse models do in fact closely recapitulate certain aspectsof transcriptomic responses in some inflammatory conditions inhumans and are useful for studying human disorders and con-ducting translational research.In some commentaries (4, 6–8), the use of only a single strain
(i.e., C57BL/6) has been criticized as a potential caveat of thestudy by Seok et al. (1). This may not be the case, at least forsome datasets we analyzed, because the similarity of the dataderived from C57BL/6 with human data are comparable to oreven greater than those from other strains of mice, although ourstudy does not exclude the possibility that some other mousestrain mimics the human disorders better. Our results extendthose criticisms about the use of only a single strain and indicatethe importance of searching for optimal conditions in multipledimensions, including time course, dose, age, induction protocol,organ, and so on, by using appropriate and nonbiased data-mining methods to establish animal models of human disorders.Seok et al. (1) pointed out that all of the nearly 150 clinical
trials testing candidate agents intended to block the inflam-matory response in critically ill patients have failed (1). Whyhave these trials failed when such agents have proven effective inmouse models? Although it is beyond the scope of the study todiscuss this serious issue in detail, we would like to suggest somepossible reasons. First, it is possible that some of the failed drugswere targeting genes with distinctly different responses betweenhumans and mice, although the targeted pathways were indeedenriched for differentially expressed genes in humans and mice.Genes that show altered expression with a P value less than 0.05in at least three of the conditions are listed in Dataset S2. TNF-alpha, a major proinflammatory cytokine that activates theNF-κB pathway, is a therapeutic target for sepsis (22, 23). Can-didate agents, such as the TNF antibody, are effective in mousemodels (24) but failed to reduce mortality in clinical trials (25).In fact, the mRNA of TNF-alpha is significantly increased in thedatasets from mouse models (fold change, 4.8; P = 0.002 inmouse sepsis and fold change, 1.39, P = 0.0023 in mouse in-fection) but not in the datasets of human conditions assessed inthe present study (P > 0.05 in all of the four human conditions;TNF-alpha is not present in Dataset S2, because the table showsthe genes significantly changed in three or more biosets), con-sistent with previous studies that showed little increase in thegene (26, 27) or its product (28, 29) in human sepsis. Most ex-perimental therapies for sepsis have focused on attenuating theproinflammatory response, but the idea that immunosuppressionis a major contributor to the conditions has recently becomemore accepted (30). In this sense, the failure of the trials couldbe due to an inadequate strategy/target of the therapy. Second,compared with animal models, human sepsis patients comprise
heterogeneous populations in terms of the site or microbiologyof infection, genetic makeup, age, comorbidity of the patients,and so forth. Such heterogeneity in human conditions could di-lute the positive effect of the treatment, if any, in a subgroup ofpatients in a clinical setting. A review of studies using animalmodels of TNF neutralization in sepsis suggests that anti-TNFtherapies are most efficacious in models of systemic endotox-emia or Gram-negative infection, ineffective in models of com-plex polymicrobial infections, and harmful when the challengeorganism is Streptococcus pneumoniae or an opportunistic path-ogen (31). This implies that heterogeneity of the efficacy ofa treatment exists even among animal models with relativelyhomogeneous populations and that assessment of a treatmentrequires rigorous characterization of the heterogeneity of humanpatients. Third, the index for the outcome of clinical trials isfocused on patient mortality. If a trial is evaluated based solelyon mortality, it provides no information on the effect on theclinical or biological events that are not lethal but may be im-portant to patients who survive their sepsis episodes (23). It ispossible that one therapy cannot improve patient mortality butimproves the prognosis of survivors, such as the incidence rate ofamputations (23, 32) and level of organ dysfunction (32, 33).Taken together, it is not surprising that Seok et al. (1) found
almost no correlation between the genomic responses in humandisease conditions and those of mouse models, because they usedmethodologies with limited detection ability, which are rarelyused by researchers who study animal models. With reasonablysophisticated and commonly used methods, we found highlysignificant correlations between human and mouse data. Itshould be noted that in our analyses we used the data from theexact same studies as those used in Seok et al. (1), and still wefound highly significant similarities, demonstrating that theirfailure to detect correlations is purely a result of the inappropriatelybiased methodologies they used. Efforts to find more optimal age/time-course/organs/strains/induction protocols that recapitulate hu-man gene expression for each disease condition, might lead tobetter mouse models of these disorders than the ones evaluatedhere. Given that even these suboptimal models provide highly sig-nificant correlations, however, Seok et al. (1) might considermodifying their conclusions, which have drawn the attention ofthe popular media as well as researchers in broad areas of bio-medical sciences.
Materials and MethodsDatasets for Human Diseases and Mouse Models. The datasets that we ana-lyzed in the present study were the same as those used in the study by Seoket al. (1) and are registered in NextBio. In the present study, the followingdatasets were used for gene expression pattern analyses: “leukocytes ofpatients with severe burns on >20% of total body surface area vs. healthycontrols” from GSE37069 is referred as “Human Burn”; “white blood cells ofsevere blunt trauma patients 14 d after injury vs. healthy subjects” fromGSE36809 is referred as “Human Trauma”; “whole blood of sepsis patientswith community-acquired infection vs. healthy subjects” from GSE28750 isreferred as “Human Sepsis”; “WBC from blood at 7 d after burn injury vs.burn injury sham” from GSE7404 is referred to as “Mouse Burn”; “WBC fromspleen at 3 d after trauma hemorrhage vs. trauma hemorrhage sham” fromGSE7404 is referred to as “Mouse Trauma”; “Blood of C57BL6J mice 4 h afterStaphylococcus aureus infection vs. uninfected” from GSE19668 is referredto as “Mouse Sepsis”; “Blood from 8-wk-old BALB-c mice 1 d after tail veininjection - Candida albicans vs. saline control” from GSE20524 is referredto as “Mouse Infection.” Only genes with a P value of 0.05 or less and anabsolute fold change of 1.2 or greater were considered to be differen-tially expressed.
Comparison of Gene Response Between Datasets. Fold changes of datasetsused in the present study were calculated by dividing the value of humandiseases/mouse models by that of healthy controls/normal control mice.Values were converted into the negative reciprocal, or −1/(fold change) if thefold change was less than 1. In Fig. 1, genes meeting the criteria of P < 0.05and fold change >1.2 are plotted in the graph. The normality of eachdataset was assessed using the Kolmogorov–Smirnov test, in which none ofthe datasets had a normal distribution (P < 0.0001). Because the distribution
Takao and Miyakawa PNAS | January 27, 2015 | vol. 112 | no. 4 | 1171
of the datasets was not normal, Spearman’s rank correlation (ρ) was calcu-lated between each dataset. In nonparametric ranking analysis of the geneexpression signature by NextBio (Fig. 2), genes with a P value of 0.05 or lessand an absolute fold change of 1.2 or greater were used, which is the de-fault criterion used in analyses in NextBio. Pearson correlation (R) betweenthe human burn condition and other mouse models was calculated usinggenes with a P value of 0.05 or less and an absolute fold change greater than4.0 (Fig. 3, solid red circles) or 2.0 (Fig. 3, open red circles) in the human burndataset (as a reference) and genes with a P value of 0.05 or less and anabsolute fold change greater than 1.2 in other conditions of mouse models(Fig. 3). The percentage of genes changed in the same direction as in thehuman burn condition was also calculated using genes with the same cri-teria. The five datasets of human disease and five datasets of mouse modelsthat were used in Seok et al. (1) and are registered in NextBio as well are alsoplotted in Fig. 3.
Comparison of Pathways/Biogroups Altered in the Human Diseases and MouseModels.Meta-analyses of the human andmouse gene sets using NextBiowereconducted to find the pathways/biogroups commonly changed in the genesets. Pathways/biogroups commonly altered across datasets of human dis-eases and mouse models were determined through a combination of rank-based enrichment statistics (see the following section) and biomedicalontologies using NextBio. The datasets used in the meta-analysis were asfollows: Human Burn, “leukocytes of patients with severe burns on >20% oftotal body surface area vs. healthy controls”; Human Trauma, “white bloodcells of severe blunt trauma patients 28 d after injury vs. healthy subjects”;Human Sepsis, “whole blood of sepsis patients with community-acquiredinfection vs. healthy subjects”; Mouse Burn, “WBC from blood at 7 d afterburn injury vs. burn injury sham”; Mouse Trauma, “WBC from spleen at 3 dafter trauma hemorrhage vs. trauma hemorrhage sham”; Mouse Sepsis,“blood of C57BL6J mice 4 h after Staphylococcus aureus infection vs. un-infected”; Mouse Infection “blood from 8-wk-old BALB-c mice 1 d after tail
vein injection - C. albicans vs. saline control.” Pathways/biogroups from GOand canonical pathways of Broad MSigDB were included in this analysis.
For identification of the pathways/biogroups showing significant enrich-ment of the genes changed in Human Burn, a pathway enrichment analysiswas conducted among the 6,176 pathways/biogroups registered in GO andcanonical pathways of Broad MSigDB, with a P value corrected by theBonferroni correction for multiple comparisons (P = 8.1 × 10−6 = 0.05/6,176).The pathways/biogroups showing significant enrichment of the geneschanged in six other human and mouse conditions were also determined bypathway enrichment analyses using NextBio. Among the pathways/bio-groups with significant enrichment in human burn conditions, those alsoshowing significant enrichment in each condition were determined. Path-ways/biogroups with P values less than the alpha-level corrected by theBonferroni correction were considered significantly enriched.
Computing Overlap P Values of Gene Expression Patterns from DifferentDatasets. NextBio was used to compare gene expression patterns (signatures)between experiments from human diseases and mouse models. NextBio wasaccessed on December 31, 2013, for Figs. 1–4, unless otherwise noted. NextBiocompares the signatures in publicly available microarray databases with a sig-nature provided by the user using a “Running Fisher” algorithm, as previouslydescribed (15, 18). The overlap P value, that is, the direction of the correlationbetween two given gene signature sets, and the P values between subsets ofgene signatures, is calculated with the algorithm. For detailed method forcalculation of overlap P values, see SI Materials and Methods.
ACKNOWLEDGMENTS. This work was supported by Grants-in-Aid for Scien-tific Research 25116526, 10301780, and 80420397 from the Ministry of Edu-cation, Culture, Sports, Science, and Technology of Japan and by grants fromCore Research for Evolutional Science and Technology of the Japan Scienceand Technology Agency.
1. Seok J, et al.; Inflammation and Host Response to Injury, Large Scale CollaborativeResearch Program (2013) Genomic responses in mouse models poorly mimic humaninflammatory diseases. Proc Natl Acad Sci USA 110(9):3507–3512.
2. Cauwels A, Vandendriessche B, Brouckaert P (2013) Of mice, men, and inflammation.Proc Natl Acad Sci USA 110(34):E3150.
3. de Souza N (2013) Model organisms: Mouse models challenged. Nat Methods 10(4):288.
4. Drake AC (2013) Of mice and men: What rodent models don’t tell us. Cell Mol Im-munol 10(4):284–285.
5. Leist M, Hartung T (2013) Inflammatory findings on species extrapolations: Humansare definitely no 70-kg mice. ALTEX 30(2):227–230.
6. Osterburg AR, et al. (2013) Concerns over interspecies transcriptional comparisons inmice and humans after trauma. Proc Natl Acad Sci USA 110(36):E3370.
7. Remick D (2013) Use of animal models for the study of human disease–a shock societydebate. Shock 40(4):345–346.
8. (2013) Of men, not mice. Nat Med 19(4):379-379.9. Schaff M (2013) Mouse study findings prove contentious among scientists. Pitt News.
Available at www.pittnews.com/news/article_41cd755a-a081-577a-b752-2c83fc6628e2.html. Accessed January 17, 2014.
10. Collins DF (2013) Of mice, men, and medicine. NIH Director’s Blog. Available at http://directorsblog.nih.gov/2013/02/19/of-mice-men-and-medicine/. Accessed January 17,2014.
11. Kolata G (2013) Health testing on mice is found misleading in some cases. NY Times.Available at www.nytimes.com/2013/02/12/science/testing-of-some-deadly-diseases-on-mice-mislead-report-says.html. Accessed January 17, 2014.
12. Engber D (2013) Septic shock. Slate. Available at www.slate.com/articles/health_and_science/the_mouse_trap/2013/02/lab_mouse_models_of_inflammation_for_burns_sepsis_trauma_animal_testing.html. Accessed January 17, 2014.
13. Cossins D (2013) Do mice make bad models? The Scientist. Available at www.the-scientist.com/?articles.view/articleNo/34346/title/Do-Mice-Make-Bad-Models-/. Accessed January 17,2014.
14. Chu E (2013) This is why it’s a mistake to cure mice instead of humans. Popular Sci-ence. Available at www.popsci.com/science/article/2013-02/how-scientists-got-lost-curing-mice-instead-humans. Accessed January 17, 2014.
15. Kupershmidt I, et al. (2010) Ontology-based meta-analysis of global collections ofhigh-throughput public data. PLoS ONE 5(9):e13066.
16. Tseng GC, Ghosh D, Feingold E (2012) Comprehensive literature review and statisticalconsiderations for microarray meta-analysis. Nucleic Acids Res 40(9):3785–3799.
17. Powell CM, Miyakawa T (2006) Schizophrenia-relevant behavioral testing in rodentmodels: a uniquely human disorder? Biol Psychiatry 59(12):1198–1207.
18. Takao K, et al. (2013) Deficiency of schnurri-2, an MHC enhancer binding protein,induces mild chronic inflammation in the brain and confers molecular, neuronal, andbehavioral phenotypes related to schizophrenia. Neuropsychopharmacology 38(8):1409–1425.
19. Ohira K, et al. (2013) Synaptosomal-associated protein 25 mutation induces imma-turity of the dentate granule cells of adult mice. Mol Brain 6:12.
20. Walton NM, et al. (2012) Detection of an immature dentate gyrus feature in humanschizophrenia/bipolar patients. Transl Psychiatr 2:e135.
21. Hagihara H, Ohira K, Takao K, Miyakawa T (2014) Transcriptomic evidence for im-maturity of the prefrontal cortex in patients with schizophrenia. Mol Brain 7:41.
22. Wesche-Soldato DE, Swan RZ, Chung C-S, Ayala A (2007) The apoptotic pathway asa therapeutic target in sepsis. Curr Drug Targets 8(4):493–500.
23. Marshall JC (2014) Why have clinical trials in sepsis failed? Trends Mol Med 20(4):195–203.
24. Beutler B, Milsark IW, Cerami AC (1985) Passive immunization against cachectin/tumornecrosis factor protects mice from lethal effect of endotoxin. Science 229(4716):869–871.
25. Abraham E, et al.; NORASEPT II Study Group (1998) Double-blind randomised con-trolled trial of monoclonal antibody to human tumour necrosis factor in treatment ofseptic shock. Lancet 351(9107):929–933.
26. Wong HR, et al.; Genomics of Pediatric SIRS/Septic Shock Investigators (2009) Genomicexpression profiling across the pediatric systemic inflammatory response syndrome,sepsis, and septic shock spectrum. Crit Care Med 37(5):1558–1566.
27. Wong HR (2013) Genome-wide expression profiling in pediatric septic shock. PediatrRes 73(4 Pt 2):564–569.
28. Hotchkiss RS, Karl IE (2003) The pathophysiology and treatment of sepsis. N Engl JMed 348(2):138–150.
29. Oberholzer A, Oberholzer C, Moldawer LL (2000) Cytokine signaling—regulation ofthe immune response in normal and critically ill states. Crit Care Med 28(4, Suppl):N3–N12.
30. Hotchkiss RS, Coopersmith CM, McDunn JE, Ferguson TA (2009) The sepsis seesaw:Tilting toward immunosuppression. Nat Med 15(5):496–497.
31. Lorente JA, Marshall JC (2005) Neutralization of tumor necrosis factor in preclinicalmodels of sepsis. Shock 24(Suppl 1):107–119.
32. Levin M, et al. (2000) Recombinant bactericidal/permeability-increasing protein(rBPI21) as adjunctive treatment for children with severe meningococcal sepsis: Arandomised trial. rBPI21 Meningococcal Sepsis Study Group. Lancet 356(9234):961–967.
33. Marshall JC, et al. (1995) Multiple organ dysfunction score: A reliable descriptor ofa complex clinical outcome. Crit Care Med 23(10):1638–1652.
1172 | www.pnas.org/cgi/doi/10.1073/pnas.1401965111 Takao and Miyakawa
![Taqleed (Mimic) [English]](https://static.documents.pub/doc/80x56/577ce0831a28ab9e78b37d12/taqleed-mimic-english.jpg)
