Registered report: Intestinal inflammation targets cancer-inducing activity of the microbiota

The Reproducibility Project: Cancer Biology seeks to address growing concerns about reproducibility in scientific research by conducting replications of 50 papers in the field of cancer biology published between 2010 and 2012. This Registered report describes the proposed replication plan of key experiments from ‘Intestinal Inflammation Targets Cancer-Inducing Activity of the Microbiota’ by Arthur et al. (2012), published in Science in 2012. Arthur and colleagues identified a genotoxic island in Escherichia coli NC101 that appeared to be responsible for causing neoplastic lesions in inflammation-induced IL10−/− mice treated with azoxymethane. The experiments that will be replicated are those reported in Figure 4 (Arthur et al., 2012). Arthur and colleagues inoculated IL10−/− mice with a mutated strain of E. coli NC101 lacking the genotoxic island, and showed that those mice suffered from fewer neoplastic lesions than mice inoculated with the wild type form of E. coli NC101 (Figure 4). The Reproducibility Project: Cancer Biology is a collaboration between the Center for Open Science and Science Exchange, and the results of the replications will be published by eLife. DOI: http://dx.doi.org/10.7554/eLife.04186.001


Introduction
In their 2012 Science paper, Arthur and colleagues examined the interplay between colitis and colon cancer. They identified a shift in the composition of the microbiota in Il10 −/− mice, which develop chronic colitis; amongst other changes noted, Escherichia coli was over 100-fold more represented in Il10 −/− mice than wild type. After treatment with azoxymethane (AOM), a carcinogen that induces colon cancer, germ-free Il10 −/− mice mono-associated with the colitis-inducing E. coli NC101 strain developed invasive mucinous carcinomas, while mice mono-associated with Enterococcus faecalis, another colitis-inducing bacterial strain, did not. E. coli NC101 harbors a 54 kb polyketide synthases (pks) genotoxic island encoding several enzymes involved in the production of toxin called Colibactin. This island has been previously shown to induce DNA damage, double strand breaks and aneuploidy (Nougayrede, 2006;Cuevas-Ramos et al., 2010) and was not found in E. faecalis or another noncolitic E. coli strain, K12.
In Figure 4, Arthur et al. inoculated germ-free IL10 −/− mice treated with or without AOM with either wild-type E. coli NC101, or with a mutant of E. coli NC101 lacking the pks island (NC1010Δpks). Arthur and colleagues first confirmed that loss of the pks island did not impair bacterial growth (Supplemental figure 7, replicated in Protocol 1). Presence or absence of the genotoxic pks island had no effect on colonic inflammation in IL10 −/− mice alone or treated with AOM ( Figure 4A). However, at 12 and 18 weeks, mice treated with AOM and inoculated with E. coli NC101Δpks had many fewer neoplastic lesions than mice inoculated with wild-type E. coli NC101 ( Figure 4B). At 18 weeks, invasion ( Figure 4C), tumor burden ( Figure 4D) and tumor size ( Figure 4E) were all reduced in mice monoassociated with E. coli NC101Δpks as compared to NC101. Taken together, the data indicate that loss of the pks genotoxic island from E. coli NC101 strongly reduces the incidence of colon cancer in mice with chronic colitis. These experiments are replicated in Protocol 3. Buc et al. (2013) performed an experiment similar to Figure 3B (not included for replication in this study), wherein Arthur et al. (2012) examined if the pks genotoxic island was more prevalent in patients with colorectal cancer. Both the dataset from Arthur et al. (2012) and the dataset presented by Buc et al. (2013) support the hypothesis that the genotoxic pks island is more prevalent in patients with colorectal cancer. Cougnoux et al. (2014), while not performing a direct replication, extended the findings of Arthur and colleagues to explore the mechanism of pks-produced colobactin toxicity effects on colorectal cancer. Finally, Arthur and colleagues have since published further work exploring in greater detail the genetic mechanisms behind the association of colorectal cancer with colitis and colonization by E. coli NC101 with or without the pks island, in which they demonstrate that colon inflammation itself has an influence on the expression of the genotoxic pks island (Arthur et al., 2014).

Materials and methods
Unless otherwise noted, all protocol information was derived from the original paper, references from the original paper, or information obtained directly from the authors. An asterisk (*) indicates data or information provided by the Reproducibility Project: Cancer Biology core team. A hashtag (#) indicates information provided by the replicating lab.
Protocol 1: comparing the growth curves of E. coli NC101 and E. coli NC101Δpks This protocol describes how to grow both E. coli NC101 and E. coli NC1010Δpks to compare their growth curve, as seen in Supplemental figure 7.
Sampling c This experiment will be repeated a total of three times.
a. Power calculations were not performed, as no significant difference was reported in the original study.

Materials and reagents
c Reagents that differ from those used originally are indicated with an asterisk (* Known differences from the original study c Lab will use in-house bacterial genomic DNA extraction protocol. a. Original was unspecified.

Provisions for quality control
All data obtained from the experiment-raw data, data analysis, control data and quality control data-will be made publicly available, either in the published manuscript or as an open access dataset available on the Open Science Framework (https://osf.io/y4tvd/). c We will sequence the 16S rRNA of the bacteria to confirm the identity of the strain. The sample purity (A 260 /A 280 ratio) of the extracted DNA will be recorded.
Protocol 3: mono-associate mice with E. coli NC101 or NC101Δpks and analyze intestinal tumorigenesis and inflammation This protocol describes the inoculation of germ-free Il10 −/− mice with E. coli as well as treatment with the carcinogen azoxymethane (AOM), as seen in Figure 4.

Materials and reagents
c Reagents that differ from those used originally are indicated with an *.

Procedure
Notes c Experiment should be conducted by experimenters blinded to genotype and treatment group. c Azoxymethane (AOM) can show lot-to-lot variation in potency, and will lose potency and gain toxicity over time.
In order to minimize these effects, a single lot of AOM will be used throughout the experiment. AOM will be aliquoted into 25 mg/ml aliquots and stored at −80˚C. A fresh aliquot will be used each time AOM is needed to avoid repeated freeze-thaw cycles. 3. Randomly assign Il10 −/− mice to two groups. As each mouse becomes eligible for induction of colitis/ colorectal cancer, randomly assign to a treatment group using the adaptive randomization approach with the gender of the mice as the covariate that is assessed as mice are sequentially assigned to a particular treatment group. Assignment will aim for a similar distribution of genders in each cohort while also taking into account the pre-determined size of each treatment group. a. Group 1: E. coli NC101; n = 14. b. Group 2: E. coli NC101Δpks; n = 16. 4. Colonize mice by oral gavage and rectal swabbing (dip sterile Q-tip in culture and swab anus) with log phase growth bactera: a. Use 200 μl of an overnight bacterial culture with a concentration of 2 × 10 9 CFU/ml. ■ Dilute enough to resolve single colony forming units (CFUs). iv. Calculate the total number of CFUs per 200 mg fecal matter. 5. At the same time as first stool culture, intraperitoneally inject with 10 mg/kg azoxymethane (AOM). 6. Repeat AOM injections every week for 5 more weeks (6 weeks total). 7. 18 weeks after last AOM injection, sacrifice mice.
a. #Mice are anesthetized with isofluorane in a drop jar. Once respiration has ceased, the mice are exsanguinated. b. *Collect stool and quantify colonization as performed in Step 5b. 8. Macroscopically examine tumor formation: a. Remove colons from the cecum to the rectum, flush with PBS, and splay longitudinally. b. Blindly count tumors per mouse and measure tumor diameter macroscopically. Image colon and tumors. 9. Prepare tissue for histological analysis: a. Collect distal colon tissue samples. b. Swiss-roll colon tissue samples from the distal to the proximal end and fix overnight in 10% formalin. 10. Paraffin-embed tissues.
a. #The replicating lab uses an automated embedding station: i. Samples are passed through a dehydration series consisting of 70%, 80%, 2 × 95% and 3 × 100% ethanol for 30 min each wash.
ii. Samples are washed into xylene; first wash is 30 min, the second is an hour.
iii. The samples are washed into 57˚C Paraplast; four washes of 30 min each. iv. Samples are mounted in mold and allowed to cool and harden. 11. Cut 6 μm sections and mount on slides. 12. Stain with hematoxylin and eosin for histologic analysis.
ii. 1 = mild dysplasia characterized by aberrant crypt foci (ACF), +0.5 for small gastrointestinal neoplasia (GIN), or multiple ACF. iii. 2 = moderate dysplasia with GIN, +0.5 for multiple occurrences or small adenoma. iv. 3 = severe or high grade dysplasia restricted to the mucosa. v. 3.5 = adenocarcinoma, invasion through the muscularis mucosa. vi. 4 = adenocarcinoma, full invasion through the submucosa and into or through the muscularis propria. c. Score invasion as follows: i. 0 = no invasion.
Deliverables c Data to be collected: a. Mouse records (gender used in each group, type of colonization procedure, health records, condition for early euthanasia, etc). b. OD 600 of overnight bacterial cultures used in colonization. c. (Compare to Supplemental figure 10): Raw data and Kaplan-Meier survival curve of mice for all conditions. d. Sequencing chromatograms and gel image of amplicons from stool sample confirming colonization with E. coli. e. (Compare to Figure 4D): Images, raw numbers and dot plot graph of macroscopic tumor number per mouse (multiplicity) at 18 weeks of colon tissue from mice for all conditions. f. (Compare to Figure 4F): Micrographs of H&E histology for each mouse at 18 weeks for all conditions. g. (Compare to Figure 4E): Raw numbers and dot plot graph of mean macroscopic tumor diameter in each mouse at 18 weeks of colon tissue from mice for all conditions. h. (Compare to Figure 4A): Raw numbers and dot plot graph of inflammation scores at 18 weeks of colon tissue from mice for all conditions. i. (Compare to Figure 4B): Raw numbers and dot plot graph of neoplasia scores at 18 weeks of colon tissue from mice for all conditions. j. (Compare to Figure 4C): Raw numbers and dot plot graph of invasion scores at 18 weeks of colon tissue from mice for all conditions. k. Counts of bacterial colonization from stool sample per mouse at 4 week time point and at sacrifice.

Confirmatory analysis plan
c Statistical analysis of the replication data: a. (As seen in Supplemental figure 10): Compare survival of AOM-treated Il10 −/− mice inoculated with E. coli NC101 relative to NC101Δpks. ■ Log-rank test (Mantel Cox). b. (As seen in Figure 4A, right panel): Compare mean inflammation scores of AOM-treated Il10 −/− mice monoassociated with E. coli NC101 relative to NC101Δpks. ■ Unpaired two-tailed Student's t-test. c. (As seen in Figure 4B): Compare mean neoplasia score of AOM-treated Il10 −/− mice mono-associated with E. coli NC101 relative to NC101Δpks. ■ Unpaired two-tailed Student's t-test. d. (As seen in Figure 4C): Compare mean invasion score of AOM-treated Il10 −/− mice mono-associated with E. coli NC101 relative to NC101Δpks. ■ Unpaired two-tailed Student's t-test. e. (As seen in Figure 4D): Compare mean macroscopic tumor number (multiplicity) of AOM-treated Il10 −/− mice mono-associated with E. coli NC101 relative to NC101Δpks. ■ Unpaired two-tailed Student's t-test. f. (As seen in Figure 4E): Mean macroscopic tumor diameter per mouse of AOM-treated Il10 −/− mice monoassociated with E. coli NC101 relative to NC101Δpks. ■ Unpaired two-tailed Student's t-test. g. Compare mean number of CFUs per 200 mg stool pellet between E. coli NC1010-colonized and E. coli NC1010Δpks-colonized mice at 4 weeks post-inoculation and at sacrifice. ■ Two-way ANOVA. c Meta-analysis of original and replication attempt effect sizes: a. This replication attempt will perform the statistical analysis listed above, compute the effects sizes, compare them against the reported effect size in the original paper and use a meta-analytic approach to combine the original and replication effects, which will be presented as a forest plot.
Known differences from the original study c The replication attempt will be restricted to the 18 week time point. c The microscope used in the original lab was an Olympus CX41; the replicating lab will use an Olympus BX41. c We will be using a different lot of azoxymethane that used by the original authors. AOM is known to have lot-tolot variation in efficacy that may affect the absolute numbers of tumors generated per mouse. c The original study collected data on 4 female and 8 male IL10 -/mice inoculated with E. coli NC101 and 8 male IL10 -/mice inoculated with E. coli NC101Δpks. While the gender of the mice in the replication is not currently known, the mice will be randomly assigned when they reach the age for inoculation with the aim of a similar gender distribution in each group. This will likely generate a different gender distribution than the original study.

Provisions for quality control
All data obtained from the experiment-raw data, data analysis, control data and quality control data-will be made publicly available, either in the published manuscript or as an open access dataset available on the Open Science Framework (https://osf.io/y4tvd/).
c The experiment will be performed by a Science Exchange lab with expertise in germ free mouse studies. c Experimenters will be blinded to the genotype and treatment group. Mice will be randomly assigned to treatment groups.

Power calculations
Protocol 1 c Not applicable.
Protocol 2 c Not applicable.

Protocol 3
Summary of original data Figure 4A: Test family c Two-tailed t-test, difference between two independent means, alpha error = 0.05. a. Sensitivity calculations were performed using G*power software (Faul et al., 2007).

Power calculations
c Because the original data shows a non-significant effect, we will not be powering this replication to detect an effect. Based on the sample size of 10 mice per group derived from Figure 4C, with α of 0.05 we will be powered to 80% to detect a Cohen's d of 1.3249474.

Summary of original data
c Note: Raw data values obtained from scatterplot with confirmation of accuracy from the authors.
c Stdev was calculated using formula SD = SEM*(SQRT n).

Test family
c Two-tailed t-test, difference between two independent means, alpha error = 0.05. a. Power calculations were performed for statistically significant effects reported in original study using G*power software (Faul et al., 2007).

Power calculations
Summary of original data c Note: Raw data values obtained from scatterplot with confirmation of accuracy from the authors.  Figure 4C, we will use 10 mice per group. This brings the a priori power to 91.3%.
c Stdev was calculated using formula SD = SEM*(SQRT n).

Test family
c Two-tailed t-test, difference between two independent means, alpha error = 0.05. a. Power calculations were performed for statistically significant effects reported in original study using G*power software (Faul et al., 2007).

Power calculations
Summary of original data c Note: Raw data values obtained from scatterplot with confirmation of accuracy from the authors.
c Stdev was calculated using formula SD = SEM*(SQRT n).

Test family
c Two-tailed t-test, difference between two independent means, alpha error = 0.05. a. Power calculations were performed for statistically significant effects reported in original study using G*power software. Figure 4E: c Two-tailed t-test, difference between two independent means, alpha error = 0.05. a. Sensitivity calculations were performed using G*power software (Faul et al., 2007).

Power calculations
c Because the original data shows a non-significant effect, we will not be powering this replication to detect an effect. Based on the sample size of 10 mice per group derived from Figure 4C, with α of 0.05 we will be powered to 80% to detect a Cohen's d of 1.3249474.

Supplemental figure 10:
Test family c Two-tailed t-test, difference between two independent means, alpha error = 0.05. a. Sensitivity calculations were performed using G*power software (Faul et al., 2007).

Power calculations
c Because the original data shows a non-significant effect, we will not be powering this replication to detect an effect. Based on the sample size of 10 mice per group derived from Figure 4C, with α of 0.05 we will be powered to 80% to detect a Cohen's d of 1.1453705.