The study presents an analysis of 307 mice generated from eight genetically different CC lines, with both sexes (156 male, 151 female). These mice were exposed to a dietary challenge (HFD vs. CHD) with or without oral co-infection challenge during a 12-week experimental period. The variations in the number of polyps within each sex after 12 weeks on either HFD or CHD under infection or non-infection conditions were determined, as well as the length and size of intestine.
Polyp number variation in different CC lines in response to dietary and infection challenge
Our observations show that there is a variation in polyp number between the different CC lines on HFD and CHD. To evaluate the effect of the HFD on the number of polyps developed, we compared the control group (CHD/no-infection) with the non-infected group on HFD (HFD/no-infection). Females showed higher polyp number on CHD in infected condition. The effect of HFD was recognized in IL711 and IL3912 by comparing infected mice on HFD versus non- infected mice on HFD. Surprisingly, these two lines responded very differently to diet. In IL711, infected mice maintained on standard diet developed significantly (P<0.05) more polyps compared to infected mice at HFD with values of 9.80±1.2 and 6.80±0.86, respectively. In contrast, infected females of IL3912 maintained on standard diet developed significantly (P<0.01) less polyps (5.38±0.48) compared to infected mice on HFD (10.80±2.85) (Figure 1A).
Males of IL6018 were resistant to diet and infection challenges, while females of IL6018 on HFD without infection presented a significant variation (P<0.01) in polyp development compared to the other groups. In male mouse population, overall, we found a significant (P<0.01) effect of infection on males on CHD, where infected males on CHD developed less polyps (6.45±0.70) when compared to non-infected mice on CHD (9.18±1.13).
Furthermore, we found that there is a significant opposite effect (P<0.05) of HFD without infection, males maintained on HFD without infection developed less polyps (7.29±1.22) compared to mice on standard diet without infection (9.18±1.13). Contrarily, infected males maintained on HFD developed more polyps (8.34±1.17) when compared to infected males maintained on CHD (6.45±0.70) (Figure 1B).
Males of IL72, IL711 and IL3912 also showed a significant variation with P<0.05, P<0.05, P<0.01, respectively, between infected males on HFD and infected males maintained on CHD (Figure 1B). In both small and large intestines, separately, female mice on HFD developed more polyps compared to control mice (Figure 2A). In overall male mouse population, there were a significant variation (P<0.01) between non-infected males on CHD and infected males on CHD, where infected males developed less polyps in small intestines, 6.37±0.98 and 4.53±0.49, respectively. In addition, non-infected males on CHD developed significantly (P<0.01) more polyps compared to non-infected males on HFD, 6.37±0.98 and 4.64±0.95, respectively. (Fig. 2B). IL3912 and IL4141 males maintained on CHD without infection were, significantly different (P<0.01) from all the other 3 experimental groups as shown in whole intestines, as well.
Infected males of IL3348 maintained on HFD developed, significantly (P<0.05) more polyps (8.33±2.40) compared to infected males on CHD (5±0.45) (Figure 2B). The effect of the infection on polyp development, along the whole intestines was highly significant in the female population (Figure 3A). The effect of infection was observed in males of IL3912 and IL4141, since infected males on HFD developed significantly (P<0.01) less polyps in colon compared to non-infected males at CHD (Figure 3B). On the other hand, the combination of HFD and infection affected males of IL72 and IL711 and, significantly development of more polyps (P<0.05) in colon compared to infected mice maintained on (Figure 3B).
The combination of HFD and infection significantly affected (P < .05) the whole intestines and the small intestine in the female population compared to other groups. Overall, infected females maintained on HFD showed, significantly the shortest intestines with P<0.01 when compared to female mice on CHD and female mice at HFD without infection (P<0.05.) (Figure 4A), while in the overall male mouse population, infected mice maintained on CHD showed, significantly longer intestines compared to the other three groups when compared to control male mice maintained on CHD (P<0.05) and when compared to male mouse maintained on HFD (P<0.01) (Figure 4B).
Non-infected females of IL6018 showed the longest small intestines (40.37±0.37 cm) significantly (P<0.01), compared to other three experimental groups of challenges (Figure 5A), while infected males of IL72 on CHD was significantly (P<0.01) shorter small intestines (25.5±0.47) compared to infected males maintained on HFD (30.78±0.39 cm) (Figure 5B). Overall infected male and female mice maintained on HFD presented shorter colons, 6.76± 0.49 cm and 6.53± 0.34 cm compared to other three groups of challenges (Figure 6A and 6B).
Combinatorial effect of the diet and infection on length and size of the intestine
Five tested different lines from the eight, showed significant variation between infected females maintained on HFD and infected females on CHD. Infected females maintained on HFD in all these lines showed shorter colon when compared to infected females on CHD.
The combinatorial effect of HFD and infection observed varied in most of the male lines. Similar to female response, infected males of IL72, IL557, IL3912, IL4141 and IL6018 maintained on HFD showed, significant short colon compared to infected males maintained on CHD.
In females’ population (Figure7A): Overall infected females on CHD had, significant (P<0.01) the biggest size of intestines (36.15±1.35 cm2) compared to non-infected females maintained on CHD (31.54±1.88 cm2), Infected females at HFD (28.25±2.17 cm2) and non-infected females at HFD (31.75±2.53 cm2). Infected females of IL1912 and IL711 maintained on CHD also showed the biggest size of intestines compared to the other three experimental groups.
Results have showed that Infected females of IL557, IL711 and IL1912 on CHD developed significantly (P<0.01) bigger size of intestines, 39.21±0.1.08 cm², 40.06±0.3.5 cm², and 38.11±0.2.23 cm², respectively, when compared to infected females on HFD as showed records of 26.33±2.20 cm², 26.47±0.1.75 cm², and 26.36±4.59 cm², respectively.
The effect of oral infection on males was observed when mice maintained on CHD. Results have shown that infected males of IL557, IL1912 and IL3912 on CHD developed bigger size of intestines, with records of 43.36±3.26 cm², 37.45±1.49 cm² and 40.31±1.24 cm², respectively, when compared to non-infected males on CHD, which showed lower records, and were 32.37±0.1.50 cm², 26.95±1.58 cm² and 35.19±0.2.63 cm², respectively. However, infected males of IL72 on CHD responded, differently to infection and developed, significantly (P<0.05) smaller size of intestines (26.07±1.69 cm²) compared to non-infected males of IL72 on CHD (31.07±0.63 cm²) (Figure 7B)
Overall, infected females maintained on CHD significantly showed the biggest size of small intestines (29.67±1.24 cm²) when compared to males maintained on CHD without infection, 25.11±1.68 cm² with P<0.01, infected males on HFD, 23.81±1.94 cm² with P< 0.05 and non-infected males on HFD, 27.19±2.26 cm² with P<0.01.
Non-infected females of IL6018 and IL557 maintained on CHD, had a significant (P<0.05), bigger size of small intestines with records of 34.01±1.91 cm² and 33.14±1.29 cm², respectively, when compared to non-infected females maintained on HFD as showed records 25.74±1.56 cm² and 27.78±1.63 cm², respectively (Figure 8A)
Similarly, to overall females’ mice response, infected males maintained on CHD showed significantly (P<0.01) the biggest size of small intestines (27.76±1.95 cm²) compared to non-infected males maintained on CHD as showed records 25.70±1.79 cm², to infected males maintained on HFD with records 24.22±1.51 cm², and to non-infected males when maintained on HFD with showed records 24.23±1.26 cm² (Figure 8B).
Overall males and females showed significant (P<0.01) bigger size of colon in CHD+No.inf and CHD+Inf compared to HFD+No.Inf and HFD+Inf experimental groups.
Non-infected females of IL557, IL3348, L4141 and IL6018 maintained on CHD showed significantly (P<0.01) bigger size of colon compared to non-infected females maintained on HFD (Figure 9A). Non-infected males of IL1912, IL3912 and IL6018 when maintained on CHD showed bigger size of colon with records of 6.90±0.31 cm², 8.18±1.15 cm² and 8.53±0.31 cm², respectively, compared to non-infected males maintained on HFD as showed records 2.58±0.54 cm², 5.05±0.37 cm² and 5.32±0.31 cm², respectively (Fig.9B). Another significant variation (P<0.01) was observed between non-infected males maintained on CHD and infected males on HFD in IL1912, IL3912 and IL6018 (Figure 9B).
Variation in the effect of HFD on glucose tolerance
The results presented in (Figure 10) revealed, significantly higher values of AUC compared to the control group (P < .01) in both male and female populations. The combination between HFD and infection induced high levels of AUC values, while male mice showed higher values of AUC (51805.95 ± 3483), when compared with males with males maintained on CHD with oral infection (37772.97± 2076.38) (Figure 10B).
The AUC profiles observed in females of IL72, IL4141 and IL3912, IL557 at HFD, which showed a significant difference comparing to females maintained on CHD of the same lines with levels of *P < 0.05 (Figure 10A). Interestingly, males of the same lines at HFD also show a significant variation in AUC (mg/dL*min) values compared to males at CHD (P <0.01).
Multimorbidity heatmaps of polyp counts, AUC, BW, length and size
A major aim of our proposed research was to study the effect of the host genetic background on diseases multimorbidity, to better understand the coexistence of multiple disease conditions in an individual.
Heatmaps were developed of these traits and searched for an association in development and severity between these traits. As presented in Figure 11, overall, the non-infected females and males when maintained on CHD (Figure 11A) showed negative correlation while infected mice maintained on CHD showed positive correlation between total AUC and length of intestines (Figure 11B), which indicates, to be more resistant to the effects of the dietary and infection challenges on the development of these traits.
This figure further analyzes the data showing which traits were found to be non-significant or significant at levels P < .05 and P < .01 in female and male mice of the studied lines in the different experimental groups. The results presented here, show that body weight gain was highly significantly (P < .01) increased in female IL557 mice compared with IL711 females in all experimental groups, while in male mice body weight gain was highly significant (P < .01) in all groups except the CHD + No-Inf. group. The same significance level (P < .01) was also observed in AUC values between females in the HFD + No-Inf. and CHD + No-Inf. group, and males in HFD + No-Inf. and in CHD + Inf groups. Intestine size was highly significantly different (P < .01) between IL557 and IL711 females in the CHD + No-Inf. group only, and between the male mice in the CHD + Inf group. In the case of the length of the intestines, a significant variation (P < .05) was observed between IL557 and IL711 females in the CHD + No-Inf. group only, and between male mice in two experimental groups, HFD + No-Inf. and CHD + Inf. Finally, the number of polyps was found to be a significant variant (P < .05) between the two lines only in HFD + No-Inf. Group, with no significant difference between sexes.
Heritability and genetic coefficient of variation
Table 2 presents heritability and genetic coefficient of variation values for a variety of traits studied. The heritability values of these traits are generally in the range of 0.33–0.92, while the CVG has been observed to be in the range 0.00–0.62, much higher than the benchmark of 0.071. Thus, the data shows that an absolute magnitude of genetic variation among the CC lines observed is higher than found within a typical outcrossing population.
Classification and Regression Results
The results presented in Table 3, the size or area of the intestine could be classified with high ROC AUC values for four lines. The model that performed best across the board was Random Forest with four values greater than 0.8 (lines 557, 711, 1912 and 3348) and a minimal value of 0.663 (line 6018). The single highest score was for Logistic Regression in line 557 – AUC = 0.936. In Supplementary Table 1, the classification of the mouse intestine length does not produce many high scores. The only line with multiple high values was 3912, for which the Logistic Regression performed best with a score of 0.859, while Naïve Bayes and Random Forest produced scores of 0.832 and 0.835 respectively. In lines 72 and 557 significantly lower scores – No model was able to predict with a score greater than 0.6.
The results presented in Supplementary Table 2 are the AUC classification scores also varying between different lines. While for most lines, we see no model with a ROC AUC greater than 0.8, in lines 72 and 1912 there are 2-3 models that produce such values. The best model for these lines was Logistic Regression with scores of 0.861 and 0.867. The observed results presented in Supplementary Table 3 are of the final body weight classified with exceptionally high AUC values for almost all lines. The model that performed best was Logistic Regression, with a maximum of 0.948 for line 4141 and a minimum of 0.682 for line 6018. The lines 72 and 6018 stand out as no model has produced a ROC AUC of over 0.8, while for most lines 3-5 models were able to produce such scores. These results are not surprising, as there is a strong positive correlation between initial body weight and final body weight in most lines. In Supplementary Table 4 a significant difference between two groups of lines: Lines IL72, IL711, IL1912 and IL3348 all had at least two models with high AUC values (0.823 and above), whereas lines IL557, IL3912, IL4141 and IL6018 had no model with an AUC of 0.75. The best model for this prediction was Random Forest with a peak of 0.985 for line 3348. In Supplementary Table 5 as we can see in the results, the number of polyps is very hard to classify by the chosen models. The results presented in Supplementary Table 6 indicate that diet can be classified with high AUC values for three lines using most models. For lines IL72 and IL4141 we were able to observe five models with AUC larger than 0.84. However, for some lines all models were unable to produce a high AUC value like the best result for line IL711 was 0.619.
The classification results are significantly different among different lines, models, and input-output combinations. While some features, such as polyp number were completely unpredictable, other features like final body weight are highly predictable with AUC values of ~0.9. Out of the 6 models we used, Logistic Regression, Naïve Bayes and Random Forest performed better than Decision Trees, Nearest Neighbors and SVC. Most lines obtained high AUC values for some combinations, yet line 6018 stands out as the significantly most unpredictable line, with only one AUC value larger than 0.8 – 0.801 for classifying the length using Random Forest.
Regression
In Supplementary Table 7, to predict the size of intestine, for most lines the behavior was completely unpredictable, with values of zero or close to zero. However, the linear models produced correlative predictions for line IL72 (median score of ~0.46) and for line IL1912 (median score of 0.229-0.299). The only line with correlative prediction was line IL6018. The Neighbors model performed best with an average score of 0.424 and a median score of 0.48. The linear models, Linear Regression and Lasso also produced correlative predictions with median scores of 0.299 and 0.336 respectively in predicting the length of intestine as presented in Supplementary Table 8.
The results presented in Supplementary Table 9 are the regression results for AUC varying drastically among the lines. For lines IL72, IL1912 and IL4141 both median and average results are very high, with a peak of 0.775 for Linear Regression in line IL72. On the contrary, the scores for the other 5 lines are virtually 0. In Supplementary Table 10, much like the high classification results for the final body weight, the numerical prediction for it is also very high. The peak was in Linear Regression for line IL4141 with a very high 0.877 median score. For 6 lines we were able to obtain correlative results in all models, and for the remaining 2 lines 1-2 models produced weak correlative results – these results indicate some predictability. As the classification for the number of polyps produced the lowest predictability. The input combination of diet and infection status only produced higher results than any other subset of the original input. However, the results are still non-correlative as presented in Supplementary Table 11.
The Regression results vary drastically among models. Extra Trees Regression performed significantly worse than Nearest Neighbors and the 2 linear models – Linear Regression and Lasso. For most input-output combinations, we were able to observe the correlative results for at least 2 models indicating that these features are somewhat predictable. The number of polyps was completely unpredictable, with only one score larger than 0.2 in line IL1912. Most of lines obtained correlative results for some combinations, yet lines IL3348 and IL6018 stand out as the significantly most unpredictable lines, with only one predictable feature – final body weight and intestine length respectively.