An improved method for experimental induction of ulcerative colitis in Sprague Dawley rats

Abstract

Ulcerative colitis (UC) is a chronic inflammatory manifestation of the human colon that is linked with colorectal cancer. Development of an appropriate animal model is crucial to study the immunopathophysiology of UC wherein chemical induction is the most popular method of choice. However, unavailability of an optimum experimental model limits the success of this method. The present study aims to establish an optimized model for acetic acid-induced colitis in Sprague Dawley rats. Response Surface Methodology (RSM) with a six-factors Box-Behnken design was employed to generate an improved method of inducing UC in rat, predicting the case statistics, apposite investigation of quadratic response surfaces, and construction of a second-order polynomial equation. UC was diagnosed through three responses viz. weight loss, severity of diarrhea, and appearance of blood in the stool. Analysis of variance alongside RSM jointly revealed that induction of UC can be achieved with highest probability using the combination of parameters that includes 120 gm body weight, 1.5 ml of 4% acetic-acid v/v in distilled water with a single dose of treatment for 24 h including a pre-induction of 5 mins. This optimized UC-induction model was validated in-vivo through disease scoring index and hematological assessments with satisfactory level of desirability.

• •An improved experimental method for inducing ulcerative colitis (UC) in Sprague Dawley rats has been developed.
• •Box-Behnken Design-fitted Response Surface Methodology (RSM) was implicated in optimizing the experimental parameters for generating UC.
• •This statistically optimized and experimentally validated method resembles the recipe for the generation of UC in animal model with the highest possible desirability.

Graphical abstract

Specifications Table

Subject area:	Immunology and Microbiology
More specific subject area:	Inflammatory diseases, Experimental model
Name of your method:	Induction of ulcerative colitis in Sprague Dawley rats
Name and reference of original method:	Not Applicable
Resource availability:	Design-Expert Software v8.0 (Stat-Ease Inc., Minneapolis, USA), Glacial Acetic acid (MERCK), Sprague Dawley rats

Method details

Method overview

Inflammatory bowel disease (IBD) is a chronic, relapsing, and remitting inflammatory condition of the gut and colon distinguished into two major types, viz., ulcerative colitis (UC) and Crohn's disease. UC is characterized by continuous inflammation over the mucosal and submucosal layers initiating at the rectum and extending proximally towards the colon [1]. However, inflammation in Crohn's disease is transmural and can occur in a patchy pattern in any part of the gastrointestinal tract, especially in the terminal ileum, ascending colon, and rectum [1]. The disease manifestation of UC begins with abdominal pain, chronic diarrhea, and weight loss which further transforms into tissue fibrosis, stenosis, fistulas, and colorectal cancer [1,2]. The etiology of UC remains enigmatic. Nevertheless, several studies on both human and murine models suggested being a multifactorial process including, genetic factors, environmental factors, food and lifestyle habit, alteration of the gut mucosal barrier, microbial dysbiosis, and dysregulation of immune homeostasis play a key role in the pathogenesis of UC [3]. The alteration in the immune response is a critical contributor to the initiation and perpetuation of UC. It is characterized by the substantial infiltration and thickening of neutrophils, monocytes, macrophages, and T-cells across the mucosal layer followed by immunogenic cascade for activation of Th-2 driven immune responses directed by the release of IL-4, IL-5, IL-10, and IL-13 [1,3,4]. Thus, changes in the tissue architecture and alteration in immune homeostasis portray the hallmark for distinguishing UC.

Chemically-induced UC in animal models mimics the key features of the immunological and histopathological characteristics of human UC [3,5]. It serves as a significant model in the research of pathophysiological attributes of UC and open-up new areas to explore therapeutic intervention strategies [3,5]. There are several different techniques for inducing UC chemically in the animal model, viz. acetic acid-induced colitis, Dextran sodium sulfate (DSS)-induced colitis, Oxazolone-induced colitis, and 2,4,6 Trinitrobenzene sulfonic acid (TNBS)/ethanol induce colitis [3,5]. All these methods are well-established and highly validated for generating the UC model. Herein, acetic-acid-induced colitis is the most common and easily inducible murine model resembling with pathogenesis, histopathology, and immunology of human disease [5]. The intra-anorectal administration of acetic acid at a diluted concentration induces non-transmural inflammation signifies by massive necrosis of the mucosal and submucosal layer, vascular dilation, edema, and ulceration followed by infiltration of neutrophils [3], [4], [5]. The mechanistic insight in the induction process suggests that the acid dissipates protons over the intracellular spaces leading to acidification that results in damage to the intestinal layer [3]. Moreover, the induction of oxidative stress triggers the production of reactive oxygen species including superoxide anion, hydrogen peroxide, hydroxyl radical, and nitrogen species that arouse the pathogenesis of UC [6]. However, several factors are essential to be monitored and highly influence the development of an appropriate colitis model wherein the level of each determinant has to be selected carefully for standardizing the experimental design to get an optimal result. However, the lack of information regarding the optimum level of the experimental parameters for generating UC in the animal model limits the success of the experimental UC model as a tool to study the human UC and the conception of appropriate intervention strategies. Taking this as a background, in this study, we have exploited a Box-Behnken Design-based Response Surface Methodology (RSM) approach for the statistical optimization of six critical factors viz. initial body weight, acetic acid concentration, the volume of acetic acid, number of doses, period of induction, and time of holding in Trendelenburg position that is considered as an influential determinant in the experimental induction of UC in laboratory animal [7]. The optimized statistical model depicted a second-order polynomial equation that efficiently predicts the optimum value of each experimental determinant and the equation can also be utilized for predicting the level of UC induction using a given set of parameters with definite values.

Methodological protocols

Animal procurement and ethical clearance

Male Sprague Dawley rats (weighing 100–140 gm) were purchased from registered laboratory animal breeders and dealers and acclimatized under standard laboratory conditions for one week before the start of experiments for acclimatization. All animals were carefully handled following the guidelines of the Committee for the purpose of Control and Supervision of Experiments on Animals (CPCSEA), India for the Care and Use of Laboratory Animals. The study is approved by the Animal Ethics Committee of Kazi Nazrul University, Asansol, West Bengal, India.

Induction of colitis

An animal model of UC will be prepared in Sprague Dawley rats following standard laboratory methods by Millar et al., 1996 [8]. In brief, UC will be induced experimentally by intrarectal administration of 2%−4% acetic acid (pH 2.3) using a 5FG feeding tube. The feeding tube is lubricated with sterilized lubricant and inserted 5 cm into the anorectal tract. Air (2 ml) is injected keeping the rat in an upright position after the instillation of the enema such that acetic acid spreads completely in the colon. Immediately after that, the rat was kept in the Trendelenburg position as per the experimental design. UC will be physically monitored by checking the weight loss, diarrhea, anorectal bleeding, and/or fecal occult blood.

Assessment of disease activity index (DAI)

The disease activity index (DAI) was calculated as the sum of weight loss, diarrhea, and bleeding, according to the criteria described by Cooper et al., 1993 [9]. The weight loss scores were determined as follows: 0 for no loss; 1 for 1–5% weight loss; 2 for 5–10% weight loss; 3 for 10–15% weight loss; and 4 for more than 15% weight loss. The appearance of blood in the stool was measured by a benzidine test and was given a score from 0 to 4, defined as follows: 0 for no blood; 2 for positive hemoccult; and 4 for gross bleeding. The severity of diarrhea was given a score from 0 to 4, defined as follows: 0 for well-formed pellets; 2 for pasty and semi-formed stools; and 4 for liquid stools.

Experimental design for optimization of disease condition

The RSM based optimization involves a six-factor Box-Behnken Design consisting of 54 different runs. Six different parameters that are used as factors including the variable range were initial body weight (BW, ranges 100–140 gm), acetic acid concentration (AC, ranges between 1 and 4%), acetic acid volume (AV, 1–2 ml), number of doses (D, 1–3 dose), period of induction (P, 6–42 h), and time of holding in Trendelenburg position (T, 1–5 min) (Table 1). Each variable had two levels i.e., lower and upper limits as described in Table 1. The dependent response selected against these variables is weight loss (WL), the severity of diarrhea (DS), the appearance of blood in the stool (BS), and finally ulcerative colitis (UC). To carry out the experimental runs, n = 6 Sprague Dawley rats are used for induction of UC as per the designed experimental parameters (Table 2). Upon obtaining the experimental results from each run, a separate table was assigned for optimization of the UC condition. In this design, the previously established runs were included followed by only one dependent response i.e., the severity of ulcerative colitis (UC). The experimental value of the severity of ulcerative colitis (UC) was determined based on the average value of weight loss (WL), the severity of diarrhea (DS), and the appearance of blood in the stool (BS) from the previous experiment. Multiple linear regression analysis was studied and experimental data were fitted to a second-order polynomial model. All the experimental sets and responses for optimization were subjected to be tested statistically using Design-Expert Software v8.0 (Stat-Ease Inc., Minneapolis, USA). The significance of the regression model was tested statistically using ANOVA.

Table 1.

List of variable factors and responses along with their value limits.

Factor	Name of Factor	Abbreviation	Units	Minimum	Maximum	Mean	Std. Dev.
A	Initial Body weight	BW	gm	100	140	119.6296	13.04734
B	Acetic acid concentration	AC	%	1	4	2.5	1
C	Acetic acid volume	AV	ml	1	2	1.5	0.333333
D	Doses	D	No of doses	1	3	2	0.666667
E	Period of induction	P	hour	6	42	24	12
F	Time of holding in Trendelenburg position	T	minutes	1	5	3	1.333333
Response
Y1	Weight Loss	WL	Disease activity Index	0	4	1.592593	0.659288
Y2	Severity of Diarrhoea	DS	Disease activity Index	0	4	1.259259	0.994042
Y3	Appearance of Blood in Stool	BS	Disease activity Index	0	4	0.444444	0.768892
R1	Ulcerative colitis	UC	Severity	0.333333	3.333333	1.098765	0.693273

Table 2.

Design matrix for run of experiments.

Run	Factor 1 A:BW (gm)	Factor 2 B:AC (%)	Factor 3 C:AV (ml)	Factor 4 D:D (No of doses)	Factor 5 E:P (hour)	Factor 6 F:T (minutes)
1	100	2.5	1.5	1	42	3
2	120	4	1.5	2	42	5
3	140	2.5	1	2	24	5
4	120	1	1.5	2	6	1
5	100	1	1.5	1	24	3
6	120	2.5	1.5	2	24	3
7	100	2.5	1	2	24	1
8	120	2.5	1	3	24	1
9	120	2.5	1.5	2	24	3
10	140	2.5	2	2	24	5
11	100	2.5	1.5	3	6	3
12	120	2.5	1	1	24	5
13	140	2.5	2	2	24	1
14	140	4	1.5	3	24	3
15	120	1	1.5	2	42	1
16	120	2.5	1	1	24	1
17	120	1	1.5	2	42	5
18	120	4	1.5	2	6	5
19	140	1	1.5	3	24	3
20	120	4	1.5	2	6	1
21	120	2.5	1.5	2	24	3
22	120	2.5	2	3	24	5
23	120	1	1	2	42	3
24	100	4	1.5	3	24	3
25	100	2.5	1	2	24	5
26	140	2.5	1.5	1	6	3
27	120	1	1	2	6	3
28	120	2.5	1.5	2	24	3
29	120	2.5	2	3	24	1
30	120	1	2	2	42	3
31	120	4	1	2	6	3
32	140	2.5	1.5	1	42	3
33	120	4	1.5	1	24	3
34	120	1	1.5	2	6	5
35	120	2.5	2	1	24	5
36	120	2.5	1	3	24	5
37	140	2.5	1.5	3	42	3
38	100	2.5	2	2	24	5
39	120	4	2	2	6	3
40	120	2.5	2	1	24	1
41	100	2.5	1.5	3	42	3
42	140	2.5	1	2	24	1
43	120	4	1.5	2	42	1
44	120	2.5	1.5	2	24	3
45	120	2.5	1.5	2	24	3
46	120	4	2	2	42	3
47	100	1	1.5	3	24	3
48	100	4	1.5	1	24	3
49	120	4	1	2	42	3
50	120	1	2	2	6	3
51	140	1	1.5	1	24	3
52	100	2.5	2	2	24	1
53	140	2.5	1.5	3	6	3
54	100	2.5	1.5	1	6	3

Colon, spleen, and liver index

The dissected entire colon length was measured using a scale in centimeters (cm) and weighted in grams, and the ratio was calculated as follows:

• The ratio of colon length/weight = length by cm/weight in grams.

• The spleen and liver, two major organs, were weighed. The organ index was calculated based on the animal's body weight:

• Organ index (liver/spleen) = (organ weight in grams/animal body weight in grams)

Blood collection and hematological assessment

The blood samples from all the rats were drawn via cardiac puncture and collected into separate tubes. 1 ml of blood was collected in a microcentrifuge tube containing 0.1 ml of 10% EDTA added as an anticoagulant and another 1 ml in plain vail for various hematological tests including complete blood count and serum hemoglobin level.

Method validation

Fitting of model for evaluation and optimization

The design matrix generated using six variable factors was evaluated for the response surface linear model and shows no aliases for a linear model. The degree of freedom for the design matrix is shown in Table 3. A good design shows a minimum of 3Df for lack of fit and 4Df for pure error, and our design is suitable for testing. The standard error and variance inflation factors (VIF) are within the acceptable limit and the model is ideal (Table 4). The R_i-square is the multiple correlation coefficient factor and is very low for our model suggesting it is a good model.

Table 3.

Degree of freedom of the design matrix for evaluation.

Degrees of Freedom for Evaluation
Model	6
Residuals	47
Lack Of Fit	42
Pure Error	5
Corr Total	53

Table 4.

Evaluation of the design matrix for different variable factors showing standard deviation and standard errors.

Factors	Standard Error	VIF	Ri-Squared	Power at 5% alpha level to detect signal/noise ratios of
				0.5 Standard deviation	1 Standard deviation	2 Standard deviation
A	0.208978	1.003639	0.003626	21.6%	64.9%	99.7%
B	0.20431	1.00182	0.001816	22.4%	66.9%	99.8%
C	0.204124	1	0	22.4%	67.0%	99.8%
D	0.20431	1.00182	0.001816	22.4%	66.9%	99.8%
E	0.204124	1	0	22.4%	67.0%	99.8%
F	0.204124	1	0	22.4%	67.0%	99.8%

Upon exploring the graphical plots for the evaluation of the design matrix, the perturbation plot shows curvature in the factors that signify that the responses are sensitive to the different factors (Fig. 1A). The perturbation plot helps to compare the effect of all the factors at a particular point in the design space. The 3D surface plot shows the projection of the contour plot giving shape to the color and showing the response against the different selected factors. The 3D plot for the standard error of our design shows that the model is fit for testing and good for optimization (Fig. 1B-E). Next, we explore the fitness of the different responses generated against the different factors and found all three responses show a linear model having an insignificant lack of fit, high adjusted R-square, and predicted R-square value that is statistically significant (Table 5, 6). The experimental outcomes of the Box-Behnken design were fitted in a second-order polynomial equation that depicts the empirical relationship between the responses and the independent variable. The generalized equation for regression is as follows:

$$\mathrm{Y}={\beta }_0+\sum _{\mathrm{i}=1}^{\mathrm{n}}\left({\beta }_i{X}_i\right)+\sum _{\mathrm{i}=1}^{\mathrm{n}}\left({\beta }_{ii}{X}_i^2\right)+\sum _{\mathrm{i}>\mathrm{j}}^{\mathrm{n}}\left({\beta }_{ij}{X}_i{X}_j\right)$$ (1)

Fig. 1.

Graphical representation of good fit of the model for testing. A. Perturbation plot showing the distribution of different factors across the standard error of the design. B.-E. 3D surface plot for relation of various parameters against the standard error of the design.

Table 5.

Summary Table for the different responses.

Response	Model	Sequential	Lack of Fit	Adjusted	Predicted
		p-value	p-value	R-Squared	R-Squared
Y1: Weight loss	Linear	< 0.0001	0.6135	0.426004	0.329783
Y2: Severity of Diarrhoea	Linear	0.0002	0.4761	0.334451	0.21538
Y3: Appearance of Blood in Stool	Linear	0.0018	0.2914	0.267279	0.126625
Ulcerative colitis severity	Linear	< 0.0001	0.3337	0.391201	0.278234

Table 6.

Detailed evaluation of design matrix for different responses showing sequential model sum of square, lack of fit test, model statistics.

Sequential Model Sum of Squares [Type I]
Response	Model	Sum of	df	Mean	F	p-value
		Squares		Square	Value	Prob > F
Y1: Weight loss	Mean vs Total	136.963	1	136.963
	Linear vs Mean	11.31084	6	1.885139	7.555861	< 0.0001
Y2: Severity of Diarrhoea	Mean vs Total	85.62963	1	85.62963
	Linear vs Mean	21.46119	6	3.576865	5.438924	0.0002
Y3: Appearance of Blood in Stool	Mean vs Total	10.66667	1	10.66667
	Linear vs Mean	10.97383	6	1.828971	4.222186	0.0018
Ulcerative colitis severity	Mean vs Total	65.19342	1	65.19342
	Linear vs Mean	11.72078	6	1.953464	6.676098	< 0.0001

Lack of Fit Tests
Response	Source	Sum of	df	Mean	F	p-value
		Squares		Square	Value	Prob > F
Y1: Weight loss	Linear	10.39287	42	0.247449	0.927935	0.6135
	Pure Error	1.333333	5	0.266667
Y2: Severity of Diarrhoea	Linear	28.07584	42	0.668472	1.179657	0.4761
	Pure Error	2.833333	5	0.566667
Y3: Appearance of Blood in Stool	Linear	19.02617	42	0.453004	1.698766	0.2914
	Pure Error	1.333333	5	0.266667
Ulcerative colitis severity	Linear	12.77098	42	0.304071	1.549041	0.3337
	Pure Error	0.981481	5	0.196296

Model Summary Statistics
Response	Source	Std.	R-Squared	Adjusted	Predicted	PRESS
		Dev.		R-Squared	R-Squared
Y1: Weight loss	Linear	0.499493	0.490985	0.426004	0.329783	15.43981
Y2: Severity of Diarrhoea	Linear	0.810951	0.409796	0.334451	0.21538	41.09086
Y3: Appearance of Blood in Stool	Linear	0.658165	0.350228	0.267279	0.126625	27.36575
Ulcerative colitis severity	Linear	0.54093	0.460121	0.391201	0.278234	18.38573

In this equation, Y presents the assumed response value, and n represents the total number of independent variables; β₀, β_i, β_ii, and β_ij present regression coefficients of intercept, linear, quadratic, and interaction effects respectively.

Statistical optimization of response 1: body weight loss (WL)

The main focus of this study is to develop an augmented UC model by determining the optimal level of the severity of the disease activity index viz., body weight loss, the severity of diarrhea, and the appearance of blood in the stool. ANOVA for response surface reduced linear model was generated with backward elimination regression with alpha to exit=0.10 for all three different responses. ANOVA analysis for R1:WL is summarized in Table 7 and shows the model F-value of 14.90 that infers the model is significant. The F-value is the mean square for a term divided by the mean square for the residual. However, there is only a 0.01% chance that a large model F-value could occur due to noise. Prob>F value is the probability of getting an F-value of this size if the term did not have an effect on the response and it should be <0.05 for a significant model. In our model, the prob>F value is <0.05 for factors B:AC, D:D, and E:P and are significant. Moreover, the insignificant model was removed by backward elimination regression and depicted in Table 8. The F-value= 0.90 for lack of fit implies lack of fit is not significant relative to the pure error. There is a 63.05% chance that a lack of fit F-value this large could occur due to noise. However, the insignificance of lack of fit suggests the fitness of the model. The R² is the measure of the amount of variation around the mean, while adjusted R² measures the amount of variation around the mean adjusted for the number of terms in the model. In our model, the predicted R² of 0.3831 is in reasonable agreement with the adjusted R² of 0.4404 suggesting the model is significant (Table 9). Adequate precision measures the signal-to-noise ratio and it should be greater than 4 for desirability. Our model ratio of 13.037 indicates a satisfactory signal and can be used to navigate the design space. The model shows standard deviation (SD), mean, and predicted residual error sum of squares (PRESS) values of 0.49, 1.59, and 14.21 respectively (Table 9). The calculated value of the coefficient of variation (CV%) is 30.97 intitles improved the reliability of the developed model. The experimental outcomes of the Box-Behnken design for optimization of Y1: the weight loss was fitted in a second-order polynomial equation that depicts the empirical relationship between the responses: WL and three significant variable factors viz., AC, D, and P and are given in Eq. (2):

$$\text{WL}\left(\text{DAI}\right)=+1.59+0.37\times \text{AC}(\%)+0.25\times \mathrm{D}\left(\text{No}.\text{of}\text{Dose}\right)+0.50\times \mathrm{P}\left(\text{hours}\right)$$ (2)

Table 7.

Analysis of variance (ANOVA) for Y1: Weight Loss (WL).

Source	Sum of Squares	df	Mean Square	F Value	p-value Prob > F	Significance
Model	10.875	3	3.625	14.90293	< 0.0001	Significant
B-AC	3.375	1	3.375	13.87514	0.0005
D-D	1.5	1	1.5	6.16673	0.0164
E-P	6	1	6	24.66692	< 0.0001
Residual	12.16204	50	0.243241
Lack of Fit	10.8287	45	0.240638	0.902392	0.6305	Not Significant
Pure Error	1.333333	5	0.266667
Correlation Total	23.03704	53

Table 8.

Backward eliminated factors for Y1: Weight Loss (WL).

Removed	Coefficient Estimate	t for H0 Coeff=0	Prob > |t|	R-Squared	MSE
C-AV	−0.04167	−0.40866	0.684644	0.489176	0.245164
F-T	−0.08333	−0.82451	0.413728	0.481941	0.243562
A-BW	−0.09968	−0.96647	0.338555	0.472066	0.243241

Table 9.

Statistical analysis and values of three different responses.

Response	Std. Dev.	Mean	C.V.%	PRESS	R2	Adjusted R2	Predicted R2	Adequate Precision
Y1: Weight loss (WL)	0.493194	1.592593	30.96802	14.21238	0.472066	0.44039	0.383064	13.03727
Y2: Severity of Diarrhoea (DS)	0.783765	1.259259	62.2402	35.42687	0.401786	0.378326	0.323532	13.08178
Y3: Appearance of blood in stool (BS)	0.643596	0.444444	144.8091	23.70752	0.325798	0.299359	0.243377	11.53615

The adequacy of the developed model was further analyzed for various diagnostic plots. The normal probability plot indicates that the residuals follow a normal distribution following a straight line (Fig. 2A). The residual vs experimental run plot checks for lurking variables that influence the response during the experiment showing random scattering. Our data suggest the significance of the residual vs experimental plot (Fig. 2B). The predicted vs actual plot depicts the distribution of the actual responses and the predicted responses (Fig. 2C). The perturbation plot describes the comparative influence of the variable factors on the response (WL) at the midpoint (coded 0.0) in the design space (Fig. 2D). The cube plot (Fig. 2E) collectively represents the effect of the three significant factors AC, D, and P on the response Y1: WL. The correlation between the different parameters on the response was determined through the 3D plot (Fig. 2F). Our data indicate the increase in value of AC, D, and P increase the outcome of response (WL), while the other factors BW, AV, and T have not much influence on the outcome.

Fig. 2.

Diagnostic plots and graphical plots for analysis of response of weight loss (WL). Diagnostic plots showing distribution of different residues across A. normal probability plot, B. residual vs run plot. C. predicted vs actual plot. D. perturbation plot, E. Cube plot. F. 3D surface plot depicting the effects of multiple parameters on the response of weight loss (WL).

Optimization of response 2: severity of diarrhea

The ANOVA analysis for R2:DS is summarized in Table 10 and shows the model F-value of 17.13 and prob>F value of < 0.0001 signifying that the model is significant. The factors B:AC and D:D are considered to be significant having prob>F-value less than 0.05. The other parameters viz., A:BW, C:AV, E:P, and F:T are eliminated by backward elimination regression (Table 11). The lack of fit F-value of 1.09 implies its insignificance relative to the pure error suggesting the model is fit. The predicted R² of 0.3235 is in reasonable agreement with the adjusted R² of 0.3783 suggesting the model is significant. An adequate precision value of 13.082 shows the desirability of the model to navigate into the design space. The standard deviation of 0.78, mean value of 1.26, PRESS value of 35.43, and high coefficients of variance (CV%) of 62.24 further validate the significance of our model (Table 9). The model for response 2 Y2:DS is fitted in a second ordered polynomial equation showing an empirical relationship between B:AC and D:D as follows:

$$\text{DS}\left(\text{DAI}\right)=+1.26+0.88\times \text{AC}(\%)+0.33\times \mathrm{D}\left(\text{No}.\text{of}\text{Dose}\right)$$ (3)

Table 10.

Analysis of variance (ANOVA) for Y2: Severity of Diarrhoea (DS).

Source	Sum of Squares	df	Mean Square	F Value	p-value Prob > F	Significance
Model	21.04167	2	10.52083	17.12687	< 0.0001	significant
B-AC	18.375	1	18.375	29.91266	< 0.0001
D-D	2.666667	1	2.666667	4.341067	0.0422
Residual	31.3287	51	0.614288
Lack of Fit	28.49537	46	0.619465	1.093173	0.5201	not significant
Pure Error	2.833333	5	0.566667
Corr Total	52.37037	53

Table 11.

Backward eliminated factors for Y2: Severity of Diarrhoea (DS).

Removed	Coefficient Estimate	t for H0 Coeff=0	Prob > |t|	R-Squared	MSE
E-P	−2.9E-17	0	1	0.409796	0.643941
F-T	−0.04167	−0.25437	0.800294	0.409001	0.63165
C-AV	0.083333	0.513672	0.60979	0.405818	0.62235
A-BW	−0.09604	−0.58254	0.562824	0.401786	0.614288

The diagnostic plot and the graphical plot for the experimental results show the normal distribution of the experimental residue in the normal probability plot (Fig. 3A), and the fitness of residual vs experimental run variable (Fig. 3B). The distribution of actual and predicted responses shows acceptance of the response result (Fig. 3C). The perturbation plot describes the comparative influence of the variable factors on the response (WL) at the midpoint (coded 0.0) in the design space (Fig. 3D). The cube plot (Fig. 3E) collectively represents the effect of the three factors AC, D, and BW on the response Y2:DS. The correlation between different parameters on the response was determined through the 3D plot (Fig. 3F). Our data indicate the increase in the value of AC and D increases the outcome of response (DS), while the other factors BW, AV, P, and T have a little influence on the outcome.

Fig. 3.

Analysis for response of severity of diarrhea (DS). Diagnostic plots representing the distribution of residues across A. normal probability plot, B. residual vs run plot. C. predicted vs actual plot. D. perturbation plot, E. cube plot. F. 3D surface plot for the combined effect of two parameter at a same time in the response of severity of diarrhea (DS) respectively.

Optimization of response 3: appearance of blood in stool

The result of ANOVA for the statistical analysis of R3:BS demonstrates an F-value of 12.32 implies that the model is significant. Prob>F-value of B:AC and D:D are <0.0001 and 0.0309 lesser than 0.005 that collectively indicates that the model is significant (Table 12), while the other parameters are insignificant and are eliminated (Table 13). The "lack of fit F-value" of 1.61 implies that the lack of fit is not significant relative to the pure error. There is a 31.47% chance that a large "lack of fit F-value" could occur due to noise. Non-significant lack of fitness is good for the fitness of the model. The R², adjusted R^2, and predicted R² value shows the significance of the model, followed by mean (0.44), standard deviation (0.64), PRESS (23.71), and CV (144.81) (Table 9). The equation for the response Y3:BS fitting a second-order polynomial equation is presented as follows:

$$\text{BS}=+0.44+0.58\times \text{AC}(\%)+0.29\times \mathrm{D}\left(\text{No}.\text{of}\text{Dose}\right)$$ (4)

Table 12.

Analysis of variance (ANOVA) for Y3: Appearance of blood in stool (BS).

Source	Sum of Squares	df	Mean Square	F Value	p-value Prob > F	Significance
Model	10.20833	2	5.104167	12.32249	< 0.0001	significant
B-AC	8.166667	1	8.166667	19.71598	< 0.0001
D-D	2.041667	1	2.041667	4.928994	0.0309
Residual	21.125	51	0.414216
Lack of Fit	19.79167	46	0.430254	1.613451	0.3147	not significant
Pure Error	1.333333	5	0.266667
Corr Total	31.33333	53

Table 13.

Backward eliminated factors for Y3: Appearance of blood in stool (BS).

Removed	Coefficient Estimate	t for H0 Coeff=0	Prob > |t|	R-Squared	MSE
C-AV	−1.5E-17	0	1	0.350228	0.424156
F-T	−0.08333	−0.62685	0.53373	0.344909	0.418902
A-BW	−0.09887	−0.73097	0.468279	0.337766	0.415
E-P	−0.125	−0.95059	0.346387	0.325798	0.414216

The normal probability plot shows the normal distribution of all across the straight line except for the two runs (Fig. 4A). The residual vs run plot signifies the variables that influence the response (Fig. 4B). The actual vs predicted plot for BS shows the equal distribution of residues across the 45° line (Fig. 4C). Moreover, the perturbation plot (Fig. 4D) shows a steep slope signifying the response is sensitive to the factors, while the cube plot (Fig. 4E) demonstrates the collective impact of AC, D, and BW on the optimize value of BS. The 3D plots of our experimental data reveal the connection of an increase in AC and D over the increase of P enhancing the value of BS (Fig. 4F).

Fig. 4.

Representation of various plots for response of appearance of blood in stool (BS). A. normal probability plot, B. residual vs run plot. C. predicted vs actual plot. D. perturbation plot, E. cube plot, showing the different residues in the design space for experimental run. F. 3D surface plot depicting the collective influence of different parameters on the response of appearance of blood in stool (BS).

Statistical analysis of the severity of UC

The experimental value for the severity of UC was obtained from the average of the disease activity index obtained for each run. The score of severity of UC ranges from 0 to 4, where 0 for no induction, 1 for initiation of the disease, 2 for active UC, and above 3 indicates extreme colitis and 4 for the death. The summary of the design matrix was previously explained in Tables 5, 6. The statistical analysis of ANOVA for the response surface reduced linear model shows an F-value of 19.39 and prob>F-value <0.0001 inferring that the model is significant (Table 14). The factors B:AC and D:D are significant to the model and the other factors A:BW, C:AV, E:P, and F:T are eliminated due to a higher prob>F-value using backward elimination regression (Table 15). The F-value for lack of fit shows 1.49 implying that lack of fit is not significant relative to the pure error and thus the model is a good fit. The predicted R² of 0.3587 is in reasonable agreement with the adjusted R² of 0.4097 suggesting the model is significant. The mean, standard deviation, and PRESS values are 1.10, 0.53, and 16.34 respectively supporting the good fit of the model and are significant (Table 16). The adequate precision ratio of 14.382 indicates an adequate signal for the model to be used for navigating the design space. Finally, the diagnostic and the graphical plots validate the results. The normal vs residual plots indicate the distribution of all the residues across the standard line (Fig. 5A). Similarly, the residual vs run plot and the predicted vs actual plot validate that all the variables are within the limit for fitting the model (Fig. 5B, 5C). The steep slope of the perturbation plot (Fig. 5D) signifies that the variables have a significant influence over the response. Lastly, the cube plot validates the effect of the BW, AC, and D over the severity of UC (Fig. 5E). This fact was further confirmed by the various 3D plots of the contour for different parameters over the changes in the response. As per the analysis, the intensification in acetic acid concentration and the number of dose, the response of UC increases significantly (Fig. 5F). However, these factors are independent of body weight, the volume of acetic acid, and the time of holding. This validates the result for the optimization of the response. The second-order polynomial equation fitted for the response of severity of UC is given as follows:

$$\text{UC}=+1.10+0.61\times \text{AC}(\%)+0.29\times \mathrm{D}\left(\text{No}.\text{of}\text{Dose}\right)$$ (5)

Table 14.

Analysis of variance (ANOVA) for Severity of Ulcerative colitis (UC).

Source	Sum of Squares	df	Mean Square	F Value	p-value Prob > F	Significance
Model	11.00463	2	5.502315	19.39494	< 0.0001	significant
B-AC	8.962963	1	8.962963	31.59327	< 0.0001
D-D	2.041667	1	2.041667	7.196608	0.0098
Residual	14.46862	51	0.283698
Lack of Fit	13.48714	46	0.293199	1.493654	0.3515	not significant
Pure Error	0.981481	5	0.196296
Corr Total	25.47325	53

Table 15.

Backward eliminated factors for Severity of Ulcerative colitis (UC).

Removed	Coefficient Estimate	t for H0 Coeff=0	Prob > |t|	R-Squared	MSE
C-AV	0.013889	0.125786	0.900438	0.45994	0.286606
F-T	−0.06944	−0.63548	0.528135	0.455396	0.283119
A-BW	−0.09819	−0.88308	0.381505	0.446729	0.281872
E-P	0.125	1.153425	0.25422	0.432007	0.283698

Table 16.

Statistical analysis and value of ANOVA for Severity of Ulcerative colitis (UC).

Std. Dev.	Mean	C.V.%	PRESS	R2	Adjusted R2	Predicted R2	Adequate Precision
0.532634	1.098765	48.47563	16.33652	0.432007	0.409733	0.35868	14.38198

Fig. 5.

Analysis of different parameters on the response of Ulcerative colitis induction (UC). Graphical representation of experimental residues over the design space and their effect on response showing A. normal probability plot, B. residual vs run plot. C. predicted vs actual plot. D. perturbation plot, E. cube plot. F. The 3D surface plot for UC showing the effect of two different parameters on the response.

Optimization of the response to the severity of UC

All the aforesaid experimental data and validation reported the effect of different factors in response to the induction of UC. Now the optimization of the response of UC was obtained by setting-up different goals for the different factors viz., BW- in a range between the upper and lower limit, AC- maximize, AV- minimize, D- minimize, P- in a range between upper and lower limit, and T- in a range between upper and lower limit. The goal for the yield of response i.e., UC is set to be maximized. The optimized model was selected for the highest yield of the severity of UC with a value of BW= 120 gm, AC= 4%, AV= 1.5 ml, D = 1 dose, P = 24 h, and T = 5 mins. The value of UC obtained is 1.41821 and the solution desirability for the model having the maximum value of 0.712 is selected. The point prediction for the UC is 1.41821 having a standard deviation of 0.5326 and a standard error of mean value of 0.1699 (Table 17). The graphical representation of the effect of different parameters on the desirability of the response is depicted by 3D surface plot in Fig. 6.

Table 17.

Point prediction value for the optimized ulcerative condition.

Response	Prediction	Standard Deviation	Standard error Mean	95% Confidence interval low	95% Confidence interval high	Standard Error Predicted	95% Prediction Interval low	95% Prediction Interval high	95% Tolerance interval low	95% Tolerance interval high
Ulcerative colitis (UC)	1.41821	0.532634	0.169986	1.076949	1.759471	0.559101	0.295768	2.540652	−0.50869	3.345111

Fig. 6.

3D surface plots for the optimized UC model showing the effect of different parameters on the desirability of the optimum model.

Experimental validation of the optimized model for induction of UC

The experimental validation of the optimized result obtained from the previous section was carried out in n = 6 Sprague Dawley rats. The different parameters used for the induction process were maintained as per the optimal result and the group is marked as optimized UC model group. For cross-comparing the experimental result, n = 6 healthy rats were instilled with normal saline and marked as the control group. Another n = 6 rats were administered with the un-optimized values of the different parameters and marked as an un-optimized UC model group. Furthermore, to validate the success of the acetic acid-induced UC model, we have compared the efficacy of acetic acid-induced UC with DSS-induced UC experimental model. Herein, n = 6 Sprague Dawley rats were fed with 5% (w/v) DSS (M.W: 35–55 kDa) in drinking water for seven consecutive days following an earlier report [10]. The disease activity index was monitored for all the groups and the rats were euthanized after 24 h . The blood collected via heart puncture was used for the different hematological tests. The plots for the disease activity index depicted in Fig. 7A show that the results obtained from the RSM model for the optimization of UC are validated by the in-vivo study. The colon, liver, and spleen indexes for the different groups are shown in Table 18 and Fig. 7B. The volume of hemoglobin in the blood decreases significantly in the optimized model inferring a good UC model. Moreover, the plots for the complete blood count also signify a similar result for validation of a good UC model in the optimized group than the unoptimized group (Fig. 7C). The cross-comparison between the acetic acid-induced UC and DSS-induced UC model is shown in Table 19.

Fig. 7.

Graphical representation of the experimental validation of the optimized result in comparison to DSS-induced UC. A. Disease activity index showing effect of weight loss, severity of diarrhea, and appearance of blood in stool. B. Colon of normal healthy rat, unoptimized UC model, optimized UC model, and DSS-induced UC model. C. Plots for various hematological parameters for the experimental model. Here, blue color indicates control healthy group, green color displays unoptimized UC model group, red color shows optimized UC model group, and orange color denotes DSS-induced UC model.

Table 18.

Organ Index for experimentally induced ulcerative colitis in Sprague Dawley rats.

Organ Index	Control	Unoptimized UC model	Optimized UC model	DSS-induced UC model
Colon Index	1.666	1.846	1.33	1.879
Spleen Index	0.021	0.015	0.016	0.015
Liver Index	0.083	0.057	0.058	0.057

Table 19.

Comparative analysis of Optimized acetic acid-induced and DSS-induced experimental animal models for UC.

Sl. No.	Parameters	RSM optimised acetic acid-induced UC model (present study)	DSS-induced UC model (Das et al., [10])
1.	Process of induction	Single dose of 1.5 ml of 4% acetic acid injected using lubricated and sterilized feeding tube inserted into the anorectal tract keeping the rat in upside down position for 5 mins.	5% (w/v) DSS fed for 7 consecutive days.
2.	Mechanism of induction	Protonated form of acid releases protons over the intracellular spaces of the gut epithelium causing immense epithelial damage. It also causes non-transmural inflammation leading to the production of superoxide ions and ROS causing neutrophil infiltration, vascular dilation, edema and necrosis of mucosal and submucosal layers.	DSS induce dysbiosis in the gut microbial homeostasis and causes toxicity in the gut epithelium. It disrupts the gut epithelial cells integrity resulting in mucosal damage leading to immune cell infiltration and profound inflammatory responses.
3.	Time period for induction	1 day	7 days
4.	Cost for induction process	Minimum	High cost of DSS
5.	Chances of success of model	High	Moderate
6.	Advantage	It is easy to replicate, time saving, cost efficient and mimics the human UC model. It is beneficial for studying the immunological parameters associated with the pathogenesis of UC.	It is efficient for generating chronic UC model and also colitis-associated cancer model. It is beneficial for studying gut epithelial damage and changes in gut microbial homeostasis resulting in UC.
7.	Limitation and drawback	Intrarectal administration of acetic acid at un-appropriate dose may lead to burning of the epithelial layer leading to death.	It is very much time taking and depends on the amount of water intake by the rat. It has moderate chances of generating disease model.

Outcomes

This research methodology aimed to generate an easy and optimized acetic-acid-induced UC model using RSM. The Box-Behnken Design was used to optimize the disease activity index for UC including weight loss, the severity of diarrhea, and the appearance of blood in the stool. Six different independent variable factors that are critical for the induction process were selected and used to generate a model that fitted a second-order polynomial equation. Our model was found to be a statistically significant and good fit. The responses obtained from the actual and predicted model for the different disease activity index were optimized to generate the UC model. Our experimental result optimized the dose and/or criteria for the different parameters for the induction of UC in the animal model. The optimum criteria for the best-fitted model suggest the use of 120 gm body weight Sprague Dawley rats for intrarectal administration of a single dose of 1.5 ml 4% acetic acid and keeping the rat in the upside down position for 5 mins followed by 24 h of observation before euthanization. This result was further validated by an in-vivo experiment and found to be an optimized UC model. In addition to this, our study cross compare the other chemically-induced UC model viz., the DSS-induced UC model with our optimized UC model and showed that our model is a very much easy, optimized, time-saving, and cost-effective method. On the other hand, the DSS-induced UC model takes 7 days time duration with a moderate probability of success and is costlier (Table 19). Thus, our study standardizes the method for induction of chemically-induced UC in rats using acetic acid and getting the best result. This optimized method is easy, cost-effective, and can be easily replicated for the generation of a UC model. The workflow of the optimized method for acetic acid-induced UC in Sprague Dawley rats is summarized in the scheme depicted in Fig. 8. Induction of UC is a multi-factorial process involving the physiological and inflammatory changes wherein several innate and adaptive immune cells like T-helper cells, natural killer cells, macrophages, and dendritic cells play important roles in releasing the proinflammatory cytokines (TNFα, interleukin 1β (IL-1β), and IL-6) and chemokines (IL-8, lL-18, MCP-1, MIP-1) [11,12]. These cytokines mediate the activation of different gut-associated immune cells like T-helper cells, natural killer cells, macrophages, goblet cells, and dendritic cells, secretion and accumulation of mucus and different acute-phase proteins as well as infiltration of other immune cells that collectively mediate the inflammatory pathology of UC [11,12]. Therefore, this optimized UC induction model could be explored further for an in-depth dissection of the immunological parameters as well as their interplay in establishing UC in rat models and this would greatly assist in developing appropriate therapeutic studies in the future.

Fig. 8.

Schematic representation of the work-flow of the methodology used for the optimization of acetic acid-induce UC in Sprague Dawley rats.

Ethics statements

All animals were carefully handled following the guidelines of the Committee for the purpose of Control and Supervision of Experiments on Animals (CPCSEA), India for the care and use of laboratory animals. The study is approved by the Animal Ethics Committee of Kazi Nazrul University, Asansol, West Bengal, India. Further, the study complied with the ARRIVE guidelines and was carried out in accordance with the U.K. Animals (Scientific Procedures) Act, 1986 and associated guidelines; EU Directive 2010/63/EU for animal experiments; or the National Institutes of Health guide for the care and use of laboratory animals (NIH Publications No. 8023, revised 1978).

Credit author statement

RP: Performed all experiments, validate the results, and drafted the manuscript; SP: Performed the experiments, and drafted the manuscript; SM: Conceptualize and designed the study, validate the results, edited the manuscript, and supervised the study.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data Availability

• Data will be made available on request.