Sex represents a relevant interaction in Sprague–Dawley rats: the example of oesophageal length*

ABSTRACT

Background: 8-week old Sprague Dawley rats represent the standard rodent model of oesophageal surgery, which is challenging and might be eased by larger oesophageal lengths. Therefore, we aimed to analyse whether oesophageal length would linearly increase with bodyweight and ensure comparable experimental conditions. Methods: We analysed 41 8-week old Sprague Dawley rats of both sexes by linear regression of oesophageal length with sex as an interaction term for bodyweight. Based on exploratory investigations, analyses were powered to 80% for a deviation of the regression’s slope from zero. Results: Linear regression was statistically significant with F(3,37) = 3.29, P=0.0312 with an adjusted R² of 0.15 (95% CI: 0.02–0.43). Oesophageal length could be modelled by 4.56 (95% CI: 1.45–6.69) + 0.007 (95% CI: −0.002–0.019) x bodyweight in grams + 6.7 (95% CI: 1.86–11.1) x sex (1=male) – 0.02 (95% CI: −0.04–(−0.005)) x bodyweight x sex. Exploration of the interaction revealed that oesophageal length increased with bodyweight for female rats, but decreased in males. Conclusions: Sex represents a major interaction for oesophageal length in an age-adjusted cohort of Sprague Dawley rats. This may have relevant implications for reproducibility of rat models of oesophageal surgery, but may be different in inbred strains.

KEYWORDS:

• Animal model
• oesophageal surgery
• rattus norvegicus
• sex differences
• simulation analysis

Introduction

Pathophysiology of any disease should be elucidated using small rodent models before it can be applied to humans. In oesophageal carcinoma, this can be done via Levrat’s rat model. (Levrat et al. 1962) Meaningful and reproducible research requires standardisation of animal experiments to achieve methodological rigour. (Macleod 2011; Ioannidis 2017) It has recently been shown that only bodyweight determined oesophageal length and resilience to traction forces in rats (Oetzmann von Sochaczewski, Tagkalos, Lindner, Baumgart, et al. 2019).

In Levrat’s model of oesophageal adenocarcinoma, rats are usually age-standardized despite a 10% variability of bodyweight for the same age. Oesophageal surgery in rats is generally considered difficult, (Cui & Urschel 1999; Tagkalos et al. 2019) exemplified by a higher mortality (Man et al. 1988; Drescher et al. 2012; Gronnier et al. 2013) than in intestinal anastomoses. (Marjanovic et al. 2010) This difficulty might be reduced in larger and heavier animals with correspondingly larger oesophagi.

We therefore tested whether oesophageal length would be affected by bodyweight if age was controlled at typical surgical age of eight weeks. In addition, we tested whether sex would be a relevant interaction term.

Materials and methods

Rats and husbandry

We bought Sprague–Dawley (RjHan:SD) outbred strain rats (Rattus norvegicus) as weanlings from Janvier Labs (Le Genest-Saint Isle, France). They were housed at our closed facility under standard husbandry conditions in groups of 3 animals at maximum per type-VI cage until they reached the age of eight weeks (day of life 57–59 in our study). Cages were equipped with nest material from European aspen (populus tremula) with a size of 2–3 mm (Abbedd midi, Vienna, Austria), a tunnel, and environmental enrichments. Rats had access to standard rat-chow (ssniff Ratte/Maus-Haltung Extrudat, ssnif-Spezialdiäten, Soest, Germany) and water ad libitum. We used a dark–light-cycle of 12 h starting at 07:00 in a room with 67% relative humidity at 23°C and an hourly air-exchange of 12 times the room’s volume.

We included 41 rats of both sexes (30 ♀, mean bodyweight: 246 g, standard deviation: 12.3; 11 ♂, mean bodyweight: 367.2 g, standard deviation: 24.1) aged 57 (10 rats; 4 ♂), 58 (16 rats; 3 ♂), and 59 (15 rats; 4 ♂) days to resemble an experimental group of eight weeks old animals. These animals were in their adolescent phase of life according to Sengupta (Sengupta 2013). Raw data of our experiments are available as supplemental information 1.

Regulatory framework

Our experiments were compliant with the directive 2010/63/EU and the national regulations for the protection of animals. The German law for the protection of animals exempts all experiments in which laboratory animals are sacrificed to obtain isolated organs from approval by the competent state authority (exact citation: section 7 subsection 2 sentence 3 of the German law for the protection of animals [German citation: § 7 II 3 TierSchG]).

Literature search to support rat strain choice

Sprague–Dawley rats were chosen based on an assessment of rodent models. We searched MEDLINE via PubMed at the 3rd of June 2017 with the search query ‘(rat OR mouse OR mice OR murine OR rodent) AND esophagus AND surgery AND anastomo*’. We identified eight reports using mice and 70 using rats, among them 45 used a total of 2,319 Sprague–Dawley followed by 20 reports using 718 Wistar rats.

Determination of oesophageal length

We separately euthanized rats by slowly increasing carbon dioxide concentrations (20% of the cage volume until 100% of carbon dioxide were reached as described previously (Marquardt et al. 2018)). Killing was done in a chamber that was located a different room than the home cage. We explanted intact oesophagi conjointly in a cluster of viscera with larynx, trachea, hypopharynx, and parts of the cardia. Identification of the upper oesophageal sphincter was based on muscle fibres of the musculus cricopharyngeus that originate at the cricoid cartilage and marked the upper oesophageal sphincter. The lower oesophageal sphincter was identified by the transitional zone of gastric and oesophageal mucosa. The procedure of identification of the anatomical landmarks and the necessary steps to isolate the cluster of viscera has been described and depicted before in detail in swine (Oetzmann von Sochaczewski, Tagkalos, Lindner, Lang, Heimann, & Muensterer 2019; Oetzmann von Sochaczewski, Tagkalos, Lindner, Lang, Heimann, Schröder, et al. 2019; Oetzmann von Sochaczewski et al. 2020). Oesophagi were then dissected free from surrounding tissue and neighbouring organs. We measured oesophageal length using an electronic slide gauge (VWR, Darmstadt, Germany) with a resolution of 0.01 mm and an accuracy of 0.03 mm.

Statistical analysis

All analyses were conducted using the stats4-package (version 3.5.1) (R Core Team 2019) unless indicated otherwise. The assumption of normality was tested separately for both sexes using the Shapiro–Wilk-test. Likewise, the data were assessed for linearity. Simulated samples were compared to the sample mean of our cohort using Welch-test. The relationship between oesophageal length and bodyweight was investigated by linear regression using sex as an interaction term. We bootstrapped 95% bias-corrected, accelerated confidence intervals for correlation analyses using the wBoot-package (version 1.0.3) (Weiss 2016) and for our regression analysis using the simpleboot-package (version 1.1-7) (Peng 2019). All bootstrap-procedures were conducted with 10,000 repetitions as recommended elsewhere (Hesterberg 2015).

Power analysis

Exploratory preliminary investigations with five rats per group formed the basis to calculate an a-priori power of 80% for a deviation of the regression’s slope from zero with two predictors (bodyweight and sex) with 41 rats in total. The power analysis was conducted using G*Power 3.1.9.2 (Faul et al. 2007).

Simulation of reference cohort

Reference values for bodyweight were obtained from Janvier Labs growth curves for Sprague–Dawley rats separately for males ($\overline{\mathrm{x}}=368.5$ g, standard deviation 44n n = 120) and females ($\overline{\mathrm{x}}=242.25$, standard deviation 25.55, n = 120) at the age of 58 days using WebPlotDigitizer (Rohtagi 2018). Simulation assumed a normal distribution of these reference populations. The R-code for the simulation is available as supplemental information 2.

Results

Oesophageal length was correlated to bodyweight in the female subset, but did not reach statistical significance (R=0.23, 95% confidence interval −0.09–0.53, P=0.16). In the male subset, oesophageal length correlated negatively to bodyweight (R=−0.53, 95% confidence interval: −0.89–(−0.001), P=0.046).

Linear regression analysis was statistically significant (F(3,37) = 3.29, P=0.031), but had a low coefficient of determination (adjusted R²=0.15, 95% confidence interval: 0.02–0.43). Oesophageal length could be predicted by the equation: $\begin{array}{rl}& 4.56(95\mathrm{\%}CI:1.45\text{--}6.69)\\ & +0.007(95\mathrm{\%}CI:-0.002\text{--}0.019)\\ & \times \text{ bodyweight in grams }\\ & +6.7(95\mathrm{\%}CI:1.86\text{--}11.1)\\ & \times \text{ sex}(1=\text{male})\\ & -0.02(95\mathrm{\%}CI:-0.04\text{--}(-0.005))\\ & \times \text{ bodyweight in grams}\\ & \times \text{ sex}.\end{array}$

In this analysis both the moderator sex (t=2.67, P=0.011) and its interaction with bodyweight (t=−2.43, P=0.020) were statistically significant, but not bodyweight itself (t=1.11, P=0.276). Exploration of the interaction revealed that oesophageal length increased with bodyweight in females, but decreased in males (Figure 1).

Figure 1. Regression analysis of oesophageal length for bodyweight, sex, and their interaction. Male [Mars symbol] (n=11) and female [Venus symbol] (n = 30) symbols represent individual measurements. Black lines indicate regression lines of oesophageal length by bodyweight with sex as a moderator that has an interaction effect on bodyweight. Shaded area depicts the 95% confidence interval of the regression. The assumption of normality within the two sexes (W = 0.953, P = 0.208 for females and W = 0.946, P = 0.595 for males) is met.

We simulated 100,000 samples of 11 males drawn from the reference population and tested for bodyweight deviation from our sample mean, which occurred in 3.45% for a two-sided Welch-test. This indicates that our male sample is an adequate representation of the population mean. On the contrary this is not the case for our female experimental sample, which differs in 10.95% from 100,000 simulated samples of 30 females, more than 10% being different from our experimental sample.

Discussion

A stepwise approach has been suggested to meet demand for intensified research efforts to counter the ‘Barrett’s carcinoma epidemic’. (Thrift 2016) Its first step is basic research in Levrat’s (1962) model of oesophageal adenocarcinoma. (Kapoor et al. 2015) Traction forces are an important factor in oesophageal surgery (Oetzmann von Sochaczewski, Tagkalos, Lindner, Lang, Heimann, Schröder, et al. 2019) as they may affect oesophageal perfusion. (Mun et al. 2006) Consequently, traction forces are linked to a lack of oesophageal length, particularly in long-gap oesophageal atresia. (Manning 1994) Based on this finding, it has previously been demonstrated that oesophageal length in rats could be predicted by bodyweight instead of age in rats irrespective of the investigated age (Oetzmann von Sochaczewski, Baumgart, & Muensterer 2019; Oetzmann von Sochaczewski, Tagkalos, Lindner, Baumgart, et al. 2019). As rats grow throughout their life following a sigmoid curve, (Eisen 1976; Keenan et al. 1995) this result was not surprising, but concordant to previous observations in other organs. (Schoeffner et al. 1999; Bailey et al. 2004; Piao et al. 2013) Relative organ weights – either in relation to bodyweight (Schoeffner et al. 1999) or brain weight (Bailey et al. 2004; Piao et al. 2013) – did not consistently, but largely follow this trend. It was therefore surprising that this relationship between oesophageal length and organ weight did not reproduce in our sample. This result was contrary not only to our expectation, but also to the preceding exploratory investigation undertaken for our power analysis.

We conducted a simulation analysis to ensure that our sample cohort of 11 males would be an adequate representation of the reference values of bodyweight supplied by the vendor. Using these data, we generated a simulated cohort from the normal distribution that had the same mean, standard deviation, and sample size. From this simulated cohort, we sampled 100,000 samples with the same size of our experimental cohort and tested how many of the simulated cohorts would have bodyweights different from our observed experimental cohort. This difference was marginal with 3.45% of simulated male cohorts and almost equally distributed between smaller and larger bodyweights. Thus, our observed sample seems to be an adequate representation of the reference cohort and our counterintuitive result is unlikely to be influenced by animals from the extremes of the distribution. The number of deviations from the observed cohort result was much higher for females with 10.95%. However, the overwhelming majority of these deviations had a lower bodyweight than our observed cohort. An aspect, we consider less likely to exert a substantial effect on our results based on the rather small standard deviation of our observed cohort’s mean bodyweight. In contrast, the reference cohort exhibited a more than doubled standard deviation despite four times the number of animals with a comparable mean bodyweight.

As bodyweight is influenced by age, (Eisen 1976; Keenan et al. 1995) it is crucial to precisely describe the phase of life of the experimental animal and its potential equivalence to human age. For this purpose, several techniques have been described. (Sengupta 2013; Dutta & Sengupta 2016) In our case, rats had a typical age used for experiments with Levrat’s model of oesophageal adenocarcinoma, (Levrat et al. 1962) which is the eight week of life. According to Sengupta, (Sengupta 2013) rats reach sexual maturity at day of life 50 equalling a human age of 11.5 years. (Sengupta 2013). Rats in our cohort were between 56 and 58 day old and thus adolescent. Using the formula for relating adolescent age of rats to humans, in which 1 human year is comparable to 10.5 rat days, (Sengupta 2013) our cohort would be equal to a human age of around 12 years. An age that oesophageal carcinogenesis would only start in special circumstances such as post-correction of oesophageal atresia with precancerous Barrett oesophagus. (Schneider et al. 2016; Hsieh et al. 2017) Consequently, the age at which carcinogenesis is typically induced in rats in order to model oesophageal adenocarcinoma does not match those of the human disease, which mainly affects senior citizens. (Thrift 2016) This divergence of the model from the disease may however be justified by the 50 weeks that need to pass until cancerous lesions arise. (Gronnier et al. 2013)

Another potential explanation of this unexpected decrease in oesophageal length with bodyweight might be an increased fat deposition in males. However, Sprague–Dawley rats are not prone to increased fat deposition compared to other strains. (Schemmel et al. 1970) Moreover, to achieve a considerable gain in bodyfat content in Sprague-Dawleys, it is necessary to feed them a high-fat diet. (Schemmel et al. 1970) Bodyweight increase and subsequent organ growth depend on a certain amount of protein within the rat diet. (Schweisthal et al. 1982) We bought our rats as weanlings from their vendor and they have subsequently been fed the same diet irrespective of sex with a low to medium metabolisable amount of energy. Consequently, their bodyfat content should be within normal ranges observed by others using Sprague–Dawley rats of a comparable bodyweight. (Schoeffner et al. 1999) Using the estimation formula by Brown et al. (1997) although it is thought to overestimate bodyfat content in Sprague-Dawleys, (Schoeffner et al. 1999) this would derive a bodyfat content of 8-9.6% in our male cohort. Therefore, the increased bodyfat content seems to be an unlikely explanation for a decreasing oesophageal length in males, because bodyfat content was largely influenced by the type of rat chow (Schemmel et al. 1970).

Bodyweight in contrast was largely influenced by rat strain in the same study. (Schemmel et al. 1970) In Sprague–Dawley rats, bodyweight and weight gain vary up to 22% despite the same strain from the same vendor and identical experimental conditions. (Lillie et al. 1996) This report corroborated an earlier one that demonstrated a similar effect of variability on organ weights despite the exclusion of bodyweight differences. (Gur & Waner 1993) Bodyweight variability did prevent algorithmic bodyweight prediction and the authors also noted that this variability in bodyweight did not decrease with age, but persisted. (Yee et al. 1993) Variability has been suggested to be particularly present in Sprague–Dawley rats, because they expressed two different isoforms of the cardiac myosin heavy chain-α or even a mixture of them, which did not occur in other strains. (Reiser et al. 2004) A reduced variability of bodyweight has been described for F344 inbred- in comparison to Sprague–Dawley outbred-rats, (Marino 2012) but directly comparative data on this aspect is missing in the literature.

Another explanation of our results may be based on differences in sex hormone levels: Food intake and thus bodyweight increase is negatively influenced by oestrogens. (Bull et al. 1974) Testosterone, in contrast, causes an over-proprtionate growth of accessory sex organs in males starting at the onset of puberty. (Barkley 1977) Exogenous testosterone administration in the adult rat did result in a reduced bodyweight, but increased weight of accessory sex organs. (Dewan et al. 2000) Although none of these investigations assessed the contribution of sex hormones to visceral organ growth, the promoting effects of androgens and suppressing effects of oestrogens in oesophageal carcinoma have been established in rat models back in 1985 (Kobayashi 1985). Due to this effect of sex hormones, women are more likely than men to survive oesophageal cancer even in large-scale population-based studies. (Micheli et al. 2009) In-vitro studies suggest a driving role of testosterone in moderate levels on oesophageal cancer cells, whereas high levels of testosterone exerted a negative effect on cancer cell proliferation. (Palethorpe et al. 2017) This observation has been corroborated in clinical reality in a case–control study in which the highest levels of serum testosterone were associated with a decreased odds ratio of oesophageal adenocarcinoma. (Xie et al. 2020) Based on these observations, it is tempting to speculate that our observation of decreasing oesophageal length in males might be an effect of high androgen levels during puberty. (Barkley 1977) High levels of androgens suppress growth of oesophageal cancer cells, (Palethorpe et al. 2017; Xie et al. 2020) which could probably be also the case in normal oesophageal cells based on the transition of oesophageal cancer cells from normal cells. (Flejou 2005) Consequently, such suppression due to high levels of androgens would not occur in females and thus explain the different association to bodyweight than in males. Our speculation on this point could be supported by androgen-receptor expression data from normal oesophageal cells, but to our knowledge such data is missing. Nevertheless, this emphasises the importance of the specific phase of life of the experimental animal(Sengupta 2013) due its associated differences in sex hormonal levels.

Although our present study reports results that seem highly counterintuitive to preceding results and our power calculation, (Oetzmann von Sochaczewski, Tagkalos, Lindner, Baumgart, et al. 2019) we cannot add more experimental animals to the a-priori calculated number of animals as this would alter the p-curve of our experiment: Adding more animals and conducting repetitive testing would violate the assumptions of null hypothesis significance testing. It would result in an accumulation of type I error and thereby lead to unreliable results, which is therefore considered a questionable research practice in the form of P-hacking. (Simonsohn et al. 2014; Szucs 2016).

We can conclude that, contrary to previous studies, oesophageal length had a negative relationship to bodyweight in male rats. A similar relationship was absent in female Sprague–Dawley rats, although both cohorts were comparable to the reference population in bodyweight. This finding may be a relevant source of non-reproducibility in studies, in particular if only one sex is investigated. A future study will have to take not only sex, but also strain differences into account to address this moderating effect of sex in more detail. In addition, the underlying biological factors of this finding – such as hormonal differences – need to be investigated further.

Data availability

All raw data are publicly available from Zenodo (Nuber et al. 2020) and the R code for the simulation is presented in the supplementary information.

Disclosure statement

No potential conflict of interest was reported by the author(s).