Timed-pregnant, wild-type CD-1 mice were purchased from Charles River Laboratories, Inc. (Wilmington, MA). They arrived at 15–17 days of gestation and were housed in the UCSB animal facility in individual cages on a 12:12 light-dark cycle (lights on at 07:00, off at 19:00). Mice were checked every morning (before 12:00) and the delivery day was considered to be postnatal day (PD) 0. The pups were kept with the dam until used. Pups from six litters were analyzed. All litters were normal size (12–14 pups). All experiments were approved by the UCSB Institutional Animal Care and Use Committee.

A total of 63 pups from 6 litters were used in the study. At postnatal day 14 (5 litters) or 15 (1 litter), individual pups were removed from the cage, weighed, and immediately decapitated. The trunk blood was collected into a 1.5 mL microcentrifuge tube containing 25 μL of 5% Na₂EDTA dissolved in water, as previously described 1. The tube was immediately pulse-vortexed at low speed and kept at room temperature until the dissections of the brain and gut were completed. Following the dissections, the blood sample was visually examined for blood clots, the volume of the collected blood was measured, and the sample was stored in 20 μL aliquots at -75°C. The brain was dissected out of the skull with a fine rongeur, immediately weighed in phosphate-buffered saline (0.1 M PBS, pH 7.2) on a precision balance, immersion-fixed overnight in 4% paraformaldehyde in PBS at 4°C, allowed to sink in 30% sucrose in PBS at 4°C, and processed as described below. The gastrointestinal tract (the stomach through the rectum) was dissected out of the trunk and allowed to relax in PBS at 4°C for 10–13 days before photographed. The distal tail was stored at -75°C for sex-genotyping. All analyses were done blindly with respect to the values of the other variables in the same individual mice.

Following cryoprotection, 17 brains (from 2 litters) were embedded in 20% gelatin (type A; 275 bloom), immersed in formalin with 20% sucrose for 3 hours at room temperature, and sectioned coronally on a freezing microtome at 50 μm thickness into PBS in 96-well trays. Every lost section was recorded and the percentage of missing sections was not allowed to exceed 9%. The sections were immediately mounted onto glass slides coated with 0.5% gelatin and 0.05% chromium potassium sulfate, allowed to air-dry, Nissl-stained with 0.25% thionin, dehydrated in a graded series of ethanols, cleared in Xylenes, and coverslipped with Permount. Since fixation and embedding may alter tissue volume 5455, care was taken to ensure all brains and sections were exposed to the reagents for the same length of time.

Images of serial sections were captured with a Zeiss Axio Imager Z1 equipped with a color digital camera (AxioCam HRC Rev. 2) using a 1× objective (Fig. 1A), imported into the Reconstruct software (version 1.0.9.6; http://synapses.clm.utexas.edu/tools/index.stm), and the outline of the left hippocampal complex was traced from the most rostral section containing the pyramidal cells to the most caudal section containing the dental gyrus (Figs. 39–64 of 56). In this rostro-caudal block, the traced regions included the hippocampus proper, the fimbria, the dentate gyrus, the presubiculum, and the subiculum. If a section was missing, the thickness of each of the two neighboring sections was digitally increased by 50%. The quality of tracing was assessed by rotating the 3D-reconstruction and visually inspecting it for discontinuities (Fig. 1B). No digital smoothing was used.

The concentration of tryptophan and 5-HT in the blood samples was measured by high-performance liquid chromatography (HPLC) as previously described 1. Blood samples were prepared by addition of 100 μL of 25% ascorbic acid, 100 μL of 5% sodium metabisulfite, 10 μL of 10 ng/μL N-methylserotonin (NMS, internal standard) to frozen whole blood samples. After thawing and mixing, 75 μl of 3.4 M perchloric acid was added, the samples were kept on ice for 10 minutes, centrifuged, and the supernate was stored at -80°C until analyzed. Analysis was performed by HPLC with fluorometric detection: a 25 × .46 Altex Ultrasphere column heated to 40°C was eluted with a mobile phase consisting of 70% 0.1 M NaH₂PO₄ (pH 4.7) containing 150 mg/L octyl sulfate, 20% methanol. Compounds were detected with a Shimadzu RF 10-AXL fluorometer, with excitation and emission wavelengths of 285 and 345 nm, respectively, quantitated by comparing peak heights ratios (analyte/NMS) to those observed for standards, and concentrations expressed as ng per mL. The neurochemicals were determined with typical intra- and inter-assay coefficients of variation of less than 5% and 10%, respectively. In the final calculations, the dilution of the collected blood in the Na₂EDTA solution was factored out; therefore, the reported values represent tryptophan and 5-HT concentrations in the undiluted whole blood.

Since sexing of young mice based on the appearance of their external genitalia is difficult, the anogenital distance can be used as a more reliable measure. In the present study, several organs had to be rapidly extracted destroying the body in the process. Therefore, in order to avoid ambiguity and human errors, mice were sexed by genotyping. This approach also allows permanent storage of the extracted DNA which can be used for future analyses. Genomic DNA was isolated with the DNA Isolation Kit (Lamda, St. Louis, MO). Briefly, 1 cm of the tail was digested in a lysis buffer at 56°C overnight, the DNA was precipitated with ethanol, dissolved in 150 μL of Tris-EDTA buffer (TE, pH 8.0), and stored at -20°C. The sry region located on the Y chromosome was amplified by polymerase chain reaction (PCR) using the HotStarTaq Master Mix Kit (Qiagen, Valencia, CA). The sry primers were 5'-AACAACTGGGCTTTGCACATTG-3' (forward) and 5'-GTTTATCAGGGTTTCTCTCTAGC-3' (reverse), and the control (myog) primers 57 were 5'-TTACGTCCATCGTGGACAGC-3' (forward), 5'-TGGGCTGGGTGTTAGTCTTA-3' (reverse). The PCR reaction mixture (50 μL) contained 25 μL of the Master Mix, 1 μL (0.5 μM) of each primer, 0.5 μL of the DNA sample, and 20.5 μL water. The amplification was performed in a PxE thermal cycler (ThermoFisher Scientific) with the following conditions: initial activation at 95°C for 15 min, 33 cycles of amplification (94°C for 40 sec; 60°C for 40 sec; 72°C for 1 min), and final extension at 72°C for 9 min. The PCR products were run at 120 V on 1.5% agarose gel in Tris-acetate-EDTA buffer (TAE) containing 5% ethidium bromide and digitally imaged with a DigiDoc-It UV transillumination system (UVP, Upland, CA). The amplicon sizes were 146 bp/166 bp (sry, doublet) and 245 bp (myog) (Fig. 1C).

Statistical analysis was carried out in SPSS 16.0.2 (SPSS Inc., Chicago, IL). Conceptually, the analysis was based on the general linear model (GLM) that considers correlation, ANOVA and ANCOVA to be special cases of linear regression 58. Residuals were tested for normality with the Shapiro-Wilk test (with the level of significance of p < 0.05) and for independence with the Durbin-Watson test (DW; independence was assumed if the DW value was in the 1–3 range).

Unless otherwise noted, the forced-entry method was used in linear regression. A non-linear relationship between two variables was considered to be linear "in the parameters" 58, which allowed using standard GLM methods. In backward (stepwise) regression, Fisher's F critical value for variable removal was set at the standard significance level of 0.1. Pearson's correlation between two variables was used if the residuals of the linear regression between these two variables were normally distributed (otherwise, the Mann-Whitney test was used). The emphasis on the normality of residuals rather than of variables themselves follows the standard approach in GLM models. Specifically, it allows treating Student's independent t-test as a special case of Pearson's correlation between a continuous variable and a dichotomous variable (e.g., sex; Table 1).

The collected data (Additional file 1) for each of the variables were screened for quality blindly with respect to the values of the other variables. As described below, objective criteria were used to eliminate some cases from further analysis. Since in all reconstructed cases the percentage of missing sections in the hippocampal complex did not exceed 9%, all of the reconstructions were used. Cases with guts clearly damaged during dissection (e.g., split into more than two pieces) were not used in further analyses that included gut length as a variable. We investigated whether the presence of blood clots in some blood samples affected their tryptophan or 5-HT concentration. The mean tryptophan concentrations were not significantly different in the samples with and without blood clots (t(61) = 0.47, p = 0.644), but the mean 5-HT concentration was significantly lower in the former group (t(61) = 2.44, p = 0.018). Therefore, cases with blood clots (N = 5) were disregarded in all further analyses that included tryptophan or 5-HT as variables.

The collected blood volume varies from case to case and is strongly affected by the constriction of major blood vessels, the angle between the trunk and the collecting tube, and other factors. We therefore asked whether the amount of the collected blood correlated with the measured tryptophan and 5-HT concentrations (if all other parameters are equal and the measurements are accurate, no significant effect should be observed). The correlation between blood sample volume and blood 5-HT concentration was significant (r = -0.385, p = 0.003). However, the collected blood volume is also influenced by body size. Therefore, we calculated the partial correlation between blood sample volume and blood 5-HT concentration controlling for body mass and found that this correlation became non-significant (r = -0.085, p = 0.532). In contrast, the partial correlation between body mass and blood 5-HT concentration controlling for blood sample volume was highly significant (r = -0.495, p = 9·10^-5). Therefore, all of these cases (N = 58) were used in further analyses that included tryptophan or 5-HT as variables. The relationship between body mass and blood 5-HT concentration was investigated further.

ANOVA showed significant litter effects on body mass (F_5,57 = 40.3, p < 10^-6), brain mass (F_5,57 = 15.5, p < 10^-6), hippocampal volume (F_1,15 = 7.9, p = 0.013), blood tryptophan levels (F_5,52 = 10.8, p < 10^-6), blood 5-HT levels (F_5,52 = 16.0, p < 10^-6), and a near-significant litter effect on gut length (F_5,27 = 2.54, p = 0.052).

After dissection, the length of the gut increases over the course of several days until it reaches a plateau (Fig. 2A, B). We followed the relaxation of 5 guts in PBS at 4°C and plotted the rate of their length increase as a function of their current length (Fig. 2C). Linear regression revealed a strong relationship between these variables (R² = 0.811, F_1,28 = 120.5, p < 10^-6); the residuals were normally distributed (p = 0.18) and independent (DW = 2.45). This linear relationship suggests that gut relaxation can be modeled with the differential equation

dL/dt = k (γL₀ - L),

where dL/dt is the current length increase rate as a function of the current gut length (L), L₀ is the initial gut length on the dissection day, γL₀ is the final gut length (where γ > 1 is a constant independent of L₀), and k is the length increase rate constant. Based on Eq. (1), we reanalyzed the gut relaxation data with linear regression using dL/dt as the dependent variable and L₀ and L as the two predictor variables. Since according to Eq. (1) dL/dt = 0 when L₀ = L = 0, the intercept of the regression line was fixed at the axis origin. The regression revealed that Eq. (1) well describes the relaxation process, since L₀ and L explained virtually all variance in the length increase rate (R² = 0.947, F_2,28 = 250.8, p < 10^-6). The residuals were normally distributed (p = .993) and independent (DW = 1.72). The regression yielded k = 0.49 day^-1 and γ = 1.54. Since the solution of Eq. (1) is

L(t) = γL₀ + L₀e^-kt (1 - γ),

plugging the obtained numerical k and γ values into Eq. (2) shows that, at 4°C, the gut reaches 95% of its final length (1.54 L₀) in 4 days and 99% in 7 days. In the present study, we allowed guts to relax for 10–13 days; however, shorter relaxation times (6–8 days) should be sufficient to obtain accurate measurements.

In the initial analysis, the residuals of the linear regression between each of the variable pairs were tested for normality. Of 7 pairs that failed the normality test (irrespective of which of the two variables was the non-dichotomous dependent variable), 5 included at least one of the "ratio variables" ("brain mass/body mass" or "blood 5-HT concentration/blood tryptophan concentration"). This result is expected since the ratio of two normally distributed variables is not normally distributed and can even be bimodal 59. Based on these results, the Mann-Whitney test was used to compare the "ratio variables" in the two sex groups; for all other variables, Student's t-test was used. Likewise, Spearman's correlation coefficient was used to correlate two variables when either of them was a "ratio variable"; otherwise, Pearson's correlation was used. The results of these calculations are given in Table 1.

In the interpretation of the results, the Bonferroni correction for multiple tests was used. The type I error (α = 0.05) was divided by the total number of tests (36; Table 1) to yield the significance level of α* = 0.0013. As expected, highly significant positive correlations were found between body mass and brain mass (p < 10^-6) and between body mass and gut length (p = 10^-5). Also, a highly significant negative correlation was found between body mass and the brain/body mass ratio (p < 10^-6), possibly due to allometric scaling of brain size.

The cross-correlational analysis revealed a strong negative correlation between body mass and blood 5-HT levels (p = 10^-6). Also, significant correlations were found between brain mass and blood 5-HT levels, and between the brain/body mass ratio and blood 5-HT levels (p = 0.001 and 0.0002, respectively). Since body mass, brain mass and the brain/body mass ratio cross-correlate, we next tested which of the three variables or their linear combinations could best predict blood 5-HT levels by using stepwise (backward) linear regression. It revealed that body mass, alone, could best predict blood 5-HT levels (Fig. 3A) and that inclusion of the other two variables did not statistically improve this prediction. This model yielded R² = 0.352 (F_1,56 = 30.5; p = 10^-6; adjusted R² = 0.341); the residuals were normally distributed (p = 0.21) and independent (DW = 1.15). Since two males had unusually high blood 5-HT concentration (over 3500 ng/mL), we next investigated the origin of these outliers and their influence on the regression model. No blood clots were observed in these samples (all cases with blood clots had been eliminated prior to the analysis) and the tryptophan concentration measured in these same samples was in the normal range (Fig. 3B). Since Cook's distances (Cook's D) can be used to estimate the influence of individual cases on a regression model, we calculated this statistic for all regression points and found the largest Cook's D to be 0.66, below the critical value of 1. While the two cases could not be eliminated based on any of these objective criteria, we next excluded these two cases and recalculated the regression model. After this adjustment, blood 5-HT concentration was still best predicted by body mass (R² = 0.210; F_1,54 = 14.3; p = 0.00039; adjusted R² = 0.195).

All pups with body mass less than 5 g came from the same litter. Their tissues were collected at PD14, the same developmental time at which the other four litters were processed. The litter was normal size (14 pups; another litter had the same number of pups) and the pups showed no gross anatomical abnormalities. These pups were used in all the other analyses (including the reconstruction of the hippocampal complex) and did not appear to be outliers there (Fig. 3C, D). The correlation between body mass and blood 5-HT remained significant if only the litter means were considered (Fig. 4A).

Next, we analyzed the strong correlation between gut length and the brain/body mass ratio (p = 10^-6). Since the residuals of the linear regression between these two variables were not distributed normally (p = 0.03), we modeled their relationship with a quadratic polynomial (Fig. 3C). The regression yielded R² = 0.621 (F_2,30 = 24.6, p < 10^-6; adjusted R² = 0.596); the residuals were normally distributed (p = 0.26) and independent (DW = 1.16). The constant at the quadratic term was significantly different from zero (t(30) = 3.36, p = 0.002), further supporting the non-linearity of the relationship. The correlation between gut length and the brain/body mass ratio remained significant if only the litter means were considered (Fig. 4B).

The correlation between blood tryptophan concentration and hippocampal complex volume did not meet the significance criterion after the Bonferroni correction. However, it was highly significant considered separately (Fig. 3D). The linear regression yielded R² = 0.407 (F_1,14 = 9.63, p = 0.008; adjusted R² = 0.365); the residuals were normally distributed (p = 0.58) and independent (DW = 1.40).