The aorta in humans and African great apes, and cardiac output and metabolic levels in human evolution

Humans have a larger energy budget than great apes, allowing the combination of the metabolically expensive traits that define our life history. This budget is ultimately related to the cardiac output, the product of the blood pumped from the ventricle and the number of heart beats per minute, a measure of the blood available for the whole organism physiological activity. To show the relationship between cardiac output and energy expenditure in hominid evolution, we study a surrogate measure of cardiac output, the aortic root diameter, in humans and great apes. When compared to gorillas and chimpanzees, humans present an increased body mass adjusted aortic root diameter. We also use data from the literature to show that over the human lifespan, cardiac output and total energy expenditure follow almost identical trajectories, with a marked increase during the period of brain growth, and a plateau during most of the adult life. The limited variation of adjusted cardiac output with sex, age and physical activity supports the compensation model of energy expenditure in humans. Finally, we present a first study of cardiac output in the skeleton through the study of the aortic impression in the vertebral bodies of the spine. It is absent in great apes, and present in humans and Neanderthals, large-brained hominins with an extended life cycle. An increased adjusted cardiac output, underlying higher total energy expenditure, would have been a key process in human evolution.


SUPPLEMENTARY MATERIAL 1
Only adult individuals were studied from the echocardiographic and NMR samples. Age at linear growth or body mass completion, or near completion, was used as the criterion for adulthood. The age at attainment of adult height is variable between human populations 1 , but considering that the human samples come from modern European populations (UK and Belgium), an approximate age of 18 years for completion or near completion of adult height in both males and females was considered. For gorillas and chimpanzees, we follow the available data from the literature, bearing in mind that the age of attainment of adult dimensions is also variable between species and populations within those genera, for instance between wild and captive populations 2,3 . For gorillas, we followed Galbany et al. (2017) 4 , who indicate that in their sample of wild mountain gorillas, 98% of maximum body length, back width and arm length were reached by females at 11.7, 11.9 and 15.9 years old, and by males at 13.1, 15.7, 14.5 years old respectively. Thus, as an average from these ages, we selected those individuals older than 13 years (females), and 14 years (males) as adults.
For chimpanzees, Hamada and Udono (2002) 5 observed that the age of body length maturation does not differ by sex, and it occurs approximately at 12 years. Pusey et al. (2005) 6 data from Gombe show that growth in body mass slowed at 10 years of age for females and 13 years for males, while Machanda et al. (2015) 7 data from Uganda show that for males linear dimensions reached adult values by 10 years old while body areas reached adult values between the ages of 15-17 years. Thus, as an average from these ages, we selected those individuals older than 12 years old. In humans it has been observed that the diameter of the aortic root increases with age throughout adult life (see below), and it would be reasonable to expect an increase with age in other hominoids. But we decided not to put an upper limit in the age of the samples from the three species due to potential comparability problems for the length and equivalence of the adult period in the three species.

SUPPLEMENTARY MATERIAL 2
For the echocardiographic and magnetic resonance sample, assumptions of normality and equality of variances were assessed by density plots and the Shapiro-Wilk test and through the Brown-Forsythe test respectively (P value <0.05). These assumptions were not met by at least half of the variables for the different samples (by species, and by species and sex). In addition, the samples sizes were clearly unbalanced, with a human sample tenfold larger than the gorilla and chimpanzee samples, similar in size (Supplementary Table 1). From the above information we chose to test statistical significance with the Games-Howell test and, since this study was planned as hypothesis testing research and we wanted to avoid a type I error, we applied the Bonferroni-Holm correction 8 , reporting the adjusted P values. The number of tests for this correction was 36, since we considered nine subsamples (species, male, female) and four variables (ARD and BM, and the two exponents of BM that appropriately normalized ARD: ARD/BM 0.236 , ARD/BM 0.25 ). For exploratory purposes, Welch unpaired tests, and Kruskall-Wallis test with Dunn's post hoc tests, both with Bonferroni-Holm corrections, were also computed and compared against the results from the Games-Howell test. The results from the three tests were basically similar (results not shown). In Supplementary Table 1 we can observe the mean, SD, and statistical significance for BM, ARD, ARD/BM 0.236 and ARD/BM 0.25 between species and by sex. In Supplementary Figure 1, we can observe the mean difference (95% CI) for BM, ARD and ARD/BM 0.236 between species and by sex 9 . Supplementary Figure 1. Mean difference in body mass, aortic root, and aortic root/body mass 0.236 , between humans and gorillas, and human and chimpanzees, for the total sample (A-C), males (D-F) and females (G-I). The raw data is plotted on the upper axes; each mean difference is plotted on the lower axes as a bootstrap sampling distribution 9 . Statistics are shown in Supplementary Table 1.

Supplementary Figure 1 (cont).
Mean difference in body mass, aortic root, and aortic root/body mass 0.236 , between humans and gorillas, and human and chimpanzees, for the total sample (A-C), males (D-F) and females (G-I). The raw data is plotted on the upper axes; each mean difference is plotted on the lower axes as a bootstrap sampling distribution 9 . Statistics are shown in Supplementary Table 1. 9 SUPPLEMENTARY MATERIAL 3

Hearts from dissection
Basic data from the human, gorilla, chimpanzee, and orangutan hearts shown in Figure 1 in the main text are presented in Supplementary

SUPPLEMENTARY MATERIAL 4 Allometric scaling of echocardiographic variables and body and body size variables
In healthy modern human populations the main determinants of ARD are body size and age [10][11][12][13][14] , including samples of athletes, where the ARD is usually within normal ranges for the general population but with larger ARD in sports with a high dynamic component 15,16 . In medicine and sports science, cardiac and circulatory variables have been normalized allometrically (by dividing the variable by some measure of body size raised to a power), but more generally ratiometrically (by dividing the variable by some measure of body size), and critical reviews have been elaborated regarding the appropriate statistical method for normalization and the most appropriate body size measure to use 15,[17][18][19][20]

SUPPLEMENTARY MATERIAL 5 Aortic root diameter and cardiac structural variables in humans and chimpanzees
A literature search was undertaken to find articles containing basic anthropometric data (sex, age, height,  Table 5). In chimpanzees, LVEDD and LVESD statistically significantly predicted ARD, although R 2 values were lower than in humans. This difference could be related to the fact that the human data are mean values from 71 samples with an upper age limit (30 years) to avoid the effect of age on the aortic root diameter. In chimpanzees, no upper age limit was set, and the data points shown correspond to individuals.
These results show the association of the aortic root diameter with the left ventricle in great apes, supporting its association with stroke volume, a component of the cardiac output.  In previous studies the normalization of cardiac output and stroke volume with height and weight has been researched 29,30 , and scaling coefficients from theoretical estimations have been also used 31 . In our sample we applied and assessed these scaling coefficients and obtained and assessed sample specific allometrically and/or ratiometrically normalization of cardiac output, stroke volume and heart rate with height and weight.
For the latter task, log transformations of cardiac output, stroke volume, heart rate, height and weight, and linear regression were used to determine the different scaling exponents. Results are presented in Supplementary Table 6. As explained above for the ARD, to assess the scaling procedure, Pearson correlation coefficients were obtained between the scaled variable and the body size measure by which it was scaled; and for ratiometric scaling the Pearson correlation coefficient was compared to the ratio of the coefficients of variation of both variables (Supplementary Table 7). As can be observed, there were several statistically significant correlations between the scaled-variable and either weight or height, and for the ratiometric scaling the difference between the Pearson correlation coefficient and the ratio of the coefficients of variation was also considerable for several cases. From these results, we chose to normalize cardiac output, stroke volume and heart rate by height raised to the exponents calculated for our data gathered from the literature. Also, it has been shown that fat free mass (FFM) or lean body mass, a measure of the metabolically active tissue, is strongly related to cardiac output and stroke volume 32 , and to cardiac power output 33 , and should be the appropriate variable to scale cardiac output and stroke volume. In absence of FFM, height would be the preferred size measure to scale cardiac variables 18 . shown). We can conclude that cardiac output, stroke volume, and heart rate show a moderate or lack of decline with age in the athlete and control samples, especially when the variables are scaled to body size.

Differences in cardiac output, stroke volume and heart rate between control and athlete samples.
To further test the difference between control and athlete samples for the three variables, we restricted the age and height of the samples to make them comparable. The distribution of the samples across age and height shows that for cardiac output and stroke volume, the athlete samples are younger and taller, while for heart rate the height distribution is similar, but the athlete samples are also younger. As shown above, age explains a moderate but statistically significant percentage of the variation of cardiac output and stroke volume in the control samples. Also, since cardiac output and stroke volume have an allometric relation with height, the effect of the shortest samples would be stronger than the effect of the tallest samples. Thus, to avoid a possible bias of age and/or height, we restricted the samples to a maximum age of 40 years and a minimum height of 162.9 cm (the mean height of the shortest athlete sample was 163 cm). The mean difference between athlete and control samples for the non-scaled and scaled cardiac output, stroke volume and heart rate are shown in Supplementary Figure 3 9 and Supplementary  Figure 3. Mean difference in cardiac output (CO), stroke volume (SV), and heart rate (HR) (left column, A-C), and in the same variables scaled to height (right column, D-F), between control and athlete samples. The raw data is plotted on the left axis; the mean difference is plotted on the right axis as a bootstrap sampling distribution 9 . Statistics are shown in Supplementary Table 9.
For cardiac output, no significant differences were observed between the athlete and control samples for the non-scaled and scaled variable. For stroke volume, a significant difference was observed for the nonscaled variable (P=0.00163) with a higher stroke volume in the athlete sample, while for the scaled-variables the P value was 0.0676. For heart rate, significant differences were observed between samples (P <0.001), with higher values for the control samples. Variation across lifespan of cardiac output, total energy expenditure, organs' weight, and anthropometry A literature search was undertaken to find articles containing data on cardiac output, organ weight, height and weight, and TEE, across the life span or along the growth period. For cardiac output, results from research conducted on a local, healthy sample from Hong Kong, aged 0-60 years were selected [37][38][39][40] .
Anthropometric data from this study was available below the age of 17 years, and for this subsample, cardiac output was adjusted using the exponents calculated previously for height and body surface area 30 .
When considering the whole lifespan, since anthropometric data was not available, cardiac output adjusted to body surface area (bsa) was used (cardiac output/bsa) 37 . For organ weight, and taking into account the previously selected Hong Kong sample for cardiac output, results from a cadaver study on a nationwide Japanese sample were selected 41 . For height and weight, for illustrative purposes, and due to its completeness, data from the CDC Growth Charts were also selected 42  Basic data were extracted from the article's tables and figures. Data from figures were extracted with webplotdigitizer (https://automeris.io/WebPlotDigitizer/). In order to compare data from different variables, we followed a procedure described previously 44 . Continuous functions were fit separately to all the variables (absolute and adjusted cardiac output and TEE, organs weight, height, and weight), and different models were evaluated using the Akaike Information Criteria (AIC), for the age range 0-19 or 0-27 years. A Gompertz model was selected for brain weight growth, while for the other variables, a cubic spline function with four knots was selected, placed at age quantiles. Once the models were selected, predicted values were calculated at 0.2 years intervals for all the variables. Z-scores and derived variables (velocity) were then calculated from those predicted variables, and the comparative graphs were built.

Supplementary Figure 4.
Change from birth to adulthood of brain weight and organs weight (A), and change from birth to adulthood of brain weight, together with the velocity curve for height and weight (B). Original data were first obtained from the literature, and then a Gompertz (brain), or cubic spline functions (all the other organs), were fitted to the values. Predicted values from the functions were obtained (first derivative was also calculated for height and weight) and converted to z-scores for comparison between variables.

SUPPLEMENTARY MATERIAL 8 Aortic impression
The skeletal sample is shown by genera and sex in Supplementary For each specimen, first, the available vertebrae were articulated from C1 (first cervical) to L4 (great apes), Each photo was transferred to Adobe Photoshop (CS6) where, if needed, its position was rotated to further fit the previous criteria. Several vertebrae were asymmetrical and did not fit all these criteria simultaneously.
In these cases, the alignment of the vertebral image was carried out trying to maximize the correct alignment of the vertebral endplate. The rectangle tool was used to display a rectangle whose superior side This image was saved and transferred to ImageJ 46 , and after setting the scale, the surface areas of the anterior left and right quadrants were determined. These were the values used for the calculation of the asymmetry. The potential error associated to this procedure was tested by repeating the protocol again for vertebrae from 10 cases (n=158), with an intraclass correlation coefficient of 0 It is important to address the limitations of the assumption of a direct relationship between bilateral asymmetry of the anterior half of the vertebral body and the presence of an aortic impression, defined as "a variable flattening that may be found on the left side of the bodies of mid-thoracic vertebrae" 49 . This caution is warranted since clear cases of asymmetry of individual vertebrae can be observed in Figure 4A.
The vertebral column is a metameric structure that expresses information related to ontogeny and ageing, body mass, posture, and locomotion, breathing kinematics and other physiological functions related to thoracic and abdominal organs. The asymmetry measured in the present study captures more asymmetry than that strictly related to the aortic impression. As an additional check, each image was visually assessed to detect any asymmetry caused by a unilateral flattening on the left side of the body: no great ape vertebra presented this asymmetry. Finally, since the trajectories of asymmetry along the vertebral column of individuals is not smooth, where a complete vertebral column is lacking, absence of asymmetry and thus of an aortic impression in an isolated vertebra should be considered with caution.

SUPPLEMENTARY MATERIAL 9
Heart rate In a large national sample from the USA, the NHANES I, 50 observed that for adults, supine resting heart rate (RHR, measured by electrocardiogram) was not associated to height or weight in most subgroups (sex, age, ancestry). Age and ancestry were also not associated to RHR, but a consistently higher RHR was observed in women (average of 3 bpm) that was not explained by the measured covariates. Similar findings were obtained for resting radial pulse rate (RPR, measured manually) in the same sample 51 . In the latter study, slower RPR were observed in more active persons, with difference of 6-8 bpm for males and 1-4 bpm for females. A more recent study of the NHANES 1999-2008 data 52 , found a small by significant effect of ancestry on RPR, and a moderate change of RPR with age in adults. A mean difference of 3 bpm more in females than males was also observed in these studies (and in our database, 3.6 bpm, see Supplementary  Figure 6, the variation with age of heart rate and heart rate scaled to weight raised to the power -0.25 for these athlete/control samples is displayed, with negligible or statistically not significant associations. With relation to physical activity, the mean difference in heart rate between athlete and control samples was 9.02 bpm lower in the athlete sample (Supplementary Figure 5B), especially for those involved in endurance training.
Supplementary Figure 5. Mean difference in heart rate between males and females (A), and athlete and control samples (B). In both cases, groups are plotted on the left axes, while the mean difference is plotted on a floating axis on the right as a bootstrap sampling distribution. The raw data is plotted on the left axis; the mean difference is plotted on the right axis as a bootstrap sampling distribution 9 .
.  Table 1. The mean values from Peruvians residents at the high-altitude location of Junín (4.100 m) 54 , obtained from their Table 1, is also included. Red and blue circles represent respectively athlete and healthy control samples. Lines and shaded regions indicate least squares regressions and 95% confidence intervals for both samples.