Behavior is shaped by genes, environment, and evolutionary history in different ways. Nest architecture is an extended phenotype that results from the interaction between the behavior of animals and their environment. Nests built by ants are extended phenotypes that differ among species and among colonies within a species, but the source of these differences remains an open question. To investigate the impact of colony identity (genetics), evolutionary history (species), and the environment on nest architecture, we compared how two species of harvester ants construct their nests under different environmental conditions. For each species, we allowed workers from 4 colonies to excavate nests in environments that differed in temperature and humidity for seven days. We then created casts of each nest to compare nest structures among colonies, between species, and across environmental conditions. We found differences in nest structure among colonies of the same species and between species. Interestingly, however, environmental conditions did not have a strong influence on nest structure in either species. Our results suggest that extended phenotypes are shaped more strongly by internal factors, such as genes and evolutionary history, and are less plastic in response to the abiotic environment, like many physical and physiological phenotypes.

Animal collection and maintenance

We collected and brought into the lab harvester ant workers from four different colonies from each of the two species P. californicus and V. andrei (N=8 colonies). Pogonomyrmex californicus workers were collected from colonies along the Red Rock Road in Red Rock Canyon Park in the Santa Monica Mountains. Veromessor andrei workers were collected from colonies on the southwest mesa at the UC Sedgwick Reserve in the Santa Ynez Valley. For each species, we identified the four most active colonies in the population to ensure that we could collect sufficient ants for all the experimental trials. We recorded the location of each colony so that we could return to them throughout the study. We collected ants multiple times throughout the eight weeks of the study (June to August 2020) and began the experiments 48 hours after each collection so that ants were housed in the lab for no more than 48 hours before beginning digging trials. To collect the ants we wore latex gloves and placed 100-120 workers in a plastic container that had a wet paper towel for moisture.

In the lab, we placed ants from each of the eight colonies  in a separate plastic container with a lid and fluon along the walls to prevent ants from escaping. We supplied the workers with ad libitum water and 50% sugar water in glass tubes plugged with cotton to allow wicking of the fluid during the 48 hours before the digging trials began and the 7 days of the trial itself.

Experimental design

We allowed 50 workers from each of the eight source colonies to excavate a nest in one of four environmental treatments (Figure 1): cold/dry, cold/wet, hot/dry, hot/wet. To control for the potential impact of the number of workers on nest size, we used groups of identical size (50 workers) in each experiment. Temperature and humidity were controlled using environmental chambers (Caron, model 6045) with ‘hot’ temperature set at 29.4 °C and ‘cold’ at 19.5 °C. The air temperature set in the environmental chamber was very similar to that inside the soil buckets (Mean 1.24°C difference, standard deviation 0.99; see supplementary materials, Table S4).  The ‘wet’ condition was 80% humidity and ‘dry’ was on average 23% humidity (range: 17-29%). The ‘dry’ condition had some­ variability because of the technological limits of the environmental chambers. The two experimental temperature values and two experimental humidity values we tested represent the range of typical daytime conditions that the populations we collected ants from experience throughout the year (according to wunderground.com and Western Regional Climate Center, see supplementary materials). The ‘cold’ treatment represents average daytime highs during the coldest months (December -March) and the ‘hot’ treatment represents average daytime highs during the warmest months (July-October) at both field sites (Table S2). The ‘wet’ treatment represents the highest average humidity recorded at the two sites (Table S3) and the ‘dry’ treatment humidity was the driest conditions we could achieve in the environmental chambers.

Nest excavation trials

To allow ants to dig a nest, we placed the 50 workers in a five-gallon bucket filled with 47cm of Quikrete All-Purpose sand moistened with approximately 150 mL water. We placed the maximum amount of sand possible in the buckets (47cm) without having ants escape. We used store-bought sand to provide all ant groups with controlled identical digging conditions. In preliminary trials, we found that the amount of sand we used needed to be moistened with approximately 150 mL water to allow the ants to dig stable nest structures that can be casted and excavated. . To focus the digging of the ants at one location, and prevent them from digging multiple nest entrances, we covered the sand with a layer of hardened wax with one circular opening (7-10 mm diameter) in the center of the wax cover (Figure S2). This circular opening acted as the nest entrance. Throughout the nest excavation trials, we supplied the ants with ad libitum water and 50% sugar water in glass tubes plugged with cotton to allow wicking of the fluid. Ants were allowed to excavate a nest for seven days with a 12 hour day/night light regime. This experimental set up is similar to the one used in other studies of harvester ant nest excavation (Kwapich et al 2018). To create a cast of the excavated nest, at the end of the week of excavation, we evacuated workers from the nest by blowing air into the nest entrance. When most ants were evacuated, we poured melted wax down the nest entrance opening and after the wax hardened, we carefully dug out the cast for further quantification (Figure S2).

Cast quantification

To quantify the nest structures that the ants excavated we measured features of the tunnels, chambers, and of the entire nest, as described below. We then compared these features among colonies and between species, temperature, and humidity treatments.

Tunnel measurements

Tunnels are an important feature of the nest because they facilitate movement of ants and materials. To determine the capacity of each tunnel to transport material and ants we measured the length and circumference of each tunnel segment. We defined a tunnel segment as a tunnel connecting two nest features (chambers, junctions, nest entrance, or ends of cast). We laid a string along the tunnel segment and measured its length with a ruler to the nearest mm. To measure the circumference of each tunnel segment, we wrapped a string around the tunnel segment at each of its terminals and at its center, and then recorded the length of the string with a ruler to the nearest mm. We recorded the average of these three measurements as the tunnel circumference.

Chamber measurements

A chamber was defined as any section of the nest with a globular, rather than cylindrical, shape. To quantify the capacity of each chamber to house ants and other materials, we measured chamber circumference by wrapping a string around the chamber’s widest point and recording the length of the string with a ruler to the nearest mm. 

Global nest measurements

To quantify the global structure of each nest we used network analysis to quantify nest shape, combined nest-level tunnel and chamber measurements, and measured nest volume. All measures were aggregated using a PCA, as detailed below. We depicted each nest as a network of nodes representing chambers, junctions, and ends of the cast, connected by edges representing tunnels, as in (Buhl et al. 2004; Perna et al. 2008; Viana et al. 2013; Gautrais et al. 2014; Pinter-Wollman 2015) (see supplementary materials for examples). To characterize nest connectivity, which can impact a colony’s speed of recruitment to food (Pinter-Wollman 2015), we computed the average degree, which is the mean number of unique nodes that each node is connected to. To determine how well different nest elements are connected with one another, we calculated network density, which is the number of observed edges divided by the number of possible edges. To determine the distances between different nest elements, which can impact how quickly resources and information flow through the nest, we calculated average path length, which is the mean number of edges that connect all node pairs. To further quantify the potential for flow through the nest, we noted the presence or absence of cycles. A cycle is defined as a path along a network that returns to its starting node after passing through at least one other node (Balakrishan 2011; Bender and Williamson 2010; Bollobás 1998; Gross and Yellen 2005; Newman 2018). Larger colonies have more cycles (Buhl et al. 2004) and colonies in nests with more cycles have faster recruitment to food (Pinter-Wollman 2015).

We further calculated measures of nest size: total number of chambers, average circumference of all tunnel segments, total length of all tunnel segments, and nest volume. To obtain volumes we scanned each nest fragment individually using a professional structured light scanner (Artec Space Spider) with an accuracy of 100-200 microns. To help the scanner detect the otherwise semi-translucent wax, we painted exposed sections of the wax in dark blue. We then manually aligned each digitized fragment with other fragments from the same nest using Artec Studio. Once each nest was aligned, we loaded models into Meshmixer to turn the disparate fragments into a single STL file. We imported the STL files into Meshlab, which provides volumetric measurement under the Mesh option.

Data analysis:

To aggregate the various global nest measures we used a Principal Component Analysis (PCA). In the PCA we included all measures listed in the “Global nest measurements” section above, except for the presence or absence of cycles, which is a binary variable and was thus analyzed separately. Each measure was scaled by subtracting its mean and dividing by its standard deviation before running the PCA. The first principal component explained 61.9% of the variance (Table S2) therefore we used it as a single measure of ‘nest size’ (Table S2, Figure S3).

To determine the effect of species, colonies, and environmental conditions on the excavated nest tunnels, chambers, and global nest features, we conducted four separate ANOVAs and one logistic regression (for cycles in the nest). In each statistical test, the measures of tunnels (circumference or length), chambers (circumference), or global nest (nest size or presence/absence of cycles) were the dependent variables. We log-transformed the tunnel measures to meet the ANOVA assumptions; chambers and nest size measures met the ANOVA assumptions. All statistical models included species, colony ID nested within species, temperature, and humidity as independent variables. All statistical analysis was conducted in R version 4.0.3 (R Core Team 2020). The ANOVAs were conducted using the Anova() function from the package ‘car’ (Fox and Weisberg 2018). Data and analysis code are available in the supplementary materials.

Nest excavation trials

To allow ants to dig a nest, we placed the 50 workers in a five-gallon bucket filled with 47cm of Quikrete All-Purpose sand moistened with approximately 150 mL water. We placed the maximum amount of sand possible in the buckets (47cm) without having ants escape. We used store-bought sand to provide all ant groups with controlled identical digging conditions. In preliminary trials, we found that the amount of sand we used needed to be moistened with approximately 150 mL water to allow the ants to dig stable nest structures that can be casted and excavated. To focus the digging of the ants at one location, and prevent them from digging multiple nest entrances, we covered the sand with a layer of hardened wax with one circular opening (7-10 mm diameter) in the center of the wax cover (Figure S2). This circular opening acted as the nest entrance. Throughout the nest excavation trials, we supplied the ants with ad libitum water and 50% sugar water in glass tubes plugged with cotton to allow wicking of the fluid. Ants were allowed to excavate a nest for seven days with a 12 hour day/night light regime. This experimental set up is similar to the one used in other studies of harvester ant nest excavation (Kwapich et al 2018). To create a cast of the excavated nest, at the end of the week of excavation, we evacuated workers from the nest by blowing air into the nest entrance. When most ants were evacuated, we poured melted wax down the nest entrance opening and after the wax hardened, we carefully dug out the cast for further quantification (Figure S2).

Cast quantification

To quantify the nest structures that the ants excavated we measured features of the tunnels, chambers, and of the entire nest, as described below. We then compared these features among colonies and between species, temperature, and humidity treatments.

Tunnel measurements

Tunnels are an important feature of the nest because they facilitate movement of ants and materials. To determine the capacity of each tunnel to transport material and ants we measured the length and circumference of each tunnel segment. We defined a tunnel segment as a tunnel connecting two nest features (chambers, junctions, nest entrance, or ends of cast). We laid a string along the tunnel segment and measured its length with a ruler to the nearest mm. To measure the circumference of each tunnel segment, we wrapped a string around the tunnel segment at each of its terminals and at its center, and then recorded the length of the string with a ruler to the nearest mm. We recorded the average of these three measurements as the tunnel circumference.

Chamber measurements

A chamber was defined as any section of the nest with a globular, rather than cylindrical, shape. To quantify the capacity of each chamber to house ants and other materials, we measured chamber circumference by wrapping a string around the chamber’s widest point and recording the length of the string with a ruler to the nearest mm. 

Global nest measurements

To quantify the global structure of each nest we used network analysis to quantify nest shape, combined nest-level tunnel and chamber measurements, and measured nest volume. All measures were aggregated using a PCA, as detailed below. We depicted each nest as a network of nodes representing chambers, junctions, and ends of the cast, connected by edges representing tunnels, as in (Buhl et al. 2004; Perna et al. 2008; Viana et al. 2013; Gautrais et al. 2014; Pinter-Wollman 2015) (see supplementary materials for examples). To characterize nest connectivity, which can impact a colony’s speed of recruitment to food (Pinter-Wollman 2015), we computed the average degree, which is the mean number of unique nodes that each node is connected to. To determine how well different nest elements are connected with one another, we calculated network density, which is the number of observed edges divided by the number of possible edges. To determine the distances between different nest elements, which can impact how quickly resources and information flow through the nest, we calculated average path length, which is the mean number of edges that connect all node pairs. To further quantify the potential for flow through the nest, we noted the presence or absence of cycles. A cycle is defined as a path along a network that returns to its starting node after passing through at least one other node (Balakrishan 2011; Bender and Williamson 2010; Bollobás 1998; Gross and Yellen 2005; Newman 2018). Larger colonies have more cycles (Buhl et al. 2004) and colonies in nests with more cycles have faster recruitment to food (Pinter-Wollman 2015).

We further calculated measures of nest size: total number of chambers, average circumference of all tunnel segments, total length of all tunnel segments, and nest volume. To obtain volumes we scanned each nest fragment individually using a professional structured light scanner (Artec Space Spider) with an accuracy of 100-200 microns. To help the scanner detect the otherwise semi-translucent wax, we painted exposed sections of the wax in dark blue. We then manually aligned each digitized fragment with other fragments from the same nest using Artec Studio. Once each nest was aligned, we loaded models into Meshmixer to turn the disparate fragments into a single STL file. We imported the STL files into Meshlab, which provides volumetric measurement under the Mesh option.

Data analysis:

To aggregate the various global nest measures we used a Principal Component Analysis (PCA). In the PCA we included all measures listed in the “Global nest measurements” section above, except for the presence or absence of cycles, which is a binary variable and was thus analyzed separately. Each measure was scaled by subtracting its mean and dividing by its standard deviation before running the PCA. The first principal component explained 61.9% of the variance (Table S2) therefore we used it as a single measure of ‘nest size’ (Table S2, Figure S3).

To determine the effect of species, colonies, and environmental conditions on the excavated nest tunnels, chambers, and global nest features, we conducted four separate ANOVAs and one logistic regression (for cycles in the nest). In each statistical test, the measures of tunnels (circumference or length), chambers (circumference), or global nest (nest size or presence/absence of cycles) were the dependent variables. We log-transformed the tunnel measures to meet the ANOVA assumptions; chambers and nest size measures met the ANOVA assumptions. All statistical models included species, colony ID nested within species, temperature, and humidity as independent variables.. All statistical analysis was conducted in R version 4.0.3 (R Core Team 2020). The ANOVAs were conducted using the Anova() function from the package ‘car’ (Fox and Weisberg 2018). Data and analysis code are available in the supplementary materials.

README file includes instructions for using the dataset: code needed for analysis with documentation is included, as well as an Excel file containing the dataset and a ReadMe sheet explaining how to interpret the datasheets.