Journal of Mammalian Evolution

, Volume 19, Issue 2, pp 135–153

From Desert to Rainforest: Phenotypic Variation in Functionally Important Traits of Bushy-Tailed Woodrats (Neotoma cinerea) Across Two Climatic Extremes

Authors

    • Oregon State University, Environmental Sciences
    • Museum of Vertebrate ZoologyUniversity of California, Berkeley
  • Clinton W. Epps
    • Department of Fisheries and WildlifeOregon State University
Original Paper

DOI: 10.1007/s10914-012-9187-0

Cite this article as:
Cordero, G.A. & Epps, C.W. J Mammal Evol (2012) 19: 135. doi:10.1007/s10914-012-9187-0

Abstract

Changes in body size inversely related to ambient temperatures have been described in woodrats (Neotoma) over time scales ranging from decades to millennia. However, climate-mediated variation in other traits has not been evaluated, and the effects of precipitation have been overlooked. We assessed variation in skull morphology among bushy-tailed woodrats (Neotoma cinerea) over two sampling transects spanning coastal rainforest and interior desert environments to determine whether skull morphology varied with climate. We also tested whether previously described size-temperature relationships could be generalized to our study populations. In both transects, linear measurements of functionally significant traits differed between coastal and interior populations. Geometric morphometric analyses of shape confirmed some of those differences and revealed additional patterns of skull variation. Variation in some linear measurements, including body size, was predicted by climate. However, body and skull size, as well as measurements of skull components, displayed varying responses. Although longitudinal patterns of body size variation supported Bergmann’s rule, skull size variation was only weakly associated with climate. The strongest phenotypic responses to climate were those of auditory, dental, and palatal skull traits. Altogether, our findings suggest that geographic variation in temperature and precipitation mediated selective heterogeneity and plasticity in skull traits associated with food processing and sensory organs in N. cinerea. This was consistent with our expectation of resource-dependent phenotypic variation among populations in environments with highly contrasting climatic regimes.

Keywords

WoodratsMorphometricsLocal adaptationNeotomaSkullPlasticity

Introduction

Intraspecific variation in phenotypic traits is driven by several factors, including local adaptation to climatic gradients (Mayr 1963; Endler 1977; Russell and Bauer 2005) and adaptive plasticity as a response to different environments during development (Travis 1994; Gotthard and Nylin 1995; Parichy 2005). Packrats or woodrats (Neotoma) have recently emerged as a mammalian model for studying these processes, particularly with respect to climate change in western North America (Matocq 2009). Woodrats build shelters that fossilize, thereby enabling the reconstruction of temperature trends dated as far back as 40,000 years (Jackson et al. 2005). This, in turn, has facilitated the description of some of the most striking patterns of vertebrate phenotypic responses to climate change. Body size in the bushy-tailed woodrat (Neotoma cinerea) has varied inversely with ambient temperature over a time span of 25,000 years (Smith et al. 1995). Remarkably, similar patterns have been detected on intervals of less than a decade in other woodrats (Smith et al. 1998).

A series of field, laboratory, and museum-based studies first established that woodrat body size variation follows Bergmann’s rule, presumably as an adaptive thermoregulatory strategy (Hooper 1940; Lee 1963; Brown 1968; Brown and Lee 1969). Bergmann’s rule predicts that natural selection favors a larger body in cold environments where temperatures are frequently lower than that of a physiological threshold, and the opposite holds true where it is warmer. This relationship has been supported within and among species of the genus Neotoma (reviewed in Smith et al. 2009). However, woodrat size evolution also may be driven by other climatic or ecological factors. Paleontological evidence suggests that moisture or primary productivity may be correlated with woodrat size variation (Lyman and O’Brien 2005). Also, because intestinal tract length is positively correlated with body size and digestive efficiency, there may be selection for larger body size in environments where resource availability and thus food consumption rates are high (Smith 1992, 1995). Alternatively, in some mammals, larger body size is the result of increased developmental growth in moist environments (where food availability is higher) relative to arid ones (Patton and Brylski 1987).

The best-supported hypotheses concerning the drivers of woodrat microevolutionary responses to climate change have considered ambient temperature effects (Smith et al. 1995; Smith and Betancourt 1998). These studies were based primarily on single-trait measurements of body size inferred using fecal pellets as proxies, but direct measurements were taken for some modern populations. Other studies used limited fossil evidence of mandibular elements to infer body size changes in populations and recommended further research on the effects of precipitation on modern woodrat populations (Lyman and O’Brien 2005). Although precipitation has been overlooked in many studies of phenotypic responses to climate in extinct and modern mammalian fauna (Blois and Hadly 2009), several studies have demonstrated its effect on rodent morphology (Yom-Tov and Geffen 2006; Pergams and Lawler 2009).

Examining relationships between climate and phenotypic variation is fundamental to the study of morphological adaptation in mammals (Barnosky 2005). Such research requires a two-fold approach (Gienapp et al. 2008). First, phenotypic variation and its environmental drivers need to be identified. Secondly, whether such variation is driven by natural selection or environmentally induced plasticity must be determined. Here, we address the first part of that approach by examining the effects of precipitation and temperature on phenotypic variation of modern (1912–2007) populations of the bushy-tailed woodrat (Neotoma cinerea) distributed across moist coastal and arid interior environments. Assuming that climate determines the resources (i.e., food) available to an organism and these, in turn, affect survival (fitness), then we would expect phenotypic variation in traits that interact with resources to be associated with climate. Because skull phenotypes in different environments are expected to vary in a resource-dependent manner (Hanken and Hall 1993; Skulason and Smith 1995), we hypothesized that variation in functionally significant components of the N. cinerea skull would be predicted by climate. We also considered whether previously described size-temperature relationships could be generalized to woodrats within our study area, and discussed such patterns in relation to those that we identified in skull components.

Materials and Methods

Study Species

The widely distributed N. cinerea is a large nocturnal and semi-arboreal species, found as far south as New Mexico and north to the Canadian Yukon territory (Hall 1981). Several skull polymorphisms, possibly associated with climatic variation, have been identified across its geographic distribution (Hooper 1940; Verts and Carraway 1998). Additionally, field and laboratory studies revealed strong correlations among behaviors, body size, and thermoregulatory ability in coastal, desert, and montane populations (Brown 1968; Brown and Lee 1969). Neotoma cinerea is strictly herbivorous and feeds mostly on foliage, from which it may acquire most of the water that it consumes (reviewed in Escherich 1981). Dietary preferences vary geographically or according to available resources (Escherich 1981; Smith 1997; Verts and Carraway 1998). Males are larger than females (Escherich 1981).

Study Area

We studied four putative N. cinerea subspecies distributed across 363,965 km2 within the Pacific Northwest region of the United States (Fig. 1). Because of ecological differences (described below), we divided this area into two discrete, but adjacent, transects across the Cascade Mountains (CSD; N 42°5 to 46°2) and Klamath Mountains (KM; N 42°5 to 39°) (Fig. 2). The transitions between moist coastal and arid interior climatic zones in this region are the most drastic temperature and precipitation clines in North America (Price 1978; Taylor 1999).
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-012-9187-0/MediaObjects/10914_2012_9187_Fig1_HTML.gif
Fig. 1

Localities for N. cinerea specimens examined in this study, mapped according to the subspecies distributions in Hall (1981). Specimens were referenced from the following collections: Los Angeles County Museum of Natural History (LACM), University of California, Berkeley Museum of Vertebrate Zoology (MVZ), Smithsonian Institution National Museum of Natural History (NMNH), Oregon State University Fisheries and Wildlife (OSUFW), University of California, Los Angeles (UCLA), and the University of Washington Burke Museum (UWBM)

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-012-9187-0/MediaObjects/10914_2012_9187_Fig2_HTML.gif
Fig. 2

Average vegetation growth (2005) map of the Pacific Northwest region of the United States. The Cascades (CSD) and Klamath (KM) transects, as well as their characteristic geological features are labeled. Areas in Green represent high levels of vegetation growth (i.e., near the coast), lighter tones denote a decrease in growth. Areas in yellow to brown represent low levels vegetation growth (i.e., Great Basin Desert). Map source: United States Geological Survey (http://seamless.usgs.gov/)

Coastal Zones

The coastal zones include the narrow Pacific Ocean coastal fog zone (W −124°4 to −123°), western slopes of the Cascade Mountains, and most of the Klamath Mountains (W −123° to −121.5°). Ecosystems include coniferous temperate rainforests as some areas receive the highest precipitation levels in North America. Means for annual precipitation range from 510 to 2,140 mm with maximum annual totals of 3,170–5,080 mm in some places (Anderson et al. 1998). The KM coastal zone can be differentiated from the CSD coastal zone according to forest community composition. Mixed-pine-mixed fir forests characterize the KM coastal zone, whereas the CSD coastal zone features mixed-fir-hemlock forests (Anderson et al. 1998).

Interior Zones

The interior zones are due east of longitude W −121.5°. Here, there is a rapid turnover from temperate coniferous rainforest to dry coniferous forest communities. Dry forests then rapidly transition into high elevation desert shrub steppe. The KM interior zone is entirely characterized by Great Basin Desert ecosystems (Fig. 2), where temperatures may vary by 16–17°C during a daily cycle (Walsberg 2000), and mean annual precipitation is as low as 225 mm (Anderson et al. 1998). In contrast, although the CSD interior zone includes some Great Basin Desert habitat, it also includes relatively moist mountainous regions interspersed with desert shrub steppe.

Phenotypic Variation

Data Acquisition

I. Linear morphometrics

We assessed variation in N. cinerea skulls across the coastal and interior zones of the CSD and KM transects using 13 characters (Fig. 3), measured to the nearest 0.01 mm, in 281 voucher museum specimens (Appendix I; Table 5) from 158 localities (Fig. 1). All specimens were categorized as adults based on whether the third molar was fully erupted. To reduce error, all of approximately 3,640 measurements (some characters were unavailable in some specimens) were recorded by GAC. We also recorded data for total body length (TBL) from museum tags of 126 CSD specimens.
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-012-9187-0/MediaObjects/10914_2012_9187_Fig3_HTML.gif
Fig. 3

Dorsal (top), ventral (middle), and lateral (bottom) views of the N. cinerea skull. The characters selected for measurement are abbreviated as follows: bullar breadth (BB); braincase depth (BCD); braincase length (BCL); basilar length (BL); interorbital breadth (IB); incisive foramen length (IFL); nasal length (NL); palatal bridge length (PBL); rostrum breadth (RB); rostrum depth (RD); upper tooth row length (UTL); zygomatic breadth (ZB). Illustration credit: Sarah Yarwood

II. Geometric morphometrics

Our primary method for studying the morphology of woodrats used the traditional linear morphometrics approach (Marcus 1990) to quantify size variation. However, more modern methods such as geometric morphometrics have proven successful in studies of rodent morphology (Barciova 2009) and are preferable when describing shape variation in highly complex anatomical units such as the skull (MacLeod and Forey 2002). Therefore, we complemented our linear morphometric approach with geometric morphometrics as a powerful means to quantify both size and shape. For this procedure, we photographed the right ventral skull in 103 specimens, including 70 previously sampled specimens (from our original sample size of 281) in addition to 33 previously unsampled ones (Appendix I; Table 5; resulting in 314 total specimens for the study).

We fitted each image with 21 clearly identifiable corresponding landmarks (Table 1), and digitized them in program tpsDig (http://life.bio.sunysb.edu/ee/rohlf/software.html). We chose landmarks so that their placement was nearly identical and in the same plane of orientation for all specimens, then superimposed them using the generalized least squares Procrustes method in the Integrated Morphometrics Package (IMP) (www3.canisius.edu/~sheets/morphsoft.html). The Procrustes method scales the original landmark configurations to a standard size, translates them to a common center of gravity, and rotates them to determine an optimal fit. The remaining variation among landmark configurations represents shape variation in the sample. Differences in the landmark coordinates of groups can then be assessed statistically (reviewed in Zelditch et al. 2004; Klingenberg 2010).
Table 1

Landmarks used in geometric morphometrics

Number

Landmark Position

1

Orale

2

Anterior midline premaxillary-palatal process suture

3

Posterior midline maxillary-palatal process suture

4

Posterior nasal spine

5

Midline presphenoid-basisphenoid suture

6

Midline basisphenoid-occipital suture

7

Anteriormost midline on the occipital

8

Posteriormost midline on the occipital

9

Occipital condyle

10

Auditory bulla-basisphenoid suture

11

Anterior hamular process

12

Posteriormost M3 contact with maxillary

13

Lateralmost M2 contact with maxillary

14

Anteriormost M1 contact with maxillary

15

Premaxillary-maxillary suture proximal to incisive foramen

16

Lateralmost incisive contact with premaxillary

17

Capsular projection for incisor

18

Anteriormost projection of antorbital bridge of the maxillary

19

Posteriormost jugal-squamosal suture

20

Auditory bulla on left side posteriorly to the auditory meatus

21

Auditory bulla on right side posteriorly to the auditory meatus

Statistical Analyses

I. Linear morphometrics

To allow comparing our findings to previous work on N. cinerea, we used regression analysis to determine whether skull size (CBL) could be predicted by body size (TBL). We then used principal component analysis (PCA) to explore patterns of variation in linear morphometric measurements of the skull, and found that most variation was due to the effects of skull size (Appendix II; Table 6). Therefore, we tested for differences in the log-transformed means of linear morphometric measurements of skulls from coastal and interior groups using analysis of covariance (ANCOVA) with CBL as a covariate. After determining that means for bullar breadth (BB), palatal bridge length (PBL), nasal length (NL), and upper tooth row (UTL) differed (all P < 0.001; Table 2) but there were no size effects, we selected those characters for subsequent climate-morphology analyses. Summary data on non-significantly different character means are given in Appendix III; Table 7.
Table 2

Summary statistics of linear morphometric measurementsa used in climate-morphology analyses

 

Coastal

Interior

ANOVA (Coastal versus Interior)

N

Mean (mm ± SE)

CVc

N

Mean (mm ± SE)

CVc

F

P

CSDb

        

TBL

58

398(4.95)

9.46

65

365 (3.41)

7.53

31.1

< 0.0001

CBL

81

48.4 (0.36)

6.61

79

46.7 (0.31)

5.89

12.7

0.001

BB

80

8.48 (0.05)

5.51

79

8.62 (0.05)

5.34

20.9

< 0.0001d

NL

80

19.6 (0.16)

7.47

79

18.6 (0.16)

7.73

10.1

0.002 d

PBL

80

9.59 (0.09)

8.31

70

8.55 (0.09)

9.59

54.2

< 0.0001 d

KMb

        

CBL

38

45.1 (0.28)

3.96

82

46.8 (0.33)

6.32

11.9

0.001

BB

38

8.02 (0.06)

4.79

82

8.62 (0.05)

5.72

48.1

< 0.0001 d

PBL

38

8.51 (0.08)

6.17

82

8.51 (0.06)

6.95

12.8

0.001 d

UTL

38

8.98 (0.06)

4.28

82

9.69 (0.04)

4.21

65.8

< 0.0001 d

aBB bullar breadth; CBL condylobasal length; NL nasal length; PBL palatal bridge length; UTL upper tooth row length

bCascades (CSD) and Klamath (KM) transects

ccoefficient of variation (CV)—the standard deviation standardized by the mean

dskull size represented by CBL was used as a covariate

II. Geometric morphometrics

We tested for differences in the geometric morphometric means for coastal and interior groups using Goodall’s F-test (Goodall 1991). That statistic, adapted for geometric morphometric data, assesses significance of the distance among landmark configurations after Procrustes superimposition. To visualize shape differences between coastal and interior groups, we generated deformation plots that depicted the magnitude and direction of shape variation at each landmark (reviewed in Zelditch et al. 2004; Klingenberg 2010).

After determining that skull size (represented by centroid size) did not have an effect on shape (Goodall’s F-test; P > 0.05), we conducted further analyses without adjusting for size. Centroid size is the square root of the summed squared distances of landmarks from the chosen center of gravity in the Procrustes superimposition (reviewed in Zelditch et al. 2004; Klingenberg 2010). We chose that metric as a proxy for skull size because values for skull size (CBL) were not available for all specimens. We used IMP for statistical analyses and visualization of geometric morphometric data, and JMP 9 (SAS, Inc) for other statistical analyses. Geometric morphometric analyses required resampling-based methods due to small sample sizes relative to the number of variables examined; therefore, significance tests were derived using 2,500 bootstraps.

Phenotypic Variation and Climate

Data Acquisition

We used climate data (1 km resolution) generated in the WORLDCLIM algorithm (Hijmans et al. 2005) to test whether variation in the skull morphology of N. cinerea was predicted by the climatic regimes associated with the coastal and interior zones. We selected the three climatic variables used by Smith et al. (2009) in an N. cinerea ecological niche model: (1) maximum temperature of the warmest month (BIO 5), (2) minimum temperature of the coldest month (BIO 6), and (3) annual mean precipitation (BIO 12). In addition, we included annual mean temperature (BIO 1) because it was used in previous analyses of body size variation (Brown 1968; Brown and Lee 1969). We also considered temperature annual range (BIO 7) because the interior zones are characterized by a highly fluctuating temperature regime (Price 1978), a factor that presumably underlies adaptive evolution in desert rodents (Walsberg 2000).

We downloaded museum-recorded geographic data for N. cinerea localities from the Mammal Networked Information System (http://manisnet.org/). For specimens lacking precise locations, we used Google Earth 5.0 (Google, Inc.) to determine the coordinates of the geographic center of the locality (e.g., town or topographic landmark) denoted on specimen tags. Localities (N = 18) that could not be accurately identified or located were excluded. We then used DIVA-GIS 5.2 (Hijmans et al. 2001) to extract WORLDCLIM data for 140 N. cinerea localities within our study area. We log-transformed climatic values and standardized them to range from 0 to 1 because units of measure differed among variables.

Statistical Analyses

I. Correlation of climate and skull morphology

To demonstrate climatic differences between coastal and interior N. cinerea localities of each transect, we tested for correlation among climatic variables and altitude (ALT), latitude (LAT) and longitude (LONG) (Appendix IV; Table 8). We then used Mantel tests to assess correlation between differences in skull morphology and the climates of coastal and interior zones for each transect. Because that test requires comparison of similarity (i.e., distance) matrices, we estimated Mahalanobis distance values (reviewed in Marcus 1990) in JMP 9 to represent variation in the combined suite of linear morphometric measurements for each specimen, as well as the set of climatic variables for specimen locality. We also used partial Mantel tests to assess correlation between skull morphology and climate on each transect, while controlling for the combined effects of LAT and LONG on climate and morphology by including a similarity matrix of geographic distance (log km between localities). We restricted our climate-morphology analyses to linear morphometric measurements due to the limited size of the geometric morphometric data set.

II. Phenotypic variation as a function of climate

We employed a multivariate approach to test whether phenotypic variation in woodrat skulls was predicted by coastal-interior climatic (BIO 1, BIO 5, BIO 6, BIO 7, BIO 12) and geographic (ALT, LAT, LONG) variables. We first used univariate regression to guide construction of multivariate models as well as to explore the relative contribution of each of those variables to explaining morphological variation (Appendix V; Table 9). We selected only highly significant associations (P < 0.0001) with moderate explanatory power (R2 > 0.20) for inclusion in our multivariate modeling. To further maximize the informative power of multivariate models, we restricted our analyses to dependent morphological variables with at least two univariate regression models that met the above criterion. Though variation in CBL was weakly predicted in the CSD transect (Appendix V; Table 9), we included it in order to compare patterns to those of TBL. Our variable screening approach enabled us to construct models with high informative value and decreased bias (i.e., parsimony in model construction; Anderson 2008). We then ranked the informative value of each multivariate model using the Akaike information criterion with second-order bias correction (AICc). Model ranking was conducted in a reverse stepwise manner whereby the parameter with the least significant effect was removed sequentially. We considered models within two AICc units of the highest-ranking model as competing models.

Results

Morphological Variation

Linear Morphometrics

Using linear skull measurements, we detected two patterns of skull phenotypic variation in N. cinerea. Mean values for BB, PBL, and NL differed between coastal and interior populations within the Cascades (CSD) transect, while within the Klamath transect, BB, PBL and UTL differed between coastal and interior populations (Table 2). However, CSD coastal populations were larger than interior ones with respect to most measured characters used in climate-morphology analyses, while coastal KM woodrats were smaller than interior woodrats (Table 2, 7; Appendix III). Mean skull size (CBL) in coastal and interior populations differed in both transects. Mean body size (TBL) differed in coastal versus interior CSD populations (Table 2) and this factor was a significant predictor of skull size (CBL) variation (F1,125 = 210, P < 0.0001, R2 = 0.63). Data on TBL were not available for the KM transect.

Geometric Morphometrics

Skull shape of coastal and interior groups differed within both the CSD (Goodall’s F38,2166 = 3.19, P < 0.001) and KM (Goodall’s  F38,1596 = 10.6, P < 0.001) transects. Deformation grids of coastal and interior CSD populations differed the most with respect to the shape of the jugal-squamosal bones (posteriorly) and the dental row (Fig. 4; Table 1). In addition, those populations differed in the position of the auditory bulla-basisphenoid suture (landmark 10), as well as in the position of the incisive foramen relative to that of the anteriormost molar tooth (M1) (landmark 3). We observed the most pronounced variation in the position of the incisive foramen relative to M1 in CSD specimens compared to those of the KM transect. Coastal and interior KM populations also differed prominently with respect to the shape of the jugal-squamosal bones (landmark 19), but differences in the shape of the tooth row were detected on the posteriormost molar (M3). We noted a shift in the orientation of the foramen magnum of KM specimens as a result of pronounced shape deformation of the occipital (landmarks 6–8). Also, these populations differed considerably in the shape of the premaxillary-maxillary (landmarks 1, 17–18) and auditory bulla (landmarks 10, 20–21) regions (Fig. 4; Table 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-012-9187-0/MediaObjects/10914_2012_9187_Fig4_HTML.gif
Fig. 4

Shape deformation plots of coastal and interior populations within the CSD (left) and KM (right) transects. Each black circle represents the mean landmark position of coastal specimens and vectors (above) depict the magnitude and direction of variation relative to interior specimens. Landmark 9 (gray circle) was used as the reference landmark. A picture of the right ventral skull of N. cinerea skull, with number-labeled landmarks (see Table 1), is provided for reference

Phenotypic Variation and Climate

Correlation of Climate and Skull Morphology

Specimen localities for coastal and interior N. cinerea populations had highly contrasting climatic regimes. Annual mean precipitation (BIO 12), temperature annual range (BIO 7), maximum temperature of the warmest month (BIO 5), and minimum temperature of the coldest month (BIO 6) were strongly correlated with longitude in the KM transect. Those variables, except BIO 7, were also strongly correlated with longitude in the CSD transect. Annual mean temperature (BIO 1) was weakly correlated with longitude in both transects (Appendix IV; Table 8). In the CSD transect, Mantel tests demonstrated a stronger correlation between morphology and climate than between morphology and geographic distance or group classification (coastal-interior categories; Table 3). A similar pattern was observed in the KM transect but the correlation of morphology and group classification was stronger than that of morphology and climate. We confirmed these results using partial Mantel tests: for both transects, the correlation of morphology with climate, controlled for geographic distance, was almost as strong as the correlation with climate alone. Correlations of morphology and climate, controlling for coastal-interior categories, were moderate and those of morphology and coastal-interior categories, controlling for climate, were weak (CSD) to moderate (KM) (Table 3).
Table 3

Mantel and partial Mantel tests of the correlation of morphologya with climateb, geographic distance and Coastal-Interiorc classification

 

CSDe

KMe

Test

Variables

r

P

r

P

Mantel

Morphology*Log(km)

0.16

<0.0001

0.20

<0.0001

Mantel

Morphology*Climate

0.32

<0.0001

0.44

<0.0001

Mantel

Morphology*Coastal-Interior

0.26

<0.0001

0.51

<0.0001

Partial Manteld

Morphology*Climate.Log(km)

0.29

<0.0001

0.40

<0.0001

Partial Manteld

Morphology*Climate.Coastal-Interior

0.22

<0.0001

0.13

<0.0001

Partial Manteld

Morphology*Coastal-Interior.Climate

−0.06

<0.0001

0.31

<0.0001

a“Morphology” BB bullar breadth; NL nasal length; PBL palatal bridge length; UTL upper tooth row length

b“Climate” annual mean temperature (BIO 1); maximum temperature of the warmest month (BIO 5); minimum temperature of the coldest month (BIO 6); temperature annual range (BIO 7); annual mean precipitation (BIO 12)

c“Coastal-Interior” 0 for pairwise comparisons within the same zone, 1 for comparisons across the coastal-interior divide

dcorrelation of the first two variables (A*B) while controlling for correlation with the third (.C)

eCascades (CSD) and Klamath (KM) transects

Phenotypic Variation as a Function of Climate

Climatic variables, excluding BIO 1, were significant predictors of body and skull linear measurements as determined by univariate regression (Figs. 56; Appendix V; Table 9). In the CSD transect, we tested for variation in CBL, PBL, and TBL using multivariate regression models that included ALT, LAT, BIO 5, BIO 6, BIO 7, and BIO 12 as parameters. Those same parameters, in addition to LONG but excluding ALT and LAT, were used to test for variation in BB and UTL in the KM transect. In the CSD transect, models that consistently included BIO 7 and BIO 12 were highest ranked by AICc (Table 4). Models that consistently included BIO 12 were highest ranked in the KM transect. In both transects, highest ranking models did not include BIO 6, and BIO 5 was only included in highest-ranking models for PBL (CSD transect) and UTL (KM transect). However, BIO 5 and BIO 6 were often included in competitive models (≤ 2 AICc units from highest ranked model). In the CSD transect, inclusion of geographic variables ALT and LAT resulted in decreased information loss and optimization of fit in high-ranking models. Similarly, LONG was included as a parameter in all highest-ranking models in the KM transect.
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-012-9187-0/MediaObjects/10914_2012_9187_Fig5_HTML.gif
Fig. 5

Regression plots of palatal bridge length (PBL; ad) and total body length (TBL; eh) on maximum temperature of the warmest month (BIO 5), minimum temperature of the coldest month (BIO 6), temperature annual range (BIO 7), and annual precipitation (BIO 12) in the Cascades transect. Black and hollow circles represent coastal and interior localities respectively. The proportion of variation explained (R2) by each regression model is indicated in the lower left hand corner of each plot

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-012-9187-0/MediaObjects/10914_2012_9187_Fig6_HTML.gif
Fig. 6

Regression plots of bullar breadth (BB; ae) and upper tooth row length (UTL; fj) on maximum temperature of the warmest month (BIO 5), minimum temperature of the coldest month (BIO 6), temperature annual range (BIO 7), and annual precipitation (BIO 12) in the Klamath transect. Black and hollow circles represent coastal and interior localities respectively. The proportion of variation explained (R2) by each regression model is indicated in the lower center of each plot

Table 4

Summary of multiple regression model selection on linear morphometrica variation explained by climaticb and geographicc variables

Trans.d

Dependent variable

Model

N

Adjusted R2

AICce

∆AICc

wf

Kg

CSD

CBL

LAT + BIO 7

157

0.20

−704.0

0.0

0.68

2

CSD

CBL

ALT + LAT + BIO 7

157

0.20

−701.9

2.1

0.24

3

CSD

CBL

ALT + LAT + BIO 5 + BIO 7

157

0.19

−699.8

4.2

0.08

4

CSD

CBL

BIO 7

157

0.12

−690.2

14

0.00

1

CSD

PBL

LAT + BIO 5 + BIO 7 + BIO 12

156

0.47

−606.4

0.0

0.63

4

CSD

PBL

ALT + LAT + BIO 5 + BIO 7 + BIO 12

156

0.47

−604.4

2.0

0.23

5

CSD

PBL

ALT + LAT + BIO 5 + BIO 6 + BIO 7 + BIO 12

156

0.47

−603.4

3.0

0.14

6

CSD

PBL

LAT + BIO 7 + BIO 12

156

0.43

−596.3

10

0.00

3

CSD

PBL

LAT + BIO 7

156

0.42

−593.7

13

0.00

2

CSD

PBL

BIO 7

156

0.32

−571.0

35

0.00

1

CSD

TBL

ALT + BIO 12

120

0.33

−467.2

0.0

0.49

2

CSD

TBL

ALT + BIO 6 + BIO 12

120

0.33

−466.2

1.0

0.30

3

CSD

TBL

ALT + BIO 5 + BIO 6 + BIO 7 + BIO 12

120

0.33

−464.0

3.0

0.11

5

CSD

TBL

ALT + BIO 6 + BIO 7 + BIO 12

120

0.33

−464.2

3.2

0.09

4

CSD

TBL

BIO 12

120

0.27

−458.7

8.5

0.01

1

KM

BB

LONG + BIO 12

118

0.31

−537.9

0.0

0.47

2

KM

BB

LONG + BIO 6 + BIO 12

118

0.31

−536.7

1.2

0.26

3

KM

BB

LONG

118

0.29

−535.5

2.4

0.14

1

KM

BB

LONG + BIO 5 + BIO 6 + BIO 12

118

0.30

−534.8

3.1

0.10

4

KM

BB

LONG + BIO 5 + BIO 6 + BIO 7 + BIO 12

118

0.30

−532.0

5.9

0.03

5

KM

UTL

LONG + BIO 5 + BIO 12

118

0.39

−588.1

0.0

0.54

3

KM

UTL

LONG + BIO 5 + BIO 6 + BIO 12

118

0.39

−586.2

1.9

0.21

4

KM

UTL

LONG + BIO 12

118

0.38

−585.9

2.2

0.18

2

KM

UTL

LONG + BIO 5 + BIO 6 + BIO 7 + BIO 12

118

0.39

−584.0

4.1

0.07

5

KM

UTL

LONG

118

0.32

−576.9

11

11

1

aBB bullar breadth; CBL condylobasal length; PBL palatal bridge length; TBL total body length; UTL upper tooth row length

bmaximum temperature of the warmest month (BIO 5); minimum temperature of the coldest month (BIO 6); temperature annual range (BIO 7); annual mean precipitation (BIO 12)

caltitude (ALT); latitude (LAT); longitude (LONG)

dCascades (CSD) and Klamath (KM) transects

emultiple regression models were ranked using the Akaike information criterion with second-order bias correction (AICc). Competing models (bold text) within 2 ≤ AICc units (see ∆AICc) of the highest ranking model were also considered as having high explanatory power with least information loss.

fWeighted model average

gNumber of model parameters

Discussion

Morphometric Differentiation of Coastal and Interior Woodrats: Traditional Versus Modern Approaches

Coastal and interior populations of the bushy-tailed woodrat (N. cinerea) in the Pacific Northwest region of the United States differed in skull morphology; this finding was consistent with previous qualitative surveys (Hooper 1940; Verts and Carraway 1998). Using traditional linear morphometrics, we found that populations differed in the size of functionally significant traits (Table 2). Those patterns were partially validated and further characterized by geometric morphometric analyses on shape variation. Thus, our combined linear and geometric morphometrics approach proved highly efficient. For example, while coastal and interior populations of the KM transect did not differ in zygomatic breadth (ZB; Appendix III; Table 7), we detected the strongest pattern of shape variation in the posterior region of the zygomatic arch (landmark 19, Table 1; Fig. 4). Because this is the attachment site of chewing (masseter) muscles, bone shape differences in this region are hypothetically related to the biting force and diet of rodents (reviewed in Samuels 2009). Similarly, while coastal and interior CSD woodrats did not differ in upper tooth row length (UTL), we did detect molar shape variation. In contrast, KM coastal and interior woodrats differed in UTL but molar shape variation was not pronounced. We cannot elaborate on all observed patterns of shape variation, such as those of the occipital region of KM woodrats (Fig. 4), because their adaptive significance is unclear. Moreover, we could not test whether skull shape variation depended on climate due to our small sample size. Nonetheless, we provided strong evidence to support variation in the skull morphology of coastal versus interior N. cinerea ecotypes.

Differences between the CSD and KM Transects

Phenotypic responses to climatic variation between coastal and interior populations of N. cinerea were different in the CSD versus KM transects. Could these differences be due to climate or phylogeny? Phylogeographic analyses suggest that our coastal and interior populations may have had disjunct geographic distributions during the last glacial period (Hornsby and Matocq 2012). Thus, some of the coastal and interior differences could result from lack of historical gene flow across the coastal-interior divide in the KM transect (N. c. pulla versus N. c. alticola). The distribution of woodrats in the CSD transect is not discontinuous (N. c. fusca + N. c. occidentalis versus N. c. alticola + N. c. occidentalis), but it may have been during recent glaciation events over the area that divides coastal and interior zones (Hornsby and Matocq in 2012). Testing whether phenotypic variation in our study populations depends on phylogenetic history would require a population genetics approach or perhaps a comparison of coastal and interior populations to coastal-interior contact zone populations. We were unable to include comparisons to intermediate climatic zones because of the abrupt transitions from coastal to interior environments. In this region, means for annual precipitation may transition drastically from 980 to 230 mm going from coastal to interior climates as a result of a “rain shadow” effect (Anderson et al. 1998; Taylor 1999). Therefore, intermediate climatic zones were geographically constrained and specimen sample sizes would not have been large enough for statistical comparisons.

Because the climate, geology, and vegetation of coastal-interior gradients of CSD versus KM transects differ (Anderson et al. 1998; Taylor 1999), we can assume that woodrats in the coastal and interior zones of those transects interacted with different resources and skull phenotypes responded accordingly. Even so, both transects shared expected commonalities: maximum precipitation of the warmest month (BIO 5) and temperature annual range (BIO 7) increased with distance from the coast; annual mean temperature (BIO 1), minimum temperature of the coldest month (BIO 6), and annual mean precipitation (BIO 12) decreased with the distance from the coast. We found strong support for correlation of morphological and climatic differences among woodrat specimens and their localities, even after controlling for geographic distance (Table 3). Interestingly, the correlation of morphology and coastal-interior classification, controlled for climate, was moderate in the KM transect but weak in the CSD transect (Table 3). Furthermore, including longitude as a parameter optimized the explanatory power of multivariate models of morphological variation (Table 4). These results suggest that some factor besides climate differences could contribute to morphological differences between coastal and interior populations of the KM transect. For instance, the distribution of woodrats on the KM transect appears to have a gap from W −122° to W −121.5°, coinciding with our delineation of coastal and interior zones. This observation is intriguing because boundaries of climatic zones are hypothesized to act as barriers to gene flow during mammalian speciation events (Barnosky 2005).

Variation in Food Processing Traits is Associated with Climate

Differences in food preference and feeding structures of vertebrates are expected to arise in accordance with environmental variation (Skulason and Smith 1995). As such, food is one of the most important environmental factors underlying intraspecific variation in skull morphology (Hanken and Hall 1993). In mammals, genetically- (heritable) or non-genetically- (due to environmental plasticity) based variation in dental size and shape is often a response to differences in food processing (Butler 1983; Samuels 2009; Ungar 2010). Therefore, examination of teeth and associated structures could reveal a signature of local adaptation when comparing populations that are subjected to different environmental conditions and, in turn, food resources. Indeed, we found that the strongest phenotypic responses to climate were those related to dental size (i.e., UTL) and palatal bridge length (PBL) of N. cinerea populations subjected to opposing climatic regimes (Figs. 56; Table 4). Those lines of evidence provided convincing support for our hypothesis: skull phenotypic variation must be associated with climatic conditions of different environments assuming that climate determines the composition of resources available to populations.

Though little is known about the dietary preferences of herbivorous N. cinerea within our study area, coastal and interior populations inhabited climatic zones with distinctly adapted floral communities (Price 1978) suggesting that populations may exploit different food resources. Not surprisingly, most of the plant species on which coastal populations feed do not occur in the interior zone of the CSD transect (Maser et al. 1981). Some woodrat species appear to have dietary preferences that are likely governed by physiological adaptation to local environments. In experimental trials, Neotoma fuscipes from moist mixed-coniferous and dry juniper forests preferentially selected highly toxic plants found in their respective environments (McEachern et al. 2006). A comparison of extinct and modern woodrats suggested that dietary shifts, as a consequence of increasing aridity, resulted in the adaptive variation of teeth (Zakrzewski 1993). Although similar evolutionary patterns are consistent across mammals (Blois and Hadly 2009), we recognize that not all patterns of phenotypic variation in food processing traits are adaptive. For example, experiments demonstrated a causal link between diet and mechanically induced plasticity of the mammalian palatal bridge during post-natal development (Menegaz et al. 2009). Furthermore, dental variation in cheek teeth is often due to wear as a result of aging or the consumption of highly abrasive foods (Ungar 2010). Nonetheless, natural selection is still expected to operate to some degree on dental size and shape (Butler 1983). We therefore propose that patterns of variation in food processing traits of N. cinerea were likely the result of selection and environmental plasticity acting in concert.

Middle Ear Size Variation and Climate

The middle ear of small mammals should scale isometrically with body size; that is, the geometric dimensions of this organ are not expected to change as body size increases (Nummela 1995). Consequently, the middle ear is expected to process similar levels of sound energy independent of body size (Nummela 1995). Interestingly, we found that although coastal CSD woodrats had larger body and skull size than interior populations, the auditory bulla was larger (i.e., bullar breadth; BB) in interior specimens and its shape differed slightly compared to coastal specimens (Fig. 4). Because the auditory bulla houses the middle ear, we considered BB as a proxy measurement for middle ear size. Therefore, we speculate that apparent differences in the isometric scaling of the middle ear between coastal and interior CSD woodrats could indicate adaptive variation in the efficiency of this sensory organ. Such phenotypic pattern is convergent in the desert woodrat (Neotoma lepida) from Southern California: coastal populations had larger skulls but with small auditory bullae relative to interior populations (Grinnell and Swarth 1913; Patton et al. 2008). Other convergent phenotypic patterns included differences in the maxillary-premaxillary skull region (Fig. 4), described as “angular” in coastal versus interior N. cinerea of the KM transect (Hooper 1940), and detected in the coastal-interior comparison of N. lepida by Grinnell and Swarth (1913). In our study, coastal woodrats in the KM transect were already smaller with respect to all linear skull measurements including BB and skull size (Table 2, 7; Appendix III), and auditory bulla shape did not vary between coastal and interior woodrats (Fig. 4). However, variation in BB was strongly explained by the climate of that transect (Fig. 6) and when accounting for interaction among multiple climatic and geographic variables, precipitation (i.e., BIO 12) still had the strongest effect on BB variation (Table 4). Surprisingly, climate did not explain BB variation in the CSD transect (Appendix V; Table 9).

Adaptive hypotheses on middle ear size variation in rodents invoke climate as having an effect (Webster and Webster 1975, 1980; Liao et al. 2007). Experimental data demonstrated that rats (Heteromyidae) in desert environments had enhanced auditory sensitivity to low frequency sound as a result of increased middle ear size (Webster and Webster 1980). Because xeric environments feature little vegetative cover and strong winds, increased sensitivity to low frequency sound could improve the use of auditory cues to detect predators (Webster and Webster 1980). Associations between moisture and auditory bulla morphology have been described in other rodent species (Monteiro et al. 2003), including N. cinerea (Hornsby 2009). Consistent with our findings, Great Basin Desert subspecies of N. cinerea have the largest auditory bullae relative to skull size and that variation was explained by precipitation (Hornsby 2009). Although increased middle ear size via inflation of the auditory bulla is a common adaptation of desert rodents to arid environments (Mares 1993), other factors need to be considered to comprehensively address whether such pattern is adaptive in N. cinerea. For example, altitude could also influence middle ear size variation (Liao et al. 2007). Within the context of our study, patterns of association among middle ear size and temperature and precipitation (KM transect), as well as middle ear size variation on its own (CSD transect), are taken as evidence to support our hypothesis. If previously described form-to-function relationships of the mammalian middle ear (Webster and Webster 1975; 1980; Nummela 1995) are generalizable to N. cinerea and climate (precipitation) determines resource characteristics (vegetative cover), then variation in middle ear size should have responded in a resource-dependent manner as predicted (Skulason and Smith 1995).

Are Previously Described Climate-Size Generalizations Applicable to our Study Populations?

Describing body size variation was not our primary objective, but we did so to compare our results to previous works that focused primarily on the relationship between climate and body size of N. cinerea (Brown 1968; Brown and Lee 1969; Smith et al. 1995; Smith and Betancourt 1998; Smith et al. 1998). In accordance with Bergmann’s rule, maximum summer temperatures were found to be below the physiological lethal threshold (36.3°C) of coastal N. cinerea in the CSD transect and body size in those populations was predicted to be larger than in interior ones (Brown and Lee 1969). We found moderate support for that prediction, as woodrat body size was associated with mean maximum temperature for the month of July (i.e., BIO 5), and also demonstrated that the influence of precipitation on phenotypic variation may be equally important as shown in other mammals (Yom-Tov and Geffen 2006; Pergams and Lawler 2009). Multivariate models of body size variation in CSD woodrats were highly informative when including annual precipitation (BIO 12) as a parameter (Table 4), while effects of temperature-related variables were in agreement with previous studies (Smith et al. 1998; Smith and Betancourt 2003, 2006).

Phenotypic Responses to Climate vary According to Organizational Level

Modularity theory predicts that correlated traits across discrete hierarchical levels of biological organization may display different patterns of variation in response to opposing or variable selective pressures associated with environment (reviewed in Klingenberg 2010). Previous studies of the mammalian skull support this assumption (Porto et al. 2009; Marroig et al. 2009; Goswami and Polly 2010), as do our findings. Skull size and body size were strongly associated but those traits responded differently to climate. Moreover, measurements of skull components were correlated with overall skull size and displayed the strongest responses to climate whereas skull size on its own did not. Was the lack of association of skull size and climate due to opposing selective demands for optimal body size versus size and shape of food processing and sensory traits of the skull? Metabolism may impose a constraint on how large a woodrat can grow and thus natural selection may operate on body size (Brown 1968; Brown and Lee 1969). Similarly, selection may operate on optimal dental size and shape in response to dietary variation (Butler 1983), though diet could also influence variation in overall skull size of mammals (see Goheen et al. 2003). Our data cannot address these complexities but they are noteworthy as studies often attempt to extrapolate climate-size relationships of skull or dental elements to body size. Our findings underscore that patterns of variation at one anatomical level do not always reflect those of another one.

Conclusions

Using a three-tier study approach focusing on climate, morphology, and climate-morphology, we provided strong evidence that phenotypic variation in modern populations of N. cinerea reflected a response to contrasting moist coastal and arid interior environments within the Pacific Northwest region of the United States. First, we confirmed that coastal and interior populations differed in skull morphology and body size as previously described. Secondly, we described patterns of climatic variation that were consistent with previous observations and these were correlated with morphology. Lastly, we showed that biologically relevant climatic variables were moderately to strongly associated with functionally important skull traits. However, we recognize that climate is one of several factors that could influence phenotypic variation and other competing hypotheses need to be addressed. In light of these complications, we argue that an integrative approach focusing on multiple climatic and hierarchical phenotypic variables is necessary to extend the current understanding on how mammals adapt to their environment. Although our conclusions can only truly be validated through tests of heritability, selection, environmental plasticity and phylogenetic history, the previously overlooked sources of phenotypic variation that we described will help inform future studies of extinct and contemporary populations of N. cinerea—a focal taxon for the evaluation of microevolutionary responses to climate change in western North America.

Acknowledgements

Funding was provided by the Oregon State University Undergraduate Research, Innovation, Scholarship and Creativity grant. Portions of this investigation were completed through the University of California Museum of Vertebrate Zoology (MVZ) research internship program. We thank the curatorial staff of the MVZ for their assistance and helpful comments, especially Chris Conroy. We thank all other museum curators for providing access to specimens.

Copyright information

© Springer Science+Business Media, LLC 2012