Materials and methods

Study species. Quercus rugosa (section Quercus) is an ecologically important tree that can reach 25 m in height. It is an evergreen or sometimes brevideciduous species. The leaves are usually cupped, rarely flat, thick and leathery with a thick yellow-brown tomentum on the abaxial surface and up to 10 cm in length. The bark is brown and scaly. The acorn of Q. rugosa is ovoid, elongated and wider at the base than at the apex. It is produced individually or in groups of 2 - 3 with a cup of 10 - 15 mm in diameter with brown or blackish scales. The cotyledons are usually reddish or pinkish, which is a distinctive trait of the species. Quercus rugosa has a wide geographical distribution, with a latitudinal range in Mexico that extends from the temperate zones of the Sierra Tarahumara in the State of Chihuahua, to the subtropics in the highlands of Los Altos de Chiapas in southern Mexico, and Guatemala at altitudes ranging from 1,700 m to 3,550 m (^{Rzedowski 1986}, ^{Uribe-Salas et al. 2008}). It is one of the dominant species over much of this range, and it can be found in monospecific stands or cohabiting with other species of oak or pine.

Acorn collection and measurement. During the last week of October and the first week of November of 2009 we collected a total of 2,810 acorns in seven populations representing the latitudinal gradient of the distribution of the species in Mexico, from Chiapas to Chihuahua (Figure 1, Table 1). Mature and undamaged acorns were collected directly off the tree or from the ground below isolated trees from five different individuals per site. The number of seeds collected per mother tree varied between 16 and 180 and per site it varied between 240 (Casas Grandes) and 625 (San Cristóbal de las Casas) (Table 2). Groups of acorns from the same mother tree were packaged together in plastic burlaps, placed in a plastic cooler at environment temperature and transported to the laboratory where they were processed immediately. Transportation time was no longer than one week in all cases.

Figure 1 Map showing the seven Quercus rugosa populations collected in Mexico along the latitudinal/climate gradient. The numbering of populations is as in Table 1. 

In a sample of 15 randomly chosen undamaged acorns from each of the five mother trees per population, we separated the cup from the nut and obtained the nut length (mm) and width (mm) with a digital caliper, and the fresh mass with an analytical balance. The volume (V) of each acorn was approximated assuming a prolate spheroid shape, using the formula V = (4/3) πa ² b, where a is the width and b is the length (^{Aizen & Patterson 1990}).

Viability and germination percentage. Acorns were sterilized by submerging them in water with 5 % chlorine bleach for ten minutes. At the same time, this served as a water flotation test of viability. In this way, for groups of acorns from each mother tree, we recorded the number and separated potentially viable seeds (the ones that sank) from inviable ones (the ones that floated). According with this test, the percentage of potential viability per population was between 27.1 and 53.1 % (Table 2). Then, all the potentially viable seeds were rinsed and humidified soaked in water for 24 h at 4 °C in a refrigerator. After that, acorns were individually sown in black plastic bags (25 cm long, 15 cm diameter, with drainage holes) filled with a substrate of 25 % agrolita, 25 % vermiculite and 50 % peat moss. Care was taken to introduce acorns into the substrate in approximately 2/3 of their length, with the apex pointing downwards. Bags were placed in a greenhouse with natural lighting in a randomized pattern and watered two times per week to keep a moisture level of approximately 30 % in the substrate. The position of the bags within the greenhouse was reshuffled every three or four weeks. The experiment was established in late November of 2009 and continued until early May 2010. Seeds were not moved or perturbed during this period and were considered germinated when the first leaves emerged. The germination percentage was recorded twice, in early March and in early May. After this date germination ceased and the results were analyzed, considering the total germination percentage.

Climatic variables. Climate data for each sampling site was downloded from the United States Department of Agriculture Forest Service web page (http://forest.moscowfsl.wsu.edu/climate/current/), a climate data set with derived variables designed for assessing plant-climate relationships (^{Rehfeldt 2006}). We focused on two variables that capture the most important energy and water resource inputs for plants: growing season precipitation (gsp) and growing degree-days above 5 °C within the frost-free period (gsdd5) (a measure of heat accumulation) (Table 1).

Statistical Analyses. Differences among mother trees and among populations in acorn length, width, volume, and fresh mass were evaluated with nested analyses of variance. In these tests, both populations and mother trees within populations were considered to represent random samples from a larger pool. For each character, we calculated the proportion of the total variation (i.e. the variance components) due to differences at each of the two hierarchical levels. To compare germination percentage among populations, χ² tests were used. These tests were performed with the JMP 8 software (SAS Institute, Cary, North Carolina).

To test the abiotic constraints hypothesis, we performed multiple regressions of acorn volume on latitude, longitude, gsp, and gdd5 using R v3.0.0 languaje. All variables were centered prior to analysis and added variable (partial regression) plots were made to visualize the partial correlation of each climate variable after controlling for all other variables. Similar tests were done without including longitude or with only one climate variable at a time, and simple linear regressions were performed for each climate variable with each dependent variable to confirm their correlation. If the two derived climate variables were significant in the multiple regression models after accounting for the spatial variables (especially latitude), then we found support for the abiotic constraints hypothesis.

Results

The nested analyses of variance showed highly significant (P < 0.001) differences among mother trees within populations and among populations for the four morphological acorn traits evaluated (Table 3). However, the proportion of the total variation explained by differences among mother trees wihin populations was in general much smaller (7.47-11.9 %, depending on the trait) than the proportion explained by differences among populations (75.11-79.53 %). The residual variance (9.13-14.39 %) was interpreted as being explained by differences among acorns within mother trees. The population mean values for the four variables evaluated clearly indicated that southern populations had larger acorns than northern populations (Figure 2). Simple linear regressions indicated significant, positive relationships of acorn volume with gsp (R ² = 0.57, P = 0.049) and with gdd5 (R ² = 0.86, P = 0.003) and a strongly negative relationship with latitude (R ² = 0.94, P = 0.0003).

Figure 2 Mean and standard error for the four acorn morphological traits in the seven study populations: (a) length, (b) width, (c) mass, (d) volume. Different letters indicate significant differences according to an analysis of variance followed by Tukey-Kramer tests. 

According to the χ² tests, populations differed significantly in germination percentage (χ² = 239.6; d.f. = 6; P < 0.0001) (Table 2). In general, germination percentage decreased latitudinally from 47.5 % in population San Cristobal de las Casas, Chiapas to 0.83 % in Casas Grandes, Chihuahua, although the relationship was only marginally significant (R ² = 0.54; P = 0.06). Importantly, we found that germination percentage was significantly correlated with acorn volume (R ² = 0.60; P = 0.04) and acorn fresh mass (R ² = 0.74; P = 0.013), highlighting the importance of seed size on germination success.

In the multiple regression model of acorn volume with the two climate variables controlling for spatial variables, both climate variables were significant (P < 0.034), whereas spatial variables were not (P > 0.23) (Table 4, Fig. 3). This model has R ² > 0.99. Multiple regression tests without longitude and with only one climate variable at a time gave similar results (not shown). These tests were not repeated for acorn fresh mass because of the high correlation (R ²=0.97; P < 0.0001) between this variable and volume.

Figure 3 Added variable (partial regression) plots showing the partial correlation of (a) growing season precipitation (gsp) and (b) growing season degree-days above 5 °C (gsdd5) with acorn volume. 

Discussion

Latitudinal variation in acorn size. Quercus rugosa showed a clinal pattern of south-north decrease in population mean acorn mass and volume across the distribution of the species in Mexico (Figure 2). According to the multiple regression analysis, growing degree-days > 5 °C and growing season precipitation are both significant when controlling for latitude and longitude, supporting the abiotic constraints hypothesis (Table 4, Figure 3). What is striking is that these relationships are highly significant even with a small sample size of seven sites. Although the relationship of latitude with acorn volume might become significant with increased sampling, this possibility would not change our interpretation. Most importantly, it is the significant correlations with climate that strongly support the abiotic hypothesis and specifically support the idea that primary water and energy inputs are important for acorn size and germination percentage in Q. rugosa.

Reduced seed size along latitudinal gradients in species exhibiting substantial northward postglacial migration have been interpreted as evidence of selection on small seed size because more dispersable seeds would be more likely to advance and establish northward (^{Cwynar & MacDonald 1987}, ^{Koenig et al. 2009}). Our results from a Mexican oak whose history is not confounded by glacial migration history question that idea and instead support studies showing that the main determinants of within-species variation in acorn size are climate factors that constrain seed development.

Germination success. Several studies have shown a positive effect of acorn size on germination success, and seedling survival and growth at the intraspecific level (^{Bonfil 1998}, ^{Tripathi & Khan 1990}, ^{Tilki 2010}, ^{Urbieta et al. 2008}, ^{Sage et al. 2011}). These effects are mostly due to the amount of resources stored in the seed, because larger seeds can survive for longer periods of time under unfavorable conditions, while smaller seeds will consume their food reserves in the process of respiration and physiological adjustment. After germination, seedlings that emerge from larger seeds will have more energy for initial growth and for recovery from herbivore attack (^{Bonfil 1998}, ^{Urbieta et al. 2008}). However, most of these studies have involved a single or few populations, and in few cases the relationship between acorn size and germination success has been examined along large geographical or environmental gradients.

One important question that arises from the detection of a clinal pattern in acorn size is whether variation is a result of plastic responses to the limiting climatic factors or is genetically determined, as a result of selection for the production of viable offspring under the constraints of local climatic conditions (^{Aizen & Woodcock 1992}). Our results showed that the latitudinal decrease in population mean acorn size was correlated with a concomitant reduction in germination percentage. In fact, the germination success in the northernmost population was extremely low, less than 1 %. However, it must be acknowledged that, in general, the germination percentages reported in this study are somewhat lower than the ones found in other studies with geographically restricted samples of the same species (40-92 %) (^{Bonfil & Soberón 1999}, ^{López-Barrera & Newton 2005}, ^{Huerta-Paniagua & Rodríguez-Trejo 2011}). These differences could be due to methodological issues related to transportation and storage of the acorns but, most probably, to the criteria used to score germination. In other studies, seeds have been considered germinated when the radicle reaches a size equal to the length of the same seed (e.g. ^{Huerta-Paniagua & Rodríguez-Trejo 2011}), while in our case we waited until the emergence of the first leaves. However, even in this case, the extremely low germination percentage in the Casas Grandes populations seems odd and it could be explained if during the year of collection trees or acorns were subjected to an extreme stress factor (i.e. frost or drought). Therefore, further studies should also consider interannual variation in seed size and viability.

Nevertheless, we consider that our data in general reflect the fact that seed production and development faces more severe constraints at the northernmost part of the range of Q. rugosa. Overall, our results do not seem to support the size-optimization-through-selection hypothesis to explain acorn size variation in Quercus rugosa. Also, according to our personal observations, populations of this oak species become progressively smaller and its distribution more fragmented towards the north, suggesting limited recruitment success at higher latitudes. As was also stated by ^{Aizen & Woodcock (1992)}, this evidence indicates that the latitudinal decline in acorn size is maladaptive for seedling establishment, with important implications for the delineation of northern limits of species ranges.