This study was carried out in the Natural Protected Area of Sierra Fría (21˚52'4''N, 102˚22'44''W and 23˚31'17''N, 102˚50'53''W), in the state of Aguascalientes, central Mexico (Figure 1). Its average temperature is 14.5˚C, with a median annual rainfall of 651.4 mm. The vegetation consists mainly of temperate oak, oak-juniper or pine-oak forests at altitudes ranging from 1900 to 2800 m.a.s.l.

Quercus sideroxyla Humb. & Bonpl., or red oak (Lobatae), have trunk heights varying between 5 and 20 m, al- though some trees of this species reach 30 m in height. It is an endemic species with narrow distribution to cool- er and wetter areas in the Northeastern (Nuevo León, Tamaulipas), Northwestern (Sonora, Chihuahua, Durango) and North-Central forests (Aguascalientes, Zacatecas, San Luis Potosí, Jalisco) of Mexico. Flowering occurs in May and fructification from September to November [15] . Its regeneration occurs by sexual reproduction (acorn production) and clonal growth (root suckers). However, field observations of these individuals showed that the origin of these individuals is mainly by clonal growth in more than 98% of cases. No acorn production was rec- orded from 2006 to 2010 in any sites with the exception of one study site (La Ciénega). However, despite acorn production, annual surveys from 2007 to 2010 recorded no recruitment of new individuals in the field at this site.

29 sites were selected in the oak forests, in each site one plot of 1500 m² was established. Field data obtained in- side each plot included: 1) latitude and longitude, recorded with a handheld GPS (Garmin 2.01), 2) altitude and slope orientation and 3) abundance and basal area.

To test the association of environmental variations with the abundance of Q. sideroxyla on 29 sites selected, we performed the Outlying Mean Index (OMI), implemented in ADE-4 software [18] , relating 13 variables with the abundance: 1) altitude (Alt), 2) Annual Average Temperature (AAT), 3) Monthly Average of Minimal Temperature (MMinT), 4) Minimal Temperature of Maximal Temperature in warmer Quarter (MiTMxTWQ), 5) Maximal Temperature of Maximal Temperatures in Colder Quarter (MaTMxTWQ), 6) Minimal Temperature of Minimal Temperatures in Warmer Quarter (MinTMiWC), 7) Maximal Temperature of Minimal Temperatures in Colder Quarter (MaTMiCQ), 8) Temperature Standard Deviation (TSD), 9) Maximum precipitation of driest quarter (MPPDQ), 10) Annual maximum precipitation (TMaP), 11) Annual minimal precipitation (TMiP), 12) Minimum precipitation of wettest quarter (MPPWQ) and 13) Monthly precipitation (MP). The last 12 variables were derived from WorldClim climatic layers.

The analysis used two matrices: a botanic matrix, containing the abundance of the species (rows) on 29 sites (columns), and an environmental matrix, containing 13 environmental variables (rows, one topographic and twelve climatic) on the same 29 sites (columns). The environmental matrix was analyzed with PCA to ordinate the sites in function of environmental variables, though the correlation matrix. The inertia variance of the aver- age environmental conditions used by the specie, was used to derive three parameters: 1) Outlying Mean Index (marginality index), which measures the distance between the mean environmental conditions used by species and the mean environmental conditions of the sampling area; 2) tolerance index (niche breadth) that described the variance of niche across the measured environmental variables and 3) residual tolerance, variance not ex- plained. We used a Monte Carlo test to compare the observed marginality of specie versus the distribution of 10,000 random permutation values, under the null hypothesis that its environment does not affect the species occurrence. Permutation values lower than 500 were interpreted, as the species marginality index is statistically different from average environmental conditions available (P = 0.05). Using field data, we employed ArcGIS 9.3 [19] software to map the geographical distribution of Q. sideroxyla in the Sierra Fría. Records were used to test for uniform slope orientation of oaks on each population with a Rayleigh’s test (R) using circular statistics Oriana for Windows, version 1.03 [20] .

Basal area data was used to obtain the population structure of Q. sideroxyla in each site and each individual was classified in six stage classes. Plants up to 5 cm² in basal area represented class 1, ranging from 5.01 to 35 cm², class 2. For reproductive individuals, basal area intervals were 35.1 - 150 cm² for class 3, 150.01 - 400 cm² for class 4, 400 - 800 cm² for class 5, and >800 cm² for class 6. In addition, one-way Kruskal-Wallis ANOVA on ranks were tests and a Tukey multiple comparisons test were used to evaluate the frequency of individuals among sites and stage class.

In August 2007 we selected nine populations of Q. sideroxyla along its distributional range in the Sierra Fría (Figure 1). In each population eighteen adult individuals were sampled to estimate main parameters according to [21] . At one site, only six adults were found. Each individual was collected every 50 m along a 900 m transect where Q. sideroxyla was found. A total of 150 individuals of the species were collected. Samples were kept in plastic bags and immediately stored in liquid nitrogen, transported to the laboratory and stored at −70˚C.

Genomic DNA was extracted from 0.1 g of foliar tissue, using the Qiagen Plant Minikit according to the manufacturer’s instructions. Twelve primers were tested, six developed by [22] , and six by [23] . Only four (qu- ru-GA-0C11, quru-GA-0C19, quru-GA-E09 and quru-GA-1F07) were polymorphic for Q. sideroxyla.

PCR reactions were performed in 25 µl containing 1× buffer (500 mM KCl and 200 mM Tris-HCl, pH 8.4) (Invitrogen), 4 mM of MgCl₂ (Invitrogen), 10 ng of DNA template, 10 mM of each dNTP, (Pharmacia), 0.4 μM of reverse and forward primers, 2 μg of BSA and 0.5 U of Taq DNA polymerase (Invitrogen). Amplification reactions were carried out using a PTC-100 MJ Research thermal cycler. PCR cycling conditions were 5 min at 95˚C, 30 cycles of 94˚C for 10 s, annealing according to the different primer temperatures (Tm˚C) for 10 s and extension at 72˚C for 10 s, with a final 8 min extension at 72˚C. PCR products were analysed by capillary elec- trophoresis using an ABI PRISM-3100 Genetic Analyzer and genotypes were scored using GENESCAN version 3.5 (Applied Biosystems, Foster City, CA, USA).

GenAlEx version 6.5 beta 3 [24] was used to estimate number of alleles (Ao), number of private alleles (Ap), observed heterozygosis (HO), and expected heterozygosis (HE). GENEPOP version 4.1 [25] was used to test for deviation from Hardy-Weinberg equilibrium, according to Fisher’s procedure, with combined probabilities from exact tests by the Markov chain method with 10,000 dememorization steps and 1000 batches (Iteration per batch = 1000). Also, linkage disequilibrium was assayed between pairs of loci, we used likelihood-ratio test, and the empirical distribution was obtained by a permutation procedure according to [26] . MICRO-CHECKER software [27] was used to test the presence either of null alleles, stuttering or large allele dropout. We used the four loci for the regional-scale analyses in order to calculate null alleles in each population. All populations showed null allele and were coded as missing data. Also, an analysis of molecular variance (AMOVA) was done using the program GenoDive version 2.0b24 [28] . The significance of these values was tested using a non-parametric permutation according to [29] , and also gene flow (N_em) was estimated according to [30] .

Demographic parameters were analysed by the Pearson Correlative Coefficient (PCC) between demographic variables as the number of individuals in each population (N), number of individuals in each population on stage class (Ns) and genetic parameters as expected heterozigosity, observed heterozigosity, genetic flow (N_em), number of observed alleles (Ao) and number of private alleles (Ap), where positive values reflect direct inci- dence, negative values reflect inversely proportional relations and values close to zero reflects no relationship.

In the Sierra Fría, this species had a fragmented distribution and was only found in nine of the 29 sites sampled (Figure 1). Q. sideroxyla was distributed in complex damp slopes located mostly along the Northern orientation (R = 0.895, P = 0.005) and between 2471 (El Águila) and 2852 (Monte Grande) m.a.s.l. The Outlying Mean In- dex (OMI) suggested that Q. sideroxyla was significantly influenced by meteorological and geographical variables (P = 0.008), with lower tolerance index and residual tolerance values (Table 1). In the Outlying index, the first two axes account for 99.8% of data variation, with the annual maximum precipitation (TMaP) and the minimal temperatures in warmer quarters (MinTMiWC) accounting for the major variation on the first axis, and the annual minimal precipitation (TMiP) in rainy months being responsible for the principal variation on the second axis (Figure 2). Abundance of Q. sideroxyla was associated to sites with higher minimal and maximal temperatures than the average on the 29 samples (11 and 17.6 mm, respectively).

Total number of individuals ranged from 30 at Barranca Macitas to 145 at La Araña and La Ciénega (Figure 3). In general, the greatest number of individuals belonged to size stage class 1. Exceptions were El Pilar, Agua Escondida and El Águila. A population size structure analysis of Q. sideroxyla showed that four populations (Barranca Macitas, El Pinal, La Araña and La Ciénega) had inverted J distribution patterns, while the other five showed different patterns.

El Águila and El Pilar sites showed scarce evidence of recent recruitment (either clonal or by seeds), with a low number of plants in size stage classes 1, 2 and 3, whereas at Monte Grande and Agua Escondida, an equal number of individuals in all size stage classes were found. El Conejo presented a considerable large number of individuals (82) in size stage class 1, suggesting a successful recruitment process for this site, also with a considerable number of individuals (50) found in size stage class 4 (see Figure 3). Additionally, significant differences in the abundance within the size stage classes were observed among the sites (H = 8.65, P = 0.05). The only site with significant differences in the size stage class populations was Barranca Macitas, (H = 9.690, P = 0.001).

The four microsatellite loci used in this study generated a total of 84 alleles across the 149 individuals. The genetic diversity across the populations is shown in Table 2. The number of alleles (Ao) ranged from 24 in Barranca Macitas to 49 in El Conejo (average 38.6 ± 8.1). A total of 21 number of private alleles (Ap) were found, ranging from 0 at Barranca Macitas and to 5 at El Conejo, and the average number of alleles found was 2.33 ± 1.5. The global mean heterozygosity expected (HE) and heterozygosity observed (HO) values were 0.855 ± 0.009 and 0.733 ± 0.021, respectively. A significant deviation from the Hardy-Weinberg equilibrium (HWE) was observed for the entire population, except at Barranca Macitas and Monte Grande (Table 2). No linkage disequilibrium was found for any loci or in any site. The average estimated value of FIS was 0.11 (P = 0.001)

Inertia = total variability, OMI = outlying mean index (%), Tol = tolerance index (%), RTol = residual tolerance (%), ^*OMI, Values with asterisk represent the percentages of variability corresponding to a specific statistic, P = frequency based on number of random permutations (out of 1000) that yielded a higher value than the observed outlying mean index (P = 0.05).

Ao—the total number of alleles detected, Ap—number of private alleles, HO—observed heterozygosis with standard error, HE—expected heterozygosis with standard error, HWE—Hardy-Weinberg equilibrium, P—Level of significance, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001.

for all sites. An ANOVA showed that 99.0% of the genetic variation was contained within the individuals of the populations, while only 1.0% occurred among the population. The global fixation index, FST, was 0.010, which was highly significant (P = 0.001), and the gene flow (N_em) among populations was 4.43 ± 0.53.

Correlative analysis showed a major relationship between alleles observed in each population (Ao) and num- ber of individuals in each population (PCC = 0.55), number of individuals on category 4 (CCP = 0.56) and number of individuals on category five (CCP = 0.52).

Q. sideroxyla presented a fragmented distribution in Sierra Fría forest, associated with specific conditions of orientation and altitudinal ranges. This pattern has been reported for this species in northern regions of Mexico [12] , as well as for other species of oak [2] [31] , where topographic and climatic factors have limited the distri- bution of those species. Poulos et al. [12] and [32] have suggested that Q. sideroxyla has evolved to survive and reproduce under a narrow range of specific climatic conditions.

According to the outlaying index, the annual maximum precipitation (TmaP), the annual minimal precipita- tion (TmiP), and the minimal temperatures in the warmer quarter months (MinTMiWC) were the principal fac- tors affecting the abundance of Q. sideroxyla, suggesting an increased sensitivity, niche specialization and habi- tat selection in response to specific meteorological conditions. Similarly, [12] suggested that the species has specific climatological and physiological requirements, such as higher levels of photosynthesis, stomatal con- ductance, relative water content and water potential values, when compared with other Quercus species.

The climatic affinity, sensitivity and specificity of the Q. sideroxyla niche in Sierra Fría forest could be fac- tors that affect its distribution and its permanence. Permanence rate scenarios indicate that this species is one of the most vulnerable species of oak in Mexico, with 35% of its current distribution presenting susceptibility to decrement [6] . Changes at the local scale can generate a reduction in population size or even local extinction.

In all of the study sites, size stage class 1 (which is primarily the result of clonal growth through root sucker- ing) represented the regeneration of Q. sideroxyla and is likely due to the constant, intense grazing and logging that has occurred in the region since the 20th century. Demographic studies of Q. eduardii and Q. potosina in Sierra Fría forest support this assertion. In these cases, the clonal offspring remained connected for a long period of time to the parent tree, favouring the offspring, most likely by allowing the clones to recover from grazing and trampling [5] [16] .

However, the different shapes found among the static population size structures of Q. sideroxyla may reflect the history of different intensive regimens of logging and grazing over the past 80 years in the Sierra Fría. Bar- ranca Macitas, El Pinal, Agua Escondida and La Ciénega sites showed reverse J-shaped patterns that suggests a stable population with good recruitment and a low probability of death in the last size stage class [5] [33] , also indicates populations of stable growth, with lower intensities of logging and grazing [5] [34] .

This J-shape pattern was also found in the Q. eduardii and Q. potosina populations in disturbed (logging and grazing) and undisturbed sites in the Sierra Fría. In both species, the population’s growth rates were above or equal to unity for both types of sites, suggesting that the populations were stable [5] suggested the possibility that Q. eduardii can protect itself by adjusting its vital rates, resulting in fewer disturbances. This scenario ap- pears similar to what was observed for Barranca Macitas, El Pinal, Agua Escondida and La Ciénega sites.

However, in El Águila and El Pilar sites little evidence of recent recruitment (either clonal or by seed) was found, with a low number of plants in size stage classes 1, 2 and 3, only adult size class individuals create a de- mographic effect in which only the largest trees still exist, suggesting a more severe and historical human impact from logging and grazing. A similar scenario was found in Bertholletia excelsa by [34] , where high harvesting intensity over time led to a size class structure characterized by a large number of cohorts and little regeneration.

However, variation in the middle and large size classes of trees would be expected to depend upon the inten- sity of the disturbance, as suggested by [35] studies on Araucaria angustifolia, showing that the presence of small individuals depends on anthropogenic history factors, such as the recovery time since abandonment, or the expected impact of logging on the proportion of small individuals in a population. Our results actually support Souza’s findings, with similar patterns being found in Monte Grande and Agua Escondida. In these two sites, it appeared that the main regeneration event was clonal growth because contemporary logging and grazing activi- ties were reduced, or even absent, compared with historical records. This is contrasting to the El Conejo site, where logging is more intense now than in the past, and the middle size class is absent in the population’s size structure, which is a typical demographic effect produced by logging [36] [37] .

The average allele diversity obtained in the four loci used for Q. sideroxyla (Ao = 21 ± 5.65) was equal to or greater than those of the other oak species that were subjected to forest management, such as Q. suber (Ao = 13.1 ± 6.62) [38] and Q. petraea (Ao = 18.7 ± 5.65) [39] . The diversity of Q. sideroxyla was also equal to or greater than those species that had been submitted to cattle or agriculture, such as Q. humboldtii (Ao = 13 ± 8.5) [40] , and that of the complex of red oaks formed by Q. rubra, Q. shumardii and Q. palustris (Ao = 17.1 ± 4.93) [41] . Additionally, the species presented a high number of private alleles (Ap = 21), with important implications for Q. sideroxyla from a local evolutionary perspective. The high number of private alleles could be the result of introgressive hybridization [42] or the constant interchange among neighboring populations. Regardless of the source, [43] noted the importance of the private alleles as a potential reservoir with adaptive capability. In addi- tion to changes in unusual environmental conditions or anthropogenic stresses, if we consider the history of wood extraction [16] , the vulnerability of Q. sideroxyla to climatic change [6] and the high sensibility and niche specificity of the species found in the present study, the private alleles might play an important role in the per- sistence of the species at the local scale.

Correlation analysis showed a narrow dependence of observed alleles and population size. Observed alleles are very sensitive to population size in Quercus species, recent studies show that endemic species have a minor number of alleles than other species [44] [45] , also [21] , showed that number of alleles are very sensitive and increase with sample size. We found a positive correlation in observed alleles with the whole population size, and we propose that observed allele number could be an indirect estimator of population’s size useful in conser- vation programs.

The clonal regenerative capability of Q. sideroxyla allows the undisturbed preservation of genotypes, a phe- nomenon that was also found in Q. eduardii and Q. potosina in the study zone [5] [8] . These typical life histories could be responsible for the preservation of the genetic diversity of a species, despite the prolonged years of in- tensive extraction and the changes in soil use that have occurred in the Sierra Fría during the 20th century [16] [17] , and could have served as a damping capability against earlier disturbances at the local scale [8] [40] .

Heterozygosity values showed that the populations of Q. sideroxyla departed from the Hardy-Weinberg equi- librium, signifying that, for most sites (except Barranca Macitas and Monte Grande) a heterozygosis deficit were common (Table 2). Possible cause for this result could be inbreeding, as has been reported in other oak species in the Sierra Fría [10] .

The gravity seed dispersal system of Q. sideroxyla, common in Sierra Fría oaks [8] [9] , could be responsible for the formation of family groups. Family groups in oaks generally occur at short distances (less than 20 meters) from the progenitor tree [39] . For example, in Q. petraea [46] , family groups could be identified by their genetic similarity and the spatial distribution among individuals. While these family groups had no significant effect on the population’s genetic structure, heterozygosis deficit were detected. Heterozygosis deficit, combined with FIS value (0.11) in Q. sideroxyla indicated some degree of consanguinity, however, it is necessary to analyse the spatial genetic structure on a very fine scale to understand the distribution of family groups and clones that es- tablish patterns of non-random pollination between individuals.

The full differentiation index of the population’s genetic structure was low (FST = 0.010) and abundant gene flow (N_em = 4.43) is indicative of populations that have not been founded recently, but are remnant populations of a once more widespread and perhaps more continues distribution, a pattern also found in other Sierra Fría oaks [8] -[10] .