A sample of 40 S. gummosus individuals from Bahía Sargento (29.212369˚N, 112.202189˚W), and 30 individuals from Piedra Blanca (28.229661˚N, 113.159389˚W) was collected for population characterization. The first locality was chosen nearby the one from previous analysis for spatial clonal distribution by Molina Freaner and Clark-Tapia [14] , allocated on mainland Mexico; the second locality was intended to represent the Peninsula populations, which may be reproductively isolated from the Sonora ones since the Gulf of California permanently separates them. Each sampled individual was distanced on-site at least for 30 - 35 m from the rest, leaving the unsampled individuals in between. Assuming in a clumped clonal architecture, as found by Molina Freaner and Clark-Tapia [14] , each individual sample might belong to a different genet. The collected individuals describe two inverted semicircles to opposite sides of each coordinate previously specified. Sampled tissue fragments were taken from the stem chlorenquima to be stored in 96˚ ethanol on 2 mL tubes, and deposited at the tissue collection of the Laboratorio de Genética para la Conservación (CIBNOR, La Paz, BCS, Mexico). The source specimens were left living on-site, for which no voucher specimen was collected.

Genomic DNA was extracted using an optimized protocol of de la Cruz et al. [24] . A sample of 50 ng of RNA-free, high quality DNA from an individual collected at El Comitán (24.135385˚N, 110.42504˚W); voucher specimen HCIB27832 deposited at the Herbarium of the CIBNOR (La Paz, Baja California Sur, Mexico) was

used for shotgun sequencing by using a 454 GS-FLX Titanium instrument (Roche Applied Sciences, Indianapolis, Indiana, USA) at the UCSC Genome Sequencing Center (Santa Cruz, California, USA); other seven barcoded species were included in the same plate.

Msatcommander [25] was used to filter sequences among 454 reads containing perfect microsatellites with 2 - 6 bp repeat units, counting at least five repeats for di-nucleotides and four for tri-, tetra-, penta-, and exa-nucleo- tides. Primers were designed for filtered microsatellite hit results using Primer3 software [26] embedded in QDD [27] when hits fulfilled the following parameters: (1) 50˚C - 70˚C melting temperatures with a 2˚C maximum difference between paired primers; (2) PCR product of 90 - 320 bp in length; (3) GC content > 40%; and (4) A primer length between 17 and 27 nucleotides. Default parameter values were maintained in Primer3 for primer self-complementarities and complement between pairs. Favoring larger motifs (repetition numbers) and lower melting temperature differs between primers [28] , 40 primer pairs were chosen for PCR amplification screening and polymorphism test. The names for the loci were designated with the prefix “Sgum” and a consecutive number (1 - 40). Polymorphic loci are listed in Table 1, describing microsatellite motifs, PCR conditions, and fluorescent dye-labels (Applied Biosystems, Foster City, California, EUA) incorporated to forward primer sequences.

PCR reactions were carried out in 10 μL volume, using an iCycler thermocycler (BIORAD Laboratories, Hercules, California, USA), containing approximately 45 ng of genomic DNA, 1 × PCR buffer (20 mM Tris-HCl, pH 8.4, 50 mM KCl), 0.25 mM of each dNTP, 0.4 μM of each primer, 100 mM of BSA (New England Biolabs, Beverly, Massachusetts, USA), and 0.3 - 0.4 units of TaqDNA polymerase (Invitrogen, Carlsbad, California, USA), testing 1.5, 2.0 and 2.5 mM concentrations of MgCl₂; reactions for Sgum24 also included 5% glycerol. Forward primers for polymorphic loci were fluorescent-labeled (Table 1). Cycling conditions were conducted as

a. Fluorescent dye-labels (Applied Biosystems) on forward primers: NED (yellow), VIC (green), 6FAM (blue) and PET (red).

follows: 95˚C for 5 min as initial denaturation, 35 cycles at 94˚C for 60 s each, 35 s at the locus-specific annealing temperature (Table 1), extension at 72˚C for 15 s, and a final extension at 72˚C for 15 min. Amplification products were visualized in ABI 310-3730 PRISM^® (Life Technologies, Carlsbad, California, USA), using LIZ600 as size standard (Applied Biosystems).

Allele assignation was automated with GeneMapper 4.1 (Applied Biosystems). The minimum peak height acceptance was set to 100 Relative Fluorescence Units (RFU), and the minimum height ratio between peaks for heterozygotes to one third. All peaks assigned were manually checked for allele shape (Figure 2).

To know if the set of loci genotyped provides enough power to discriminate all MLG present in the sample, a Monte Carlo procedure, resampling 5000 permutations, incrementing loci number for each locality was implemented in GENCLONE 2.0 [23] [29] , and its corresponding results were plotted using the R package ggplot2 [30] . Clones were identified among samples by analyzing the MLG composition, calculating the genotypic richness index (R) and identifying one-allele difference MLG. Both identical MLG and one-allele difference MLG (23 individuals) were confirmed by PCR re-amplification and visualization as described above to reduce the probability of scoring errors. Of the individual genotypes, 10% were reassessed starting from DNA extraction to approach an error rate. The probability of identity (p_id) trough all loci (average probability that two unrelated individuals, drawn from the same randomly mating population, will by chance have the same multilocus genotype) was estimated using GenAlEx 6.5 [31] . Repeated MLG samples were assumed to correspond to ramets from the same individual, thus excluding repetitions for the following estimations.

Presence of null alleles, scoring errors due to stuttering and allelic drop out was explored implementing MICRO-CHEKER software [32] . Frequency for null alleles was computed through 5000 bootstrap resampling steps over loci for global fixation index (F_ST) values by the Expectation Maximization algorithm [33] with FreeNA software [34] . Effective alleles number (A_E) and inbreeding coefficient (F_IS) were estimated with standard methods implemented in GenAlEx 6.5 [31] . Expected and observed heterozygosity (H_E and H_O, respectively), deviations from the Hardy-Weinberg equilibrium, and linkage disequilibrium between markers, were calculated or tested using default parameters in GENEPOP 4.2 [35] . Benjamini & Hotchberg [36] False Discovery Rate correction was used for multiple hypothesis tests.

Genomic sequencing has been chosen over other techniques for microsatellite markers development due to the high sequencing throughput and the possibility of selecting microsatellites and motif length. The resulting number of markers and the polymorphic loci proportion are expected to be higher with this technique [37] . Unfortunately, no previous information of any microsatellite development technique for neither the Stenocereus gummosus species nor genus is available to compare against the results reported here.

The 454 pyrosequence reaction resulted in 42,554 reads with a mean length of 311.11 nucleotides, where 1539 (3.6%) contained a microsatellite motif, regardless of their suitability for primer design. A total of 179 reads (11.6% of microsatellite loci found) were suitable for primer design. Out of the 40 chosen loci for PCR and polymorphism screening, 28 were monomorphic, 10 were polymorphic, and 2 exhibited inconsistent amplification patterns. Only polymorphic loci with more than 50% of the genotypes successfully recorded (Table 1) were used for population characterization. The polymorphic loci proportion (30%) was greater than reported for genomic libraries (15% and 21%, Table 2), but lower than enriched genomic libraries (53%, 91% and 70%, Table 2) and cross amplification (57%, Table 2); these differences may rely on the species reproduction system, demographic issues, mutation rates on microsatellite loci, specific genomic composition, or even on the markers isolation technique implemented [37] [38] .

Micro-Checker software [32] detected homozygote excess in 4 loci for Piedra Blanca locality (Sgum05, Sgum12, Sgum31 and Sgum36) and in 3 loci for Bahía Sargento (Sgum05, Sgum31 and Sgum36), scoring error due to stuttering were only predicted to happen in both localities for Sgum05 and Sgum31, and no signal of allelic

dropout was detected. The null alleles frequencies calculated for two loci within Piedra Blanca (0.33 for Sgum05 and 0.27 for Sgum31), and one for Bahía Sargento (0.21 for Sgum05), are likely to hinder population differentiation analysis with F_ST, and may overestimate F_IS [34] ; the rest of the calculated frequencies were left

a Characterized samples from Hughes et al. [42] ; b from Hardesty et al. [41] ; c from Solórzano et al. [44] ; d from Pérez et al. [43] ; e from Otero- Arnaiz et al. [22] ; f wild populations from Parra et al. [21] .

aside since they were very small (mean frequency of 0.069 ± 0.02 S.E.). When PCR amplification and product visualization were repeated, 5 alleles assignment were modified for the loci Sgum05, Sgum06 and Sgum24 (error ratio of 0.018 accounting for 270 PCR reactions repeated); these scoring errors were always identified as mistakes at the electrophoresis peak reading for allele size assignment, which allowed improving the automated analysis. An increment on sample size [39] , or a PCR primers redesign based on individuals’ sequences for those loci [40] , could be useful to avoid bias due to null alleles on further population genetics studies. In this case, for which the goal was to determine clonal membership, achieving precision and reproducibility on genotyping lets know that the loci are useful, provided that null alleles and stuttering behave evenly at the population level. Significant linkage disequilibrium was detected between Sgum05 and Sgum31 when analyzed globally over both localities individuals. This condition persisted only in Piedra Blanca when sample was analyzed by locality.

The number of alleles per locus (A) among polymorphic loci ranged from 2 - 7 (Table 3), H_O ranged from 0.000 (Sgum25 in Piedra Blanca, and Sgum05 in Bahía Sargento) to the maximums of 0.552 and 0.556 (Sgum06), and H_E from 0.034 (Sgum39) and 0.055 (Sgum25) to 0.673 (Sgum36) and 0.557 (Sgum06) (Table 3), with mean values of 0.221 and 0.234 for H_O and 0.408 and 0.306 for H_E, in Piedra Blanca and Bahía Sargento respectively. The mean F_ST was 0.082 ± 0.39 S.E. Significant deviations from Hardy-Weinberg Equilibrium (HWE) were detected in 4 loci in Piedra Blanca (Sgum05, Sgum12, Sgum31 and Sgum36), and three in Bahía Sargento (Sgum05, Sgum31 and Sgum36); deviations are due to heterozygote deficiency in all cases (Table 3), but this departure from HWE may as well be null alleles artifact, as predicted by Micro-Checker software.

Both diversity indexes, mean allele number for polymorphic loci (A = 3.78 ± 0.062 SE), and expected and observed heterozygosity (H_O = 0.228 ± 0.047 SE, H_E = 0.357 ± 0.048 SE), reached the lowest value when compared to other cacti (Table 2), having mean values (±SD) of A = 5.05 ± 1.84, H_O = 0.461 ± 0.153, and H_E = 0.559 ± 0.138 [21] [22] [41] - [44] . These low diversity indexes are intuitively explained in terms of inbreeding coefficient on a low effective population size (N_e) scenario [45] , coherent to the F_IS of 0.608 ± 0.025 SE previously reported [13] , albeit inconsistent with the Molina Freaner and Clark-Tapia [14] conclusion, where sexual recruitment is found as more frequent than clonal growth. The average F_IS here reported (0.341 ± 0.122 SE) may be less biased by selection than allozyme results [46] , but null alleles can still be artificially increasing it in a small magnitude in respect to the actual inbreeding [47] . Though attenuated, the inbreeding coefficient is yet remarkable, so that reproduction system

a. Significant heterozygote deficiency after Benjamini and Hotchberg [36] . False Discovery Rate correction (p < 0.05) is indicated in bold numbers.

aspects like pollination syndrome, mating among relatives and the possibility of geitonogamy, cannot be discarded as causes for these low diversity indices.

The MLG resampling procedure shown that 7 loci were enough to determine clonal membership for Piedra Blanca locality (Figure 3(a)), but 8 or 9 loci might be required in the case of Bahía Sargento (Figure 3(b)); the p_id values of each locality accounting the 9 loci were 2 × 10⁻⁴ and 2.6 × 10⁻³, respectively. Piedra Blanca accounted 29 unique genotypes with R = 0.966, and 36 unique genotypes were detected for Bahía Sargento with R = 0.897. These values slightly increase the ones found by Molina Freaner and Clark Tapia [14] with allozyme markers, where the maximum was R = 0.959 for a locality nearby Bahía Sargento from this study. One-allele difference MLG within locality were found only for Bahía Sargento, with 12 individuals into at least 2 putative multilocus lineages (clonal ramets carrying a somatic mutation identified as one-allele difference) [23] , where 2 of the 3 repeated MLG of that locality are members. This finding provides a scenario that is supposed to be eliminated by collecting individuals distanced by at least 30 meters, in which ramets are now somehow able to spread away from the source plant, farther than it has been previously found [13] .

Identical MLG of Lophocereus schottii individuals, distanced in field by 200 or more meters, have been defined as long-distance transportation events, where detached stems are dispersed downstream by floodwaters [7] . This clone dispersion mechanism is likely to occur along the distribution range of S. gummosus (shared with L. schotti), species mostly found on alluvial, arroyo margins and rocky hillsides [8] [12] . The latter results in an intermingled distribution of the clones and a subsequent increase of the genetic neighborhood [7] , where ramets disperse not only to the space surrounding the source specimen but, in this case, also for at least 30 meters away. Intermingled clonal architecture is a characteristic that gives a clonal plant the capability to spread more quickly [48] . It has been proposed that L. schottii and S. gummosus populations along the Baja California Peninsula have occupied their current distribution range through recent colonization events, reaching such large areas [13] [49] .

This scenario implies that no discrimination can be made about the effect of geitonogamy from that of endogamy, because the genetic neighborhood size was not exceeded by the sampling strategy followed [17] [19] . Though the impact of geitonogamy is expected to be less severe as clones dispersion capability increases on insect pollinated plants [16] , it may increment the probability of mating among relatives as it increases neighborhood size.

A population analysis of wider geographical range, employing these markers, might be useful to enhance accuracy for linkage disequilibrium tests. These 9 isolated microsatellite loci are the first identified and characte-

rized ones for the genus Stenocereus. The availability of these molecular tools allows assessing questions integrating population genetics, ecology, conservation, and reproductive biology for S. gummosus and for other species phylogenetically related.