*2.2. TC-DNA Extraction and PCR Amplification*

The TC-DNA was extracted from each five replicates of rhizosphere (0.5 g of each) from the three sweet potato genotypes (IPB-149, IPB-137, and IPB-052), and from both sampling times (t1 and t2). The Fast DNA Spin Kit for soil (Qbiogene, BIO 101 Systems, Carlsbad, CA, USA) was used according to the manufacturer's instructions. TC-DNA preparations were used in PCR reactions in order to amplify the *alp* gene (alkaline phosphatase coding gene) and *nifH* gene (nitrogenase coding gene).

The PCR conditions for *alp* gene amplification were performed as described by Sakurai, et al. [27]. The primers ALPS-F730/ALPS-R1101 were used in the PCR reactions and the reverse primer was added with a GC clamp. The reaction conditions were those previously described in Sakurai, et al. [27]. Fragments of the *nifH* gene were amplified using a nested-PCR approach. Briefly, the primers FGPH19 [28] and PolR [29] were used for the first-round of the PCR. The first-round PCR conditions were those described by Monteiro, et al. [30]. A 1:100 dilution of the first-round PCR product was used as a template for the second-round with the primers PolF/AQER using the same reaction conditions as described by Poly, et al. [29]. The PCR products were visualized by electrophoresis and then stored at −20 ◦C until DGGE analyses.

#### *2.3. DGGE and Statistical Analyses*

The INGENYphorU-2 system (INGENY International BV, Middelburg, The Netherlands) was used for DGGE. The PCR products from both genes (*alp* and *nifH*) and a standard bacterial marker (previously described by Heuer, et al. [31]) were loaded directly to DGGE. The denaturing gradient of urea and formamide varied in the range 40%–60% (the *alp* gene), and 40%–70% (the *nifH* gene). The electrophoresis conditions were performed as described by Sakurai, et al. [27] and Monteiro, et al. [30] for the *alp* and the *nifH* genes, respectively. After electrophoresis, the DGGEs were stained with SYBR Green I and visualized using a STORM apparatus (Amersham Pharmacia Biotech, Munich, Germany). The unweighted pair group with average linkages (UPGMA) was used for cluster analysis. The software BioNumerics 5.0 version (Applied Mathematics, Kortrijk, Belgium) was used to construct dendrograms based on Pearson similarity indices. Significant differences (*p* ≤ 0.05) between DGGE profiles were determined by permutation tests based on pairwise Pearson correlation indices using the PERMTEST software [32]. In addition, quantitative matrices generated from the DGGE lanes based on Dice correlation index were exported to PAST software [33] for principal component analysis (PCA).
