Cloning and evaluation of reference genes for quantitative real-time PCR analysis in Amorphophallus

Quantitative real-time reverse transcription PCR (RT-qPCR) has been widely used in the detection and quantification of gene expression levels because of its high accuracy, sensitivity, and reproducibility as well as its large dynamic range. However, the reliability and accuracy of RT-qPCR depends on accurate transcript normalization using stably expressed reference genes. Amorphophallus is a perennial plant with a high content of konjac glucomannan (KGM) in its corm. This crop has been used as a food source and as a traditional medicine for thousands of years. Without adequate knowledge of gene expression profiles, there has been no report of validated reference genes in Amorphophallus. In this study, nine genes that are usually used as reference genes in other crops were selected as candidate reference genes. These putative sequences of these genes Amorphophallus were cloned by the use of degenerate primers. The expression stability of each gene was assessed in different tissues and under two abiotic stresses (heat and waterlogging) in A. albus and A. konjac. Three distinct algorithms were used to evaluate the expression stability of the candidate reference genes. The results demonstrated that EF1-a, EIF4A, H3 and UBQ were the best reference genes under heat stress in Amorphophallus. Furthermore, EF1-a, EIF4A, TUB, and RP were the best reference genes in waterlogged conditions. By comparing different tissues from all samples, we determined that EF1-α, EIF4A, and CYP were stable in these sets. In addition, the suitability of these reference genes was confirmed by validating the expression of a gene encoding the small heat shock protein SHSP, which is related to heat stress in Amorphophallus. In sum, EF1-α and EIF4A were the two best reference genes for normalizing mRNA levels in different tissues and under various stress treatments, and we suggest using one of these genes in combination with 1 or 2 reference genes associated with different biological processes to normalize gene expression. Our results will provide researchers with appropriate reference genes for further gene expression quantification using RT-qPCR in Amorphophallus.

234 reference gene, followed also by EIF4A. Under waterlogging conditions, the results showed that EF1-a and 235 EIF4A were the best two reference genes in the two species. Additionally, UBQ and RP performed well across 236 different tissues in A. albus based on the results calculated by NormFinder. In the total set, EIF4A and EF1-a 237 showed remarkable expression stability in the two species of Amorphophallus. Overall, EIF4A and EF1-a 238 performed very well in all sets and were identified as the best two reference genes in 5 sets.
239 BestKeeper analysis 240 BestKeeper software was written by Pfaffl et al in 2004. The data were entered into the BestKeeper Excel file. 241 Then, BestKeeper calculated the standard deviation (SD) and coefficient of variance (CV) of each gene (Pfaffl 242 et al., 2004). The candidate reference gene with the lowest coefficient of variance and standard deviation 243 (CV±SD) was considered to be the best reference gene. Any reference gene with a SD＞1 was excluded because 244 gene expression was not consistent in all samples. The advantage of BestKeeper software is that it is not only 245 able to analyze the stability of reference genes, but it can also compare the expression levels of target genes.

246
The results of BestKeeper analysis were different from those of the other two programs which many . This may be because the principle of 248 this algorithm differs from that of the others. Lower CV values represent higher stability. EIF4A was the best 249 reference gene under heat stress in the two species of Amorphophallus, and ACTB and EF1-α ranked second in 250 A. albus and A. konjac, respectively. In the waterlogging sets, even though EIF4A performed well in A. albus, it 251 was identified as the worst reference gene in A. konjac. At the same time, EF1-α was one of the least stable 252 reference genes under waterlogging in both species of Amorphophallus. Across different tissues, CYP, EF1-α, 253 and EIF4A were ranked the same and were the best three genes in both species. In total, H3, TUB, and UBQ 254 were ranked above EF1-α and EIF4A, even though EF1-α and EIF4A were ranked as unstable in other sets and 255 were even the worst two genes in the waterlogging treatment of A. konjac. The stability of these two genes could 256 be better than others when SD≤1 was taken into consideration. In fact, this algorithm does not consider internal 257 differences between different plants. The calculated results are only references for selecting reference genes. The 258 ranking is shown in Table 7 However, we were concerned that selecting multiple genes that participate in related biological processes 336 may result in inaccurate results. geNorm algorithm works based on the assumption that the expression ratio of 337 two ideal internal control genes is identical in all samples, regardless of the experimental conditions or cell type. 338 This means that the absolute expression levels of these genes can change between conditions, but the ratio would 339 be maintained. Variations of the absolute levels of each of the genes would likely reflect technical variability. 340 This approach, however, only holds if the chosen genes belong to different functional classes; otherwise, one 341 may be simply scoring co-regulation and may incorrectly assume that two genes are stably expressed and are 342 appropriate reference genes, when in fact they may be responding to the treatment, but as they are part of the 343 same process, they may be responding coordinately. When Vandesompele et al. developed this algorithm, they 344 mentioned that special attention should be paid to selecting genes that belong to different functional classes, 345 which significantly reduces the chance that genes might be co-regulated . To avoid 346 the use of two reference genes that are related to the same biological process, we supposed that excluding one 347 of them and adding 1 or 2 reference genes related to different biological processes could effectively normalize 348 the relative expression levels of the target genes in Amorphophallus. To verify this conjecture, we excluded 349 EIF4A and repeated geNorm analysis, and analyzed the expression of the SHSP gene by designing different 350 normalization combinations without EIF4A in Amorphophallus under heat stress.

351
The results calculated by geNorm revealed that EF1-a was still one of the most stable reference genes. It 352 probably behaves as a stably expressed gene and it is not because it is being co-expressed with another gene due 353 to being part of the same biological process. In addition, the normalized result confirmed that the relative 354 expression of SHSP was not significantly different when EF1-a, EIF4A, H3, or UBQ was used as a single 355 reference gene in normalizing the gene expression profile under heat stress. Highly similar expression level of 356 SHSP normalized by EF1-a and EIF4A suggested the presence of the co-expression of them. Although our "best 357 genes" are functionally related, the third and fourth-ranked "good genes" also provided similar results compared 358 with the two "best genes", which further supported the accuracy of the results calculated by three algorithms and 359 the stability of the best four reference genes. The expression profiles of SHSP were in almost perfect agreement 360 when normalized by three combinations (EF1-a+H3, EF1-a+UBQ, and EF1-a+H3+UBQ). These results showed 361 that almost no difference was found when a single identified stable reference gene was used, and using one best 362 reference gene sometimes can obtain accurate result in RT-qPCR normalization. However, the result that 3 363 combinations had lower SS value probably indicate that the normalization by multiple reference genes is closer 364 to the actual expression. Furthermore, the strategy used to select combinations of reference genes also had impact 365 on the final results. The combination should avoid these co-regulated genes with biological processes and be 366 evaluated by geNorm. If selected candidate genes simply vary too much to be useful in a practical manner, 367 reliable results cannot be obtained irrespective of how many reference genes are selected.

CONCLUSION
369 In the present study, the expression of nine candidate reference genes in Amorphophallus under two major 370 stresses and across different tissues was compared and evaluated to identify stable reference genes for gene 371 expression studies. Among them, EF1-a and EIF4A appeared to be the two most appropriate reference genes. 372 Integrating the analyses by three algorithms with normalization of SHSP expression, we recognized that a single 373 reference gene may normalize the expression well under some conditions in RT-qPCR. But the result of using 374 multiple reference genes are more credible. It is indispensable to select reference genes in a practical manner 375 based on the specific experimental conditions and avoiding using multiple genes that participate in related 376 biological processes. These results will provide useful information to profile the gene expression of resistance 377 and quality-related in Amorphophallus.   Gene expression stability(M) and ranking of potential reference genes within different treatment groups as calculated by geNorm. Manuscript to be reviewed Gene expression stability(M) and ranking of potential reference genes within different treatment groupsas calculated by geNorm when EIF4A was excluded. Manuscript to be reviewed Determination of the optimal number of reference genes for normalization by pairwise variation determined by geNorm.