AbstractThe seeds of a soybean cultivar Zhonghuang 18 were subjected to accelerated aging for 0 （population G01）， 112 （population G02）， 154 （population G03） and 196 d （population G04）， whose germination percentage was found to be 98.0%， 95.0%， 81.0%， and 79.0%， respectively. Then， the four populations were regenerated twice in the field. The first descendant populations were marked as G11， G12， G13 and G14， and the second were marked as G21， G22， G23 and G24， respectively. The genetic variation between the control population （G01） and the experimental populations was analyzed using 12 AFLP primer combinations. The results showed that there was no significant difference in genetic similarity between the 11 experimental populations and the control population G01. The genetic similarity between population G24 and G01 was still as high as 0.933 3， indicating that the F2 generation of the population whose germination percentage was only 79.0% still had a high genetic similarity to the control population. The results of ttests revealed that the populations G11， G21， G12 and G22 showed no significant difference from the control population G01 in effective number of alleles per locus （Ae）， genetic diversity index （H） and Shannonюs diversity index （I）， while these indices of populations G03， G04， G13， G14， G23 and G24 were significantly reduced. ┶2 tests indicated that the populations G11 and G21 showed little difference， and the populations G02， G03， G04， G12， G13， G14， G22， G23， and G24 showed great difference in allele frequency distribution from the control population G01， and the difference was greater when the seed viability was lower. Compared with the control population G01， the number of rare alleles （Nr） of the populations G02， G11， G21， G12 and G22 showed no significant difference， while that of the populations G03， G04， G13， G14， G23 and G24 declined obviously. These results revealed that compared with the control population， the genetic diversity and Nr for the descendant populations of the populations with 98.0% and 95.0% germination percentages did not change significantly， but declined greatly for the descendant populations of the populations with 81.0.% and 79.0% percentages. The results suggested that the decline in seed viability has a greater impact than the number of generations on genetic structure of soybean germplasm. So， it is suggested that soybean seed with an initial germination percentage of 98.0% should be regenerated before its germination percentage declines to 81.0%. 　　Key wordsSoybean; Seed viability; Number of generations; AFLP; Genetic integrity
　　At present， lowtemperature genebanks are the main facilities for conserving crop germplasm resources[1]. By 2011， more than 397 000 germplasm accessions （including 25 000 accessions of cultivated soybean and 8 000 accessions of wild soybean） have been deposited in the longterm storage of China National GeneBank （CNGB）， which is the worldюs second largest genebank in terms of plant collections[2]. Low temperature can prolong the storage life of seeds， but cannot prevent the decline of seed viability. The germination percentage of carrot seeds and lettuce seeds that had been stored in CNGB for 10 to 12 years at -18→decreased by 14% and 12%， respectively[3]. According to the U.S. National Plant Germplasm System （NPGS， the worldюs largest national genebank networks）， the germination percentage of 42 000 seed accessions representing 276 species decreased from 87.5% to 58.0% after being preserved at 5 and -18→for 16 to 81 years. Among them， the germination percentage of 3 635 soybean germplasms dropped from 92% to 21% during 36 years[4]. The study of Qiu et al.[5]， showed that 36 out of 68 soybean accessions from the mediumterm storage at 4 to 10→failed to germinate in the field. It is necessary to reproduce the stock germplasms periodically[6]， and maintenance of genetic integrity is crucial during this process[7].
　　Factors affecting genetic integrity during germplasm regeneration include viability decline， number of generations， population size， place， selection pressure， competition， plant density， pollination， harvesting methods， etc.[8-16]， among which， viability decline is particularly important， and has been studied from the aspects of chromosome aberrations， point mutations， chlorophyll mutations and DNA markers. For example， by observing the chromosomes in root tip cells of the seedlings grown from artificially aged （32 →/12% moisture content and 38 →/18% moisture content） barley seeds and their progeny， Murata et al.[17]found that the chromosome aberrations induced by seed aging could not be transmitted to the next generation. The studies in barley and pea showed that point mutations increased significantly with the decrease in viability of aged seeds[18]. By inducing chlorophylldeficient mutations in barley with Xray， Roberts[19]found that aging treatment that leads to 50% loss of seed viability is comparable to 10 000 roentgens of Xray in inducing chlorophylldeficient mutations. By analyzing the genetic diversity in maize germplasms with different germination percentage with SSR markers， Zhang et al.[20]found that accelerated aging reduced the genetic diversity and changed the genetic integrity of maize. 　　The number of generations also has an impact on genetic integrity during seed reproduction. The study of Tao et al.[21]reported that after eight regeneration cycles， the dominant varieties of a heterogeneous population consisting of two local durum wheat varieties changed from durum wheat varieties to common wheat varieties. By detecting the genetic integrity of three local cabbage varieties that had been regenerated twice， Soengas et al.[22]found that the genetic structure and the allele frequency per locus of the secondgeneration population were significantly different from those of original population. By detecting genetic integrity of 30 soybean germplasms using SSR markers， Xia et al.[23]found that after being generated for several times， the allele frequency of 25 germplasms did not change， while that of five germplasm did. Zhang et al.[24]believed that consecutive regeneration or biological mixing might increase the risk of losing rare alleles. Therefore， reducing the cycles of regeneration and avoiding biological mixing are important to maintain the genetic integrity of soybean germplasm[24].
　　Amplified fragment length polymorphism （AFLP） technique is a highly polymorphic， sensitive and reliable method for detecting polymorphisms in DNA， requiring less template DNA and without prior information of the DNA sequence[25]. It has been extensively used in genetics research， DNA fingerprinting， genetic integrity evaluation， etc. For example， .in the study of Mughand et al[26]， AFLP markers were used to study the genetic diversity， Mendelian inheritance and nearisogenic lines of soybean. Hu et al[27]evaluated the genetic integrity of ultra dry seed of wheat with AFLP markers. In this study， AFLP technique was adopted to analyze and evaluate the influence of viability decline and the number of generations on the genetic integrity of a soybean cultivar Zhonghuang 18. The results are expected to provide theoretical and practical basis for preservation and regeneration of soybean germplasm.
　　Materials and Methods
　　Materials
　　The seeds of Zhonghuang 18 were provided by the Institute of Crop Sciences， Chinese Academy of Agricultural Sciences， sown and harvested in the fields in 2001， with moisture content 8%-9%. In February 2002， the seeds were sealed in aluminum foil bags， subjected to accelerated aging at （40÷2）→ for 0 （population G01）， 112 （population G02）， 154 （population G03） and 196 d （population G04）. Population G01 was considered as the control （CK）. The seeds of the four populations were planted separately in the field in 2003， and their firstgeneration populations were marked as G11， G12， G13， and G14， correspondingly， which were then regenerated in 2006， to obtain the secondgeneration populations G21， G22， G23 and G24， respectively （Table 1）. The field trials were carried out in the Tishang Experimental Station of Institute of Cereal and Oil Crops of Hebei Academy of Agricultural and Forestry Sciences in Shijiazhuang. The populations were arranged in the fields in a random block design with 3 replicates and 200 seeds in each replicate. Seed regeneration was carried out by referring to Technical Regulation on Characterization and Documentation for Crop Germplasm Resources[28]. Bulk harvesting was performed in this study. All harvested seeds were airdried and stored at 4→till analysis. 　　Determination of laboratory germination percentage
　　By referring to International Rules for Seed Testing[29]， the seeds were sown on a germination bed consisting of two layers of filter paper and at 25 →. Two replicates were prepared for each population. The number of germinated seeds was counted every day， germination potential was calculated on Day 3， and germination percentage was calculated on Day 4.
　　DNA extraction and detection
　　Two hundred seeds were randomly selected from each population， planted in vermiculite and cultured in an incubator with daily cycles of 14 h at 30→in light and 10 h at 20→in darkness， and at a relative humidity of 80%. Sixty seedlings at threeleaf stage were selected at random from each population， and DNA was extracted from each seedling using SDS method. The resulting DNA samples were diluted to 20 ng/l with 1 ≠ TE， and stored at 4→till analysis.
　　AFLP analysis
　　AFLP analysis was performed following the method of Tian et al.[30]. Twelve previously reported primer combinations for soybean AFLP analysis were used in this study， they were： ECGA/MACT， ECGA/MATT， EGGA/MATT， EACT/MCGA， ECGA/MCGG， EGGC/MCGG， ECGA/MCTC， EACT/MGGA， EAGG/MGGA， ECGA/MGGA， E GGC/MGGA and ETGA/MGGA[31]. EcoR I， Mse I restriction enzymes， T4DNA ligase， Taq DNA polymerase and pBR322 DNA/Msp I DNA ladder marker were all purchased from New England Biolabs （Beijing） Ltd. EcoR I， Mse I adapters and AFLP primers were synthesized by Beijing Sunbiotech Co. Ltd. （Table 2）.
　　Agricultural Biotechnology2019
　　Enzyme digestion
　　Double digestion of DNA was carried out in a mixture of 20 l containing 4.0 l of 100 ng/l DNA， 0.5 l of 20 U/l EcoR I， 0.5 l of 10 U/l Mse I， 2.0 l of 10≠ NE Buffer （EcoR I）， 0.2 l of 100≠ BSA and 12.8 l of ddH2O. The mixture was incubated in a water bath at 37→for 3 h， and then in a water bath at 65→for 15 min and finally on ice.
　　Ligation reaction
　　Ligation reaction was performed in a mixture of 10.0 l containing 1.0 l of 50 pmol/l EcoR I adapters， 1.0 l of 50 pmol/l Mse I adapters， 0.25 l of 400 U/l T4 DNA ligase， 1.0 l of T4 DNA ligase Buffer， and 6.75 l of ddH2O. The mixture was incubated at 16→for at least 3 h and then stored at -20 →.
　　Preamplification
　　The preamplification mixture contained 2.0 l of above ligation mixture， 2.0 l of 50 ng/l primer M00， 2.0 l of 50 ng/l primer E00， 2.0 l of 10≠ PCR Buffer， 1.6 l of 2.5 mmol/L dNTPs， 0.24 l of 2.5 U/l Taq polymerase and 10.16 l of ddH2O. The reaction was carried out in a PTC100 Peltier Thermal Cycler （MJ Research， USA）， with a procedure of consisting of predenaturation at 95→for 1 min; 35 cycles of denaturation at 95→for 30 s， annealing at 56→for 30 s， extension at 72→for 1 min; and a final extension at 72→for 10 min. The finished preamplification product was diluted 20fold with ddH2O and stored at -20 →. 　　Selective amplification
　　In detail， 5.0 l of the diluted preamplification product was mixed with 2.0 l of 50 ng/l EcoR I， 2.0 l of 50 ng/l Mse I， 2.0 l of 10≠ buffer， 1.8 l of 2.5 mmol/L dNTPs， 0.4 l of 2.5 U/l Taq polymerase， and 7.8 l of ddH2O. The reaction was carried out in a PTC100 Peltier Thermal Cycler （MJ Research， USA）， with a procedure of consisting of predenaturation at 95→for 2 min; 12 cycles of denaturation at 95→for 30 s， annealing for 30 s （the annealing temperature decreasing from 65→by 0.7→per cycle）， extension at 72→for 1 min; 30 cycles of denaturation at 95→for 30 s， annealing at 56→for 30 s， extension at 72→for 1 min; and a final extension at 72→for 10 min.
　　Analysis of PCR products
　　To detect the quality of PCR products obtained above， 4.0 l of each PCR product was mixed with 2.0 l of loading buffer， denatured at 95→for 5 min， cooled on ice， and run on a 6% polyacrylamide gel for 1 h at 85 W before silver staining[32].
　　Data statistics and analysis
　　The DNA banding patterns were converted to data matrix， with 1 for the presence， 0 for the absence and 9 for the missing of bands at the same mobility. And the length of the amplified DNA fragments was estimated with reference to DNA marker I and pBR322 DNA/Msp I DNA Ladder marker. The number of polymorphic loci （K）， the percentage of polymorphic loci （P）， the effective number of alleles per locus （Ae）， genetic diversity index （H） and Shannonюs diversity index （I） of each population were calculated using POPGENE Ver. 1.31[33]. Allele frequencies at each locus of the 12 populations were calculated using Powermarker Ver. 3.25. The genetic diversity parameters of experimental populations and the control population were subjected to paired t test using SAS V9.0， allele frequencies were subjected to ┶2 test. UPGMA （unweighted pair group method with arithmetic mean） cluster analysis based on Neiюs genetic identity among populations was performed using NTSYSpc 2.1 software[34]. The differences in rare allele number and other parameters were compared in Excel.
　　Results and Analysis
　　Analysis of population genetic structure
　　A total of 204 clearly distinguishable bands were amplified from the 12 soybean populations using the 12 AFLP primer pairs， 10 to 27 bands were generated by using each of the primer pairs. The average number of bands generated by one primer pair was 17. Fig. 1 shows the AFLP electropherogram of the control population G01 generated using the primer pair EGGC/MGGA. 　　As shown in Table 3， the genetic diversity parameters of the populations G02， G03 and G04 with 95.0%， 81.0% and 79.0% germination percentages were all lower than those of the control population G01， and the populations that had been subjected to longer duration of aging had lower seed viability and genetic diversity parameters. The genetic diversity parameters of populations G12， G13， G14， G22， G23 and G24， which were regenerated from populations G02， G03 and G04， were all lower than those of the control population G01. The t tests on Ae， H and I showed that the populations G03 and G04 with 81.0% and 79.0% germination percentages were significantly or very significantly different from the control population G01 in Ae， H and I. The genetic diversity parameters of populations G11， G21， G12， G22 which were regenerated from populations G01 and G02， showed no significant difference from those of the control population G01， indicating that the populations with 98.0% and 95.0% germination percentages maintain genetic diversity better than the control population during the two cycles of regeneration. The three indices Ae， H and I of populations G13， G14， G23 and G24 were significantly or very significantly different from those of the control population G01， indicating that the genetic composition of the populations whose germination percentage had dropped to 81.0% and 79.0% changed greatly during their regeneration， compared with the control The above results indicated that viability decline has a greater impact than number of generations on the genetic structure of soybean germplasm.
　　Analysis on allele frequency differences
　　As could be seen from Table 4， compared with the control population G01， the number of loci with significant （P<0.05） or highly significant （P<0.01） difference in allele frequency in populations G02， G03， and G04 with 95.0%， 81.0%， and 79.0% germination percentages increased sharply with the decline in seed viability. Among the three populations， population G04 had the most loci in significant difference level and the most loci in highly significant difference level （72 and 51）， followed by population G03 （49 and 28） and population G02 （47 and 28）， respectively， indicating that viability decline significantly affected allele frequency distribution within soybean germplasm populations. Compared with the control population G01， the numbers of loci in significant difference level and in highly significant difference level of populations G11 and G21 were very small， indicating that the allele frequency at each locus for the population with a germination percentage of 98.0% changed little after two cycles of regeneration. The numbers of loci in significant difference level and in highly significant difference level of populations G12 （60 and 37）， G13 （51 and 35）， G14 （78 and 62） and G22 （63 and 45）， G23 （63 and 47）， andG24 （89 and 71） were much more than those of populations G11 and G21， and the numbers for the secondgeneration populations were bigger than those for their corresponding firstgeneration populations， indicating that allele frequencies at each loci of populations with 95.0%， 81.0% and 79.0% germination percentages and their progeny populations were greatly different from those of the control population， and the difference was more significant when the seed viability was lower. The above results suggested that viability decline has a greater impact than number of generations on allele frequency distribution in soybean germplasm. 　　Analysis on genetic similarity
　　The genetic similarity between each experimental population and the control population was calculated （Table 5）. Among all the populations， G11 showed the highest genetic similarity （0.996 7）， followed by population G21 （0.993 5）， and population G24 had the lowest genetic similarity （0.933 3） to the control population. As shown in Fig. 2， the genetic similarity of population G24 was the lowest， but its absolute value was relatively high， and the population was clustered together with the control population G01 at 0.954 7， indicating that the secondgeneration population G24 of population G04 with a germination percentage of 79.0% still had high genetic similarity to the control population. Population G11 had the highest genetic similarity and the closest genetic distance to the control population G01， followed by population G21. Population G11 was regenerated from G01， and population G21 was then regenerated from population G11， indicating that the genetic information of soybean germplasm with an initial germination percentage of 98.0% changed little after being regenerated twice. Compared with populations G11 and G12， populations G02， G03 and G04 with 95.0%， 81.0% and 79.0% germination percentages had relatively low genetic similarity to the control population G01， indicating that viability decline had a greater impact than number of generations on genetic similarity of soybean germplasm. Populations G12， G13， and G14 had lower similarity than their parental populations G02， G03 and G04， and the genetic similarity of their secondgeneration populations was lower than that of the firstgeneration populations， suggesting that the genetic similarity of the progeny populations of G02， G03 and G04 with 95.0%， 81.0% and 79.0% germination percentages decreased with decline in seed viability and with increase in regeneration times.
　　Analysis on changes in number of rare alleles （P<0.05）
　　Rare alleles， which account for a small proportion of total alleles， are likely to be lost or increased due to seed aging， regeneration and population size， which leads to the changes in genetic diversity of the population. Therefore， changes in rare alleles within a population can be used as an indicator for analyzing germplasm genetic integrity. As shown in Table 6， the number of rare alleles of populations G03 and G04 with 81.0% and 79.0% germination percentages was significantly lower than that of the control population G01. The number of rare alleles shared by the experimental populations and the control population also decreased with decline in seed viability， and the number of increased/lost rare alleles changed in a similar pattern， suggesting that viability decline significantly affected and for most populations， decreased the number of rare alleles within a soybean germplasm population. Compared with the control population G01， the populations G11， G12， G21 and G22 which were regenerated from the populations G01 and G02 showed little change in the above three indicators， suggesting that the rare alleles were well maintained in the progeny of the populations with 98.0% and 95.0% germination percentages. The populations G13， G14， G23 and G24， which were regenerated from the populations G03 and G04， showed a significant decrease in the number of rare alleles compared with the control population G01， but a significant increase in the number of lost/increased rare allele， indicating that the progeny of the populations with 81.0% and 79.0% germination percentages were significantly from the control population in number of rare alleles. All the results proved that viability decline had a greater impact than the number of generations on changes in rare alleles in soybean germplasm. 　　Discussion
　　Effects of viability decline and number of generations on population genetic structure and genetic similarity
　　By analyzing the prolamin band profile in heat variety Sadovo l， Stoyanova[35]noticed that there were four types of prolamin band profile A， B， C and D in this wheat variety. Then the seeds were subjected to accelerated aging， which reduced the germination percentage to below 30%. In its offspring generation Type B profile was observed in most individuals， Type D in a few individuals， while Type C and Type A were not detected in any individuals. The study of Parzies et al.[36]showed that the germination percentage of two barley landraces that had been stored for 10， 40， and 72 years decreased over the duration of storage， while the genetic diversity of allozymes， the alleles and the average polymorphic loci at each loci also decreased significantly， and the genetic differences between the two landraces increased dramatically. The study of Roos[37]showed that six pea varieties in a population consisting of eight pea varieties disappeared after the population underwent 15 cycles of aging and regeneration. Using nine SSR pair combinations， Borner et al.[38]analyzed the genetic integrity of eight wheat germplasms that had been stored in Gatersleben genebank and regenerated for 24 times， and noticed genetic drift in one germplasm. All the above studies indicated that viability decline and number of generations have an impact on the genetic integrity of plant germplasm.
　　Our results showed that the experimental populations and the control population shared high genetic similarity， which may be due to that soybean is a typical selfpollinated crop. In addition， Zhonghuang 18 is a bred cultivar with homozygous genotype， and thus its individuals have the same genetic basis. So the probability that soybean is contaminated by exotic pollens during regeneration is extremely remote. These reasons explain why the all the experimental population have high genetic similarity to the control population. After further analysis， it was found that the number of polymorphic loci， percentage of polymorphic loci， effective number of alleles per locus， genetic diversity parameters and Shannonюs diversity index of the populations with 81.0% and 79.0% germination percentages and their progeny populations were significantly lower than those of the control population， while the allele frequencies per locus and the numbers of loci in significant or highly significant difference levels were much higher. The lower the level of seed viability， the greater the decline in genetic diversity parameters， and the more the loci in significant or highly significant difference levels. This seems inconsistent with the analysis of genetic similarity， but the phenomenon may be explained by the changes in rare alleles. 　　Effect of viability decline and number of generations on number of rare alleles
　　Marshall and Brown[39]classified the alleles according to their average frequency in a population as common （P>0.05） or rare （P<0.05）. According to their frequency and distribution， the two categories of alleles can be further grouped into four subcategories： ↘ common， widely distributed alleles; ? rare， widely distributed alleles; ? common， locally distributed alleles ? rare， locally distributed alleles. Rare alleles （P<0.05） account for a very small proportion of the total alleles within a population， but greatly increase the genetic diversity of the population. Many factors such as seed aging， regeneration， small population size and seed harvesting methods， can result in the loss of rare alleles[8， 40]. Changes in rare alleles has little influence on genetic similarity between populations， but significantly change the genetic diversity within populations. Our results showed that some rare alleles were lost in ageingtreated soybean population with 81.0% and 79.0% germination percentages， leading to decline in genetic diversity within populations. And the lost rare alleles could not be transmitted to next generations， which led to significantly decreased genetic diversity of their descendant populations， compared with the control population. Zhonghuang 18 is a bred soybean cultivar， so its rare alleles can be grouped to subcategory ?： rare， widely distributed. Since even these rare， widely distributed alleles can be lost during the decline of viability， it is speculated that the rare， locally distributed alleles in soybean landraces are more likely to be lost when seed viability decreases.
　　Germination percentage standard for regeneration of soybean germplasm
　　By analyzing four rye materials using AFLP markers and microsatellite markers， Chwedorzewska et al.[41-42]found that germination percentage has a greater impact than the number of generations on genomic DNA. The study of Zurek[43]on meadow fescue （Festuca pratensis Huds.） also showed that decreased germination percentage is the main factor affecting genetic integrity of germplasm， and reducing the times of regeneration is also important to maintain genetic integrity of germplasm. Based on the analysis of genetic structure， allele frequency distribution， genetic similarity and number of rare alleles， our results indicated that viability decline has a greater impact than the number of generations on genetic integrity of soybean germplasm. Therefore， seed viability should be the primary consideration for regeneration of soybean germplasm. The general recommendation for seed regeneration in any genebank is when seed viability is reduced 65% to 85% germination. The International Board for Plant Genetic Resources （IBPGR） recommended that stock germplasms should be regenerated before their germination percentage decreases by 5% to 10%. However， frequent regeneration of seed stocks not only increase the costs， but also increase the likelihood of genetic drift， seed mixing， contamination and genetic selection[17]. Therefore， there is a need to determine appropriate timing of seed regeneration. The International Plant Genetic Resources Institute （IPGRI） recommend that the initial germination value should exceed 85% for most seeds and regeneration should also be undertaken when viability falls below 85% of the initial value[44]. Li et al.[45]recommends regeneration of a sample when germination percentage falls below 80%， to minimize genetic drift and reduce the costs. 　　Our results revealed that the progeny of the populations with high germination percentage （98.0% and 95.0%） showed no significant difference from the control populations in effective number of alleles per locus， genetic diversity index， Shannonюs diversity index， allele frequency distribution and number of rare alleles， and they shared high genetic similarity with the control population. However， compared with the control population， the progeny of G03 and G04 with low germination percentage （81.0% and 79.0%， which were 82.7% and 80.6% of the initial germination percentage 98.0%） showed a significant decrease in effective number of alleles per locus， genetic diversity index， Shannonюs diversity index and number of rare alleles， and a significant increase in allele frequency distribution， and they shared low genetic similarity with the control population. The results confirms the recommendation by IPGR that regeneration should be undertaken when viability falls to 85% of the initial value. The experimental soybean Zhonghuang 18 is a bred cultivar with a relatively low level of heterogeneity. Therefore， for those soybean cultivars with a high level of heterogeneity， seed regeneration should be performed strictly following the recommended germination value by IPGRI. It is suggested that regeneration should be undertaken before viability falls to 81.0% of the initial value for the germplasms with a germination percentage of 98%， and if conditions permit， germplasm regeneration should be performed as recommended by IPGRI.
　　Effects of other factors on genetic integrity of soybean germplasm
　　The influence of the size of mating population and seed harvesting method on genetic integrity should also be considered during germplasm regeneration. An appropriate size of mating population can help to maintain genetic integrity of soybean germplasm. Sackville Hamilton et al.[46]concluded that in a population consisting of 100 individuals， the probability of losing a gene with a frequency of 0.05 was only 0.000 35， so they recommended a population of 100 individuals for germplasm regeneration. Schoen et al.[47]reported that the cumulative effect of harmful mutations is negligible only when the sample size is no less than 75 individuals; conversely， harmful mutations may increase significantly after 25 to 50 consecutive cycles of regeneration. To sum up， no less than 75 individuals should be included in a breeding population of soybean germplasm. There are two commonly used seed harvesting procedures： bulk harvesting and balanced harvesting[47-48]. Bulk seed harvesting is to collect all the seeds of maternal plants， and randomly select a portion of the bulk sample for storage， which is easy to practice. Balanced harvesting is to collect equal amounts of seed from maternal plants as required for storage. This method is time and labor consuming. By analyzing the accumulation of mildly deleterious mutations for 50 consecutive regeneration cycles in four mating populations of different sizes （25， 50， 75 and 100 individuals）， Schoen et al.[47]believed that bulk harvesting can reduce mutation accumulation during germplasm regeneration. Field trials on crosspollinated carrot showed that balanced harvesting has less effect than bulk harvesting on the initial gene frequency of regenerated germplasm[48]. Bulk harvesting was adopted in this study， which is easy to practice and can maintain genetic integrity of bred soybean cultivars. But for soybean landraces， balanced harvesting is more efficient in limiting the loss of genetic integrity caused by genetic drift. In summary， many factors such as seed viability， number of generations， breeding population size and harvesting methods should all be taken into account seed regeneration， to maintain the genetic integrity of soybean germplasm. 　　Conclusions
　　The viability decline and number of generations have no significant effect on the genetic similarity between soybean germplasm populations. The progeny from the population with a low germination percentage of only 79.0% still had a high similarity 0.933 3 to the control population. The populations with 98.0% and 95.0% germination percentages and their progeny showed no significant difference from the control population in population genetic structure， allele frequency distribution and number of rare alleles. However， the populations with 81.0% and 79.0%  germination percentages and their progeny were significantly different from the control population in population genetic structure， allele frequency distribution and number of rare alleles. The results suggested that viability decline has a greater impact than the number of generations on genetic structure of soybean germplasm. So， it is suggested that soybean seed with an initial germination percentage of 98.0% should be reproduced before its germination percentage declines to 81.0%.
　　References
　　[1]The Food and Agriculture Organization of the United Nations （FAO）. Report on the state of the worldюs plants genetic resources for food and agriculture[R]. Rome， Italy： FAO. 1996.
　　[2]WANG SM， LI LH， LI Y， et al. Status of plant genetic resources for food and agricultural in China （I）[J]. Journal of Plant Genetic Resources， 2011， 12（1）： 1-12.
　　[3]LU XX， CUI CS， CHEN XL， et al. Survey of seed germinability after 10-12 years storage in the national genebank of China[J]. Plant Genet Resour Sci， 2001， 2（2）： 1-5.
　　[4]WALTERS C， WHEELER LM， GROTENHUIS JM. Longevity of seeds stored in a Genebank： Species characteristics[J]. Seed Sci Res， 2005， 15： 1-20.
　　[5]QIU LJ， CHANG RZ， CHEN KM， et al. Analysis of conservation and regeneration statues for Chinese soybean germplasm[J]. Plant Genet Resour Sci， 2002， 3（2）： 34-39.
　　[6]Frankel OH， Brown AHD， Burdon JJ. The conservation of plant biodiversity[M]. Cambridge： Cambridge University Press， 1995.
　　[7]VAN HINTUM TJL， VAN DE WIEL CCM， VISSER DL， et al. The distribution of genetic diversity in a Brassica oleracea gene bank collection related to the effects on diversity of regeneration， as measured with AFLPs[J]. Theor Appl Genet， 1984， 114： 777-786.
　　[8]ROOS EE. Genetic shifts in mixed bean populations. I. Storage effects[J]. Crop Sci， 1984， 24： 240-244.
　　[9]ROOS EE. Report of the storage committee working population on effects of storage on genetic integrity 1980-1983[J]. Seed Sci Technol， 1984， 12： 255-260. 　　[10]BREESE EL. Regeneration and multiplication of germplasm resources in seed genebanks： The scientific background. Rome： IBPGR， 1989.
　　[11]STOYANOVA SD. Effect of seed aging and regeneration on the genetic composition of wheat[J]. Seed Sci Technol， 1992， 20： 489-496.
　　[12]VAN HINTUM TH JL， VISSER DL. Duplication within and between germplasm collections： II. Duplication in four European barley collections[J]. Genet Res Crop Evol， 1995， 42： 135-145.
　　[13]YONEZAWA K， ISHII T， NOMURA T， et al. Effectiveness of some management procedures for seed regeneration of plant genetic resources accessions[J]. Genetica， 1996， 43： 517-524.
　　[14]RIO AH， BAMBERG JB， HUAMAN Z. Assessing changes in the genetic diversity of potato gene banks： I. Effects of seed increase[J]. Theor Appl Genet， 1997， 95： 191-198.
　　[15]RAO NK， BRAMEL PJ， REDDY KN， et al. Optimizing seed quality during germplasm regeneration in pearl millet[J]. Genet Res Crop Evol， 2002， 49： 153-157.
　　[16]CHEBOTAR S， RODER MS， KORZUN V， et al. Molecular studies on genetic integrity of open pollinating species rye （Secale cereale L.） after long term Genebank maintenance[J]. Theor Appl Genet， 2003， 107（8）： 1469-1476.
　　[17]MURATA M， TSUCHIYA T， ROOS EE. Chromosome damage induced by artificial seed in barley[J]. Theor Appl Genet， 1984， 67： 161-170.
　　[18]DOURADO AM， ROBERTS EH. Chromosome aberrations induced during storage in barley and pea seeds[J]. Ann Bot， 1984， 54： 767-779.
　　[19]ROBERTS EH. Loss of viability： Chromosomal and genetic aspects[J]. Seed Sci Technol， 1973， 1： 515-527.
　　[20]ZHANG H， LU XX， ZHAGN ZE， et al. Effect of seed aging on change of genetic integrity in maize germplasm[J]. Journal of Plant Genetic Resources， 2005， 6（3）： 271-275.
　　[21]TAO KL， PERINO P， PIERLUIGI L， et al. Rapid and nondestructive method for detecting composition change in wheat germplasm accessions[J]. Crop Sci， 1992， 32： 1039-1042.
　　[22]SOENGAS P， CARTEA E， LEMA M， et al. Effect of regeneration Procedures on the genetic integrity of Brassica oleracea accessions[J]. Mol Breed， 2009， 23： 389-395.
　　[23]XIA B， LU XX， CHE XL， et al. Detection of genetic integrity of soybean germplasm using SSR markers[J]. Soybean Sci， 2007， 26（3）： 305-309.
　　[24]ZHANG YM， GAI JY. A study on suitable sample size in the conservation of landrace of soybeans[J]. Sci Agric Sin， 1995， 28（S）： 70-75.
　　[25]ZHENG XY， GUO YP， MA EB. Development of AFLP molecular markers[J]. Chem Life， 2003， 23（1）： 65-67. 　　[26]MAUGHAN PJ， SAGHAIMAROOF MA， BUSS GR， et al. Amplified fragment length polymorphism （AFLP） in soybean species diversity， inheritance， and nearisogenic line analysis[J]. Theor Appl Genet， 2003， 93（3）： 392-401.
　　[27]HU XR， LU XX， ZHANG YL， et al. Studies on the genetic integrity ultra dry seed of wheat with AFLP markers[J]. J Plant Genet Resour， 2003， 4（2）： 162-165.
　　[28]FANG JH， LIU X， LU XX. Technical Regulation on Characterization and Documentation for Crop Germplasm Resources[M]. Beijing： China Agriculture Press， 2008.
　　[29]ISTA. International Rules For Seed Testing 1996[M]. Beijing： China Agriculture Press， 1996.
　　[30]TIAN QZ， GAI JY， YU DY， et al. A study on amplified fragment length polymorphism （AFLP） in soybean[J]. Soybean Sci， 2000， 19（3）： 210-217.
　　[31]TIAN QZ， GAI JY， YU DY， et al. AFLP Fingerprint Analysis of G. soja and G. max in China[J]. Sci Agric Sin， 2001， 34（5）： 465-468.
　　[32]SANGUINETTI CJ， DIAS NE， SIMPSON AJ. Rapid silver staining and recovery of PCR products separated on polyacrylamide gels[J]. Biotechniques， 1994， 17： 914-921.
　　[33]YEH FC， YANG RC， BOYLE T. POPGENE Version 1.31 quick user guide[Z]. Canada： University of Alberta， and Center for International Forestry Research， 1999.
　　[34]ROHLF F J. NTSYSpc： numerical taxonomy and multivariate analysis system， Version 2.1， user guide[Z]. New York： Exeter Publications， 2000.
　　[35]StOYANOVA SD. Genetic shifts and variations of gliadins induced by seed aging[J]. Seed Sci Technol， 1991， 19： 363-371.
　　[36]PARZIES HK， SPOOR W， ENNOS RA. Genetic diversity of barley landrace accessions （Hordeum vulgare ssp. vulgare） conserved for different lengths of time in ex situ gene banks[J]. Heredity， 2000， 84： 476-486.
　　[37]ROOS EE. Genetic shifts in mixed bean populations. II. Effects of regeneration[J]. Crop Sci， 1984， 24： 711-715.
　　[38]BORNER A， CHEBOTAR S， KORZUN V. Molecular characterization of the genetic integrity of wheat （Triticum aestivum L.） germplasm after longterm maintenance[J]. Theor Appl Genet， 2000， 100： 494-497.
　　[39]MARSHALL DR， BROWN AHD. Optimum sampling strategies in genetic conservation[A]. In： Frankel O H， Hawkes J G eds. Crop Genetic Resources for Today and Tomorrow[C]. Cambridge： Cambridge University Press， 1975.
　　[40]ROOS EE. Genetic changes in a collection over time[J]. HortScience， 1988， 23： 86-90.
　　[41]CHWEDORZEWSKA KJ， BEDNAREK PT， PUCHALSKI J， et al. AFLPprofiling of longterm stored and regenerated rye genebank samples[J]. Cell Mol Biol Lett， 2002， 7： 457-463. 　　[42]CHWEDORZEWSKA KJ， BEDNAREK PT， PUCHALSKI J. Studies on changes in specific rye genome regions due to seed aging and regeneration[J]. Cell Mol Biol Lett， 2002， 7： 569-576.
　　[43]ZUREK G. Effect of seed storage on germplasm integrity of meadow fescue （Festuca pratensis Huds.）[J]. Genet Res Crop Evol， 1999， 46： 485-490.
　　[44]FAO/ IPGRI. Genebank standards[Z]. Rome： FAO/IPGRI， 1994.
　　[45]LI LZ， WANG LN， GENG XL， et al. Seed vigor influences on some agronomic characteristics of soybean[J]. J Hebei Agri Sci， 2002， 6（1）： 14-19.
　　[46]SACKVILLE HAMILTON N R， CHORLTON K H. Regeneration of accessions in seed collection[Z]. Rome： IPGRI， 1997.
　　[47]SCHOEN DJ， DAVID JL， BATAILLON TM. Deleterious mutation accumulation and the regeneration of genetic resources[J]. Proc Natl Acad Sci USA， 1998， 95： 394-399.
　　[48]LE CLERC V， BRIARD M， GRANGER J， et al. Genebank biodiversity assessments regarding optimal sample size and seed harvesting techniques for the regeneration of carrot accessions[J]. Biodivers Conser， 2003， 12： 2227-2336.