Genetic Resources and Crop Evolution

, Volume 60, Issue 4, pp 1407–1421

Preservation of seed viability during 25 years of storage under standard genebank conditions

Authors

    • Centre for Genetic Resources
    • Wageningen University and Research Centre
  • E. C. de Groot
    • Centre for Genetic Resources
    • Wageningen University and Research Centre
  • Th. J. L. van Hintum
    • Centre for Genetic Resources
    • Wageningen University and Research Centre
Research Article

DOI: 10.1007/s10722-012-9929-0

Cite this article as:
van Treuren, R., de Groot, E.C. & van Hintum, T.J.L. Genet Resour Crop Evol (2013) 60: 1407. doi:10.1007/s10722-012-9929-0

Abstract

Maintaining sufficient viability is critical to the sustainability of ex situ conserved seed collections. For this reason, accessions are regenerated when viability falls below a predefined threshold. Viability is monitored by determining the germination ability of accessions at predefined time intervals. Optimizing the frequency of these germination tests, in order to avoid waste of resources, is hampered by the scarce availability of data about seed longevity, particularly for material maintained under genebank conditions. Here we report on the analysis of nearly 40,000 germination test results collected for a wide range of crop species over a 25-years period by the centre for genetic resources, the Netherlands (CGN), where seeds of genebank accessions are dried to 3–7 % moisture content and stored for the long term under near-vacuum in aluminium foil bags at −20 °C. The results indicate that seed viability is well maintained for the large majority of seed lots during the first 25 years after regeneration as only 3.3 % of the monitoring tests revealed below-threshold germination values. It is argued that the majority of these sub-standard seed lots are due to other causes than seed ageing, including dormancy problems and estimation error in germination testing. For material, maintained under the seed management procedures and storage conditions practiced by CGN, it is therefore recommended to delay the first germination monitoring tests to 25 years after regeneration.

Keywords

Accession maintenanceEx situ conservationGenebanksGermination testingSeed agingSeed longevity

Introduction

Since the 1960s, many collections of crops and their wild relatives have been established in order to ensure availability of genetic resources for present and future utilization. These genetic resources are managed in various ways, including through in situ and on-farm conservation, or in the form of in vivo or in vitro collections, but maintenance of ex situ seed collections forms a major approach. It has been estimated that on a worldwide scale approximately 7.4 million accessions are conserved in 1,750 genebanks (FAO 2010). Since the establishment of crop collections, guidelines have been developed for effective germplasm management (e.g. Engels and Visser 2003). Maintenance of seed viability plays a critical role in managing seed collections, which includes testing for sufficient initial germination before genebank storage and subsequent germination monitoring at predefined time intervals (Rao et al. 2006). Optimization of the frequency of germination monitoring is crucial to genebank management in order to reduce costs and seed depletion, while avoiding loss of accessions due to undetected viability reduction. Standards for genebank management procedures were developed by the plant production and protection division of the food and agriculture organization of the United Nations and the International plant genetic resources institute (FAO/IPGRI 1994) and are currently being revised at the request from the commission on genetic resources for food and agriculture based on current technological and scientific knowledge (CGRFA 2012). The proposed standards for seed management include a level of at least 85 % initial germination for most seeds and monitoring intervals equalling one-third of the expected time for germination to fall to 85 % of the initial level. Furthermore, 85 % of the initial germination is proposed as a threshold for regeneration or alternative management decisions, such as recollecting seeds. However, management practices may vary among genebanks due to differing in-house policies and availability of resources.

Defining optimal seed management procedures is hampered by the fact that the life expectancy of seeds may be very high (Shen-Miller et al. 1995), while most genebanks still have a short life history. As a consequence, data on seed longevity under genebank storage conditions are scarce (Walters et al. 2005; Hay et al. 2012). Seed ageing studies have been conducted under a variety of experimental circumstances, including ambient conditions (Priestley et al. 1985; Nagel and Börner 2010) and controlled deterioration conditions (Ellis et al. 1989; Dickie et al. 1990). In general, these studies show a sigmoid survival curve when seed viability is plotted against time. The characteristics of this curve may vary among species and are a function of storage temperature and relative humidity (Ellis 1991; Nagel and Börner 2010). However, it can be questioned whether the longevity estimates obtained under the unfavourable conditions of seed ageing experiments may be extrapolated to genebank storage conditions since the underlying deterioration mechanisms might differ (Freitas et al. 2006). In one of the few studies on seed longevity under conditions of low temperature and low moisture content, using germination data of the USDA National Center for Genetic Resources Preservation, significant reductions in viability were observed in many species only after 30 years of storage (Walters et al. 2005). In addition to the storage environment, seed longevity may also depend on seed production conditions and post-harvesting procedures, whereas also genetic variation for longevity has been reported (Nagel et al. 2009, 2011). Therefore, more data on the viability of genebank collections are needed in order to improve our understanding of seed longevity and to develop sound management practices.

Here we report on the analysis of germination data collected by the centre for genetic resources, the Netherlands (CGN) over its 25-year existence. The prime objective was to investigate developments of seed viability over time, whereas the secondary aim was to evaluate options to improve the testing efficiency thereof.

Materials and methods

Germination testing

New seed lots acquired by CGN through collecting missions, donations or regeneration of existing accessions are first checked for sufficient seed quantity, and regenerated when required. Subsequently, it is determined whether the viability of the seed lot, measured as the level of germination, is sufficient for storage, applying less strict criteria to wild crop relatives as high germination levels are often difficult to reach for such material. Current germination threshold values for cultivated and wild material are 77 and 57 %, respectively. In case the germination test results are below the threshold, viability is retested, sometimes after repeated cleaning of the seed lot or after germination inducing treatments prior to testing. In case of sufficient viability, the new seed lot is stored, and the corresponding germination test result is documented as the ‘initial germination level’. For rejected seed lots, new seeds are obtained through regeneration. Seed lots showing consistent below-threshold germination levels are discarded. Only in exceptional cases, such as for rare or exotic germplasm, it may be decided to accept the seed lot even though the germination test results are below the threshold. Seed lots from the initial period after CGN’s establishment in 1985 do not necessarily meet these standards because germination thresholds were formalized not earlier than 2001. Seed lots, dried at 15 % relative humidity to 3–7 % moisture content, are vacuum-sealed in three-layered aluminium foil bags that are stored at −20 °C in the long-term storage facility. However, sub-optimal storage conditions were practiced before 1985 for most crops with the exception of lettuce.

After initial germination testing, the viability of seed lots is checked at fixed time intervals, varying between 10 and 25 years depending, amongst others, on the crop involved and on initial germination results. These tests are referred to as ‘germination monitoring’. Threshold values used for germination monitoring are identical to those applied to initial germination testing. The viability is retested if test results show below-threshold germination levels, and the accession is regenerated in case of repeatedly low germination. In case of multiple data points per germination monitoring test, only the latest measurements were used in the present study. The number of years between regeneration and observed below-threshold germination upon storage is the period a seed lot actually lasts under genebank management practices. Throughout the manuscript this time period is denoted as ‘seed lot longevity’, which should not be confused with the time to reach a predefined germination level as is more generally used in seed viability studies.

Germination tests are outsourced to agencies that follow the procedures of, and are certified by, the international seed testing association (ISTA 2012). These procedures include scoring of the number of normal seedlings, abnormal seedlings and non-germinating but healthy, hard and swollen seeds. Throughout this paper germination test data refer to the number of normal seedlings scored. The number of seeds per test may vary from 50 to 200, depending on the crop. Test procedures may vary between crops due to different seed characteristics, and may differ between years because of modifications in the scoring protocols. Since 2001, 5 % of the annually tested seed lots are re-examined by the testing agency in a blind experiment to obtain estimates of the reliability of the germination test results. These tests are referred to as ‘germination reproducibility tests’, of which the results are described in an accompanying paper (van Hintum and van Treuren 2012).

Germination data

Germination data are managed in CGN’s Oracle-based documentation system called GENIS. All germination data registered before May 2011 were downloaded from GENIS in Excel format. Subsequently, potential errors and missing values were marked and checked on a crop-by-crop basis by the responsible curators. For example, conflicting dates between regeneration and germination testing were checked and corrected where appropriate, and regeneration reports were re-examined in an attempt to find missing regeneration years. In case doubts about suspicious records could not be resolved, the data of that seed lot were removed from the data file. Subsequently, the corrected and supplemented data were uploaded in GENIS and used for the present study.

The examined data consisted of 27,931 records, each representing a seed lot documented with crop name, accession number, generation number, regeneration year (if available), population type (cultivated or wild) and associated germination test results. Regeneration years were missing for 1,468 (5.3 %) of the seed lots. The germination test results included the date, test method code, germination test type (initial or monitoring test), germination test result and, for a minority of the tests, the percentage of dormant seeds. Each record contained the initial germination test result, while the number of germination monitoring test results varied from 0 to 3 per seed lot. Due to variation in collection size, time of introduction to the genebank and frequency of germination monitoring testing, the crops included in the study contributed unevenly to the dataset (Table 1). The group ‘Others’, with less than 100 germination monitoring results per crop, consisted of bentgrass (Agrostis capillaris L.), caraway (Carum carvi L.), clover (Trifolium pratense L., Trifolium repens L.), cucumber (Cucumis sativus L.), fescue (Festuca rubra L., Festuca spp.), flax (Linum usitatissimum L., Linum spp.), Kentucky bluegrass (Poa pratensis L.), lily (Lilium spp.), Lolium (Lolium perenne L., Lolium multiflorum Lamk.), lupin (Lupinus albus L., Lupinus luteus L.), maize (Zea mays L.), melon (Cucumis melo L.), orchard grass (Dactylis glomerata L.), potato (Solanum spp.) and timothy (Phleum pratense L., Phleum bertolonii DC.). Concerning germination monitoring, 40.4 % of the tests are covered by lettuce alone. This overrepresentation was caused by the early introduction of lettuce at CGN, the rather large collection size and the high monitoring frequency of once per 5 years in the early stages of collection development due to the presumed poor seed longevity. Altogether the dataset consisted of 37,963 germination test results.
Table 1

Number of initial germination (A) and monitoring (B) tests per crop per 5-year period since regeneration

Cropa

0–5

6–10

11–15

16–20

21–25

>25

Unk.

Total

A. Initial germination

 Allium

410

41

4

1

1

 

43

500

 Barley

3578

120

1

   

398

4097

 Crucifers

1301

428

130

13

  

100

1972

 Eggplant

201

143

69

84

39

  

536

 Faba bean

785

78

    

15

878

 Lettuce

2336

208

195

1

  

270

3010

 Oat

273

 

507

2

  

2

784

 Pea

1038

4

2

   

26

1070

 Pepper

782

63

174

100

  

1

1120

 Spinach

342

152

53

3

  

24

574

 Tomato

777

300

340

13

   

1430

 Wheat

4076

1770

497

   

433

6776

 Others

4060

145

528

282

13

 

156

5184

 Total

19959

3452

2500

499

53

 

1468

27931

B. Germination monitoring

 Allium

 

38

126

44

9

3

34

254

 Barley

1

 

791

340

15

1

59

1207

 Crucifers

4

83

205

278

119

71

23

783

 Eggplant

    

53

51

 

104

 Faba bean

 

14

109

18

20

 

3

164

 Lettuce

22

1103

1157

1082

412

166

112

4054

 Oat

 

2

2

50

134

135

 

323

 Pea

  

117

81

 

2

13

213

 Pepper

 

2

100

40

141

148

 

431

 Spinach

26

4

12

38

11

4

6

101

 Tomato

 

22

83

72

243

29

 

449

 Wheat

 

5

385

798

292

72

99

1651

 Others

1

2

145

70

15

34

31

298

 Total

54

1275

3232

2911

1464

716

380

10032

‘Others’ represent a combined group of crops for which less than 100 germination monitoring tests were performed

Test data on seed lots with unknown regeneration are denoted by ‘Unk.’

aBotanical names Allium (Allium cepa L., Allium spp.), barley (Hordeum vulgare L., Hordeum spp.), crucifers (Brassica oleracea L., Brassica spp., Camelina sativa L., Descurainia sophia (L.) Webb, Eruca sativa Mill., Hesperis matronalis L., Portulaca sp., Raphanus sativus L., Sinapsis alba L.), eggplant (Solanum melongena L., Solanum spp.), faba bean (Vicia faba L.), lettuce (Lactuca sativa L., Lactuca spp., Chondrilla juncea L., Cicerbita plumieri (L.) Kirschleger, Ixeridium dentatum (Thunb.) Tzvelev, Mycelis muralis (L.) Dumort., Streptorhamphus tuberosus (Jacq.) Grossch.), oat (Avena sativa L., Avena spp.), pea (Pisum sativum L., Pisum spp.), pepper (Capsicum annuum L., Capsicum spp.), spinach (Spinacia oleracea L., Spinacia turkestanica Iljin.), tomato (Lycopersicon esculentum Mill., Lycopersicon spp., Solanum lycopersicoides Dunal), wheat (Triticum aestivum L., Triticum spp.)

Data analysis

In order to obtain a general understanding of viability as a function of time, initial germination and germination monitoring test results were combined and plotted against the time since regeneration. For this purpose, separate analyses were performed for cultivated and wild material. A linear regression was performed to examine the degree of correlation.

In order to investigate the rate of viability changes as a function of time, the change in germination percentage per time interval was determined for each pair of consecutive tests and expressed as (Gx − Gx−1)/(tx − tx−1), where G denotes the germination percentage, t the experimental year (t1 representing the year of initial germination testing) and x may range from 2 to 4 depending on the number of germination tests performed for a seed lot. Subsequently, these values were plotted against the mean time since regeneration, i.e. ((tx − tr) + (tx−1 − tr))/2, where tr denotes the regeneration year. A linear regression was performed to examine the extent of correlation.

Analyses of time trends in viability are complicated by the fact that once a seed lot falls below the threshold value, germination monitoring is discontinued because the seed lot is regenerated. As a consequence, the overall level of below-threshold germination is increasingly being underestimated the longer the time since regeneration. To correct for this bias, a straightforward model was developed that provides unbiased estimates of the probability of below-threshold germination (Pthr) as a function of time. First, the time between regeneration and the last germination test was calculated for each seed lot, and it was recorded whether the germination test result was above or equal to the threshold value, denoted by A, or below, denoted by B. It was assumed that B samples had fallen below the threshold in the year prior to testing. Subsequently, the number of A and B seed lots was calculated for each year (t) since regeneration, denoted by nA,t and nB,t, respectively. At t1, the first year of testing, the probability of below-threshold germination (Pthr,1) is simply calculated as nB,1/(nA,1 + nB,1). To obtain an unbiased estimate of this probability in the second year (Pthr,2), the number of B seed lots in the second year has to be increased with a number of B seed lots observed in the first year, proportional to the sample size in the second year. Therefore, the correction factor for the B fraction is determined by the ratio of B/A samples in the first year and the sample size in the second year. The corrected number of B samples in the second year (n’B,2) was calculated as nB,2 + (nB,1/nA,1) ∗(nA,2 + nB,2), allowing Pthr,2 to be calculated as n’B,2/(nA,2 + n’B,2). In the following years the correction factor n’B,t could always be calculated as nB,t + (n’B,t−1/nA,t−1) ∗ (nA,t + nB,t), and Pthr,t as n’B,t/(nA,t + n’B,t). The standard error of this estimate was calculated using the equation for the standard error of Bernoulli variables, i.e. √(Pthr,t ∗ (1−Pthr,t)/(nA,t + n’B,t)) (Sokal and Rohlf 1981). Seed lot longevity was visualized by plotting Pthr estimates against the time since regeneration. Analyses were also performed for separate crops and groups of material, which were compared for estimated values at 25 years after regeneration (Pthr,25).

In order to examine a potential effect of initial viability on seed lot longevity, the number of years until a below-threshold germination test result was observed, was plotted against the initial germination test result, and a linear regression was performed to examine the extent of correlation. However, this analysis does not include seed lots that have not fallen below the threshold yet, which may obscure a potential relationship between initial viability and longevity. To correct for this potential bias, initial germination data were classified into different germination level categories, and the probability of below-threshold values at 25 years after regeneration (Pthr,25) was estimated for each category using the aforementioned procedures.

An accompanying study on germination reproducibility tests for CGN accessions showed that germination testing is prone to large estimation errors (van Hintum and van Treuren 2012). It was shown in that study that the error depended on the viability, corresponding to the function y = −0.4117x2 + 0.353x + 0.061, where y is the standard deviation of the test result and x is the viability of a seed lot (expressed as the fraction viable seeds). This function was used in the present study to evaluate the effect of germination estimation error on regeneration decisions. For this purpose, viabilities ranging from 0.5 to 1.0 were considered, and the corresponding probability of a seed lot scoring twice (i.e. also in a repeated test) below the threshold germination level was estimated using the aforementioned error distribution. This procedure simulates the germination monitoring procedures of CGN, where seed lots with below-threshold germination test results are retested for consistency before deciding on regeneration. The probability of two consecutive germination tests with results below the threshold was plotted against the considered viability to evaluate the probability of correct decisions concerning regeneration.

All data analyses were performed in MS Excel v14.0.6112.5000 as part of MS Office professional Plus 2010.

Results

Initial germination testing

The total sample of initial germination tests consisted of 27,931 observations. Data were collected on average 4.0 years after regeneration, germination levels showing an average of 89.3 % (Table 2A). As could be expected based on different threshold criteria for uptake in the collection, cultivated material (90.3 %) showed a higher average germination than wild material (82.9 %). Despite the uptake thresholds, 2,661 (9.5 %) seed lots appeared accessed with below-threshold germination levels, a finding that can be explained by less strict uptake criteria in the early years after CGNs establishment. Concerning accessions with below-threshold initial germination, no substantial difference between cultivated and wild material was observed. Summing up germination data with available dormancy data showed only a minor effect as the number of below-threshold germinations dropped to 2,563 (9.2 %). No relationship was observed between the time since regeneration and initial germination levels (results not shown).
Table 2

Data overview of initial germination testing (A) and germination monitoring (B) for cultivated material, wild material and the total sample

 

Cultivated

Wild

Total sample

 

All

Norm

All

Norm

All

Norm

A. Initial germination

 Number of observations

24194

21893

3737

3377

27931

25270

 Mean number of years since regeneration (s.d.)

3.9

(4.1)

3.8

(4.0)

4.6

(5.3)

4.6

(5.3)

4.0

(4.3)

3.9

(4.2)

 Mean germination percentage (s.d.)

90.3

(12.3)

93.4

(5.9)

82.9

(19.6)

87.9

(11.4)

89.3

(13.7)

92.7

(7.1)

 Number of tests below threshold germination value

2301

(9.5 %)

0

360

(9.6 %)

0

2661

(9.5 %)

0

B. Germination monitoring

 Number of observations

8315

7509

1717

1654

10032

9163

 Mean number of years between tests (s.d.)

10.8

(3.0)

10.8

(3.1)

9.9

(2.6)

9.9

(2.6)

10.7

(2.9)

10.6

(3.0)

 Mean change in germination percentage (s.d.)

1.4

(6.9)

0.8

(6.2)

0.4

(9.4)

−0.2

(7.2)

1.2

(7.4)

0.6

(6.4)

 Number of tests below threshold germination value

676

(8.1 %)

280

(3.7 %)

51

(3.0 %)

21

(1.3 %)

727

(7.2 %)

301

(3.3 %)

For each of the three groups, results are presented separately for all material (All) and material meeting the initial germination criteria (≥77 % for cultivated and ≥57 % for wild material) that are currently used for inclusion in the collection (Norm)

Threshold values used for germination monitoring are equal to the criteria used for initial germination

s.d. standard deviation, as an indication of the variation—it should be noted that the distributions can be highly skewed

Germination monitoring

Overall, the study included 10,032 germination monitoring tests that on average were performed at a 10.7 years’ time interval (Table 2B). This test frequency is higher than what would be expected based on CGN’s currently used standard procedures concerning germination testing, due to the fact that for safety reasons some crops were tested at intervals of 3–5 years in the initial stage after CGN’s establishment. On average, a slight increase of 1.2 % germination was observed between consecutive tests. However, changes in germination levels were accompanied by large standard deviations. Regarding average changes in germination, no substantial differences were observed between cultivated and wild material. Overall, the monitoring tests revealed 727 (7.2 %) seed lots with below-threshold germination. However, 426 out of these 727 seed lots (58.6 %) also showed an initial germination test result below the threshold. The fraction of test results below the threshold in germination monitoring appeared nearly three times higher for cultivated material as for wild material. Because of their distorting influence, records with initial germination test results below the threshold were excluded from further analyses.

Time trends

Viability scores as a function of time since regeneration are presented for cultivated material in Fig. 1A and for wild material in Fig. 1B. The figure for cultivated material was cut off at 35 years as only few older observations were available. The highest time interval observed for cultivated material was 40 years. Both for cultivated and wild material, the large majority of seed lots maintained their high initial viability during the examined time interval. Neither for cultivated (R² = 0.032), nor for wild (R² = 0.000) material, regression analysis revealed a correlation between germination percentage and time since regeneration.
https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9929-0/MediaObjects/10722_2012_9929_Fig1_HTML.gif
Fig. 1

Germination test results for cultivated material (A) and wild material (B) meeting the initial germination threshold, plotted against the number of years between regeneration and germination testing. The total number of observations is denoted by n, and the area of the data points is proportional to the number of included observations. The horizontal line at 77 % in plot A and 57 % in plot B represent the threshold germination values

Changes in germination between consecutive tests are presented in Fig. 2 as a function of time. The number of tests showing increased germination appeared of similar magnitude as the number of tests showing decreased germination. If seed lots tend to age more rapidly after longer storage time, a positive relationship would be expected in Fig. 2. However, no correlation between the extent of germination change and time since regeneration was revealed by regression analysis (R² = 0.000), suggesting that during the considered time period the seed lots have not yet reached the slope of the generally accepted sigmoid survival curve. Separate analyses performed for cultivated and wild material resulted in similar findings (results not shown).
https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9929-0/MediaObjects/10722_2012_9929_Fig2_HTML.gif
Fig. 2

Annual change in germination test result between consecutive tests against the mean number of years between regeneration and germination testing for material meeting the initial germination threshold. The plot is based on a total of 8,892 observations, and the area of the data points is proportional to the number of included observations. The horizontal line at y = 0 indicates stability of germination values

Seed lot longevity

At CGN, germination monitoring of a seed lot is discontinued once that seed lot falls below the viability threshold and a new seed lot of the accession is made by regeneration. Consequently, the total number of germination test results below the threshold is increasingly under-estimated the longer the time since regeneration. Data, corrected for this effect, are presented in Fig. 3 where the unbiased probability of seed lots with below-threshold germination is plotted against the time since regeneration. The figure was cut off at 25 years as only 32 test results below the threshold were obtained for samples stored over a longer period, resulting in rather large sampling errors. In the period of 11–25 years after regeneration, the probability of a seed sample with below-threshold germination gradually increased from zero to 8.3 %.
https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9929-0/MediaObjects/10722_2012_9929_Fig3_HTML.gif
Fig. 3

Unbiased probability estimates of seed lots with below-threshold germination values (Pthr) as a function of time since regeneration. Analyses were restricted to seed lots meeting the initial germination threshold. The plot is based on a total of 24,028 observations. Error bars represent the 95 % confidence interval

Unbiased probability estimates of seed lots with below-threshold germination at 25 years after regeneration are presented for different crops and different groups of material in Table 3. The probability appeared considerably higher for cultivated material (9.1 %) than for the crop wild relatives (1.8 %), a finding that is most likely due to the difference in thresholds applied for these groups, i.e. 77 % for cultivated and 57 % for wild material. Crop-specific probability estimates regarding below-threshold germination ranged from 2.1 % for lettuce to 20.1 % for eggplant. However, the standard error of these estimates were sometimes found to be considerable, and varied among crops due to variation in the number of below-threshold observations and the share of tests performed on older seed lots. For these reasons, seed lot longevity has been presented for only eight crops. Because a large fraction of the germination monitoring tests were carried out with lettuce accessions (Table 1), a separate analysis was performed for non-lettuce material, resulting in a probability estimate and standard error slightly higher than the values observed for the analysis of all materials (Table 3). Sub-optimal storage conditions prevailing before entering the genebank facility may have negatively affected seed lot longevity. Therefore, seed lot longevity was specifically estimated for material regenerated before the establishment of CGN in 1985. Contrary to expectations, the resulting estimate of below threshold germination for these old materials was approximately half the value observed for all materials (Table 3).
Table 3

Unbiased estimate and standard error of the probability of seed lots with below-threshold germination at 25 years after regeneration (Pthr,25) for separate crops and groups of material

Material

Pthr,25

Standard error

Total observations

Tests below threshold

Observations 25 years

All

8.3

0.9

24028

255

787

Wild

1.8

1.8

3296

19

57

Cultivated

9.1

1.0

20732

236

730

Lettuce

2.1

0.9

2654

21

229

Spinach

3.8

7.6

442

2

6

Pepper

5.3

1.8

1012

16

154

Wheat

5.6

2.6

5622

41

74

Crucifers

7.6

2.9

1796

24

76

Tomato

11.6

3.5

1327

19

75

Oat

12.6

3.5

513

17

78

Eggplant

20.1

5.1

515

19

50

Non-lettuce

10.8

1.2

21374

234

558

<CGN

4.2

0.7

5327

63

772

Crop-specific estimates are limited to crops for which at least 100 germination monitoring tests were performed, and for which the standard error of the estimate was smaller than 10 %

Also presented are the total number of included observations, the number of below-threshold observations and the number of test results on seed lots of 25 years or older

Material regenerated prior to the establishment of CGN in 1985 is denoted by ‘<CGN’

Sub-standard seed lots

Within the group of seed lots with known regeneration year and initial germination meeting the standards, 287 seed lots with below-threshold germination monitoring results were observed. These seed lots will be referred to as ‘sub-standard seed lots’. This number was slightly lower than the number presented in Table 2 since 14 samples in this group were excluded because of unknown regeneration year. Twenty five sub-standard seed lots (8.7 %) no longer scored below the threshold value when available dormancy data were summed up with the germination data, while for 188 sub-standard seed lots (65.5 %) the below-threshold measurement had not been reconfirmed in a second independent test. Germination test results of at least 50 % were observed for 264 sub-standard seed lots (92.0 %).

The longevity of the sub-standard seed lots, expressed as the number of years between regeneration and observed below-threshold germination, appeared uncorrelated (R² = 0.000) with initial germination values (Fig. 4A). However, a general relationship between initial germination and seed lot longevity may have been obscured by the low representation of samples with a high initial viability among the sub-standard seed lots. Therefore, unbiased probability estimates of seed lots with below-threshold germination at 25 years after regeneration (Pthr,25), were determined for groups of samples with different initial germination values. A clear decline in Pthr,25 with increasing initial germination was observed for cultivated material (Fig. 4B). Crop wild relatives showed a similar relationship between initial germination and seed lot longevity, albeit that the standard errors of the estimates were relatively high (results not shown). The results indicated that, as could be expected, seed lots with relatively low initial viability have a higher probability on sub-standard test results during germination monitoring.
https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9929-0/MediaObjects/10722_2012_9929_Fig4_HTML.gif
Fig. 4

(A) Longevity of sub-standard seed lots, expressed as the number of years between regeneration and germination monitoring results below the threshold, as a function of the initial germination. The area of the data points is proportional to the number of included observations. (B) Probability of cultivated seed lots with below-threshold germination values at 25 years after regeneration (Pthr,25), per initial germination category. Error bars represent the 95 % confidence interval. All analyses were restricted to material meeting the initial germination threshold

In an accompanying study on 641 germination reproducibility tests it was shown that error levels were much higher than could be expected based on sampling effects. For example, germination test results of a sample with a viability of 80 % were shown to have a standard deviation of 8.0 %, whereas 2.8 % would be expected if the error was based solely on chance effects (van Hintum and van Treuren 2012). To examine the effect of estimation error on germination testing and hence on regeneration decisions, the error distribution deduced from the germination reproducibility tests was used to estimate the probability of below-threshold germination for cultivated material. Seed lots with a real viability of 77 % were estimated to have a 22.8 % probability of being regenerated based on the actual measurements in germination monitoring tests, a probability that decreased to 0.8 % at 85 % real viability. On the other hand, seed lots with a real viability of 76 % were estimated to have a 72.8 % probability of not being regenerated, a level that decreased to 15.8 % at 60 % viability and further down to 4.8 % at 50 % viability (Fig. 5). Thus, estimation error causes a significant number of unjustified decisions to regenerate, as well as unjustified decisions not to do so. The finding that decision errors are skewed towards no regeneration when the real viability is close to the threshold value of 77 % is due to CGN’s policy to retest only below-threshold germination results for consistency. Notwithstanding this skewedness, a substantial part of the observed sub-standard seed lots that could not be explained by dormancy observations, can be expected to represent seed lots with sufficient viability, falling below the germination threshold by chance as a result of estimation error, rather than to represent samples with low seed viability due to seed ageing.
https://static-content.springer.com/image/art%3A10.1007%2Fs10722-012-9929-0/MediaObjects/10722_2012_9929_Fig5_HTML.gif
Fig. 5

Effects of germination estimation error on the probability of regeneration for cultivated material. Probabilities are based on error estimates obtained from germination reproducibility tests (van Hintum and van Treuren 2012). The line at 77 % germination marks the threshold value for cultivated material, greater or equal values indicating maintenance of the seed lot and smaller values indicating regeneration

Discussion

Proper seed management is essential in collection management in order to maintain viable seeds for the purposes of conservation and utilization (Smith et al. 2003). It includes optimal seed storage conditions, germination monitoring at certain time intervals and regeneration of accessions when viability has fallen below a predefined level. However, germination testing and regeneration are costly operations, while regeneration may compromise the genetic integrity of an accession (van Hintum et al. 2007). In order to avoid the loss of accessions due to seed ageing as well as the waste of funds, optimization of seed management is critical to genebanks (Engels and Visser 2003). Nevertheless, studies to analyse seed viability upon storage over prolonged periods of time in order to optimize germination monitoring regimes and protocols are rather scarce (Trapp et al. 2012).

In the present study, seed viability was found to be well maintained for the vast majority of seed lots during the first 25 years after regeneration, and no clear indication of deterioration was observed. This finding is in accordance with a previous study on USDA-managed genebank materials, revealing significant reductions in germination only after 30 years of storage in most of the studied species (Walters et al. 2005). Seed storage conditions of CGN and USDA are quite similar with respect to seed moisture content (3–7 % at CGN, 4–8 % at USDA), packing material (vacuum sealed aluminium foil bags at both institutes) and storage temperature (−20 °C at CGN, −18 °C at USDA since 1978). Nevertheless, the general validity of the seed longevity estimates for material maintained under such storage conditions remains to be determined as regeneration conditions and post-harvesting procedures to which the materials are exposed before entering the genebank may affect seed viability (Hay and Probert 1995; Jalink et al. 1998; Groot and de Groot 2007). Furthermore, the seed ageing patterns may vary among species (Ellis 1991; Walters et al. 2005; Nagel and Börner 2010), while also intra-specific variation has been reported (Nagel et al. 2009; 2011). No species-specific analyses were performed in the present study due to the relatively small number of significant decreases in germination level observed, which would have resulted in low-confidence longevity estimations for separate species.

Evidently, longevity estimation also depends on the reliability of germination testing. In an accompanying study on the repeatability of germination tests results for which data were collected by CGN, a considerable level of estimation error was found (van Hintum and van Treuren 2012). The effect of estimation error is illustrated in Fig. 2, which shows that temporal increases in germination percentage occurred equally frequent as reductions. It may be assumed that the closer a seed lot is to the threshold germination level, the higher the probability of falling below the threshold due to estimation error (Fig. 5). Moreover, the error level increases with decreasing germination levels (van Hintum and van Treuren 2012), causing estimation errors to have larger effects on seed lots close to the threshold germination level than on seed lots with high germination values. In order to reduce the probability of unnecessary regeneration, seed lots with germination test results below the threshold are retested for repeatability of the measurements. Due to this policy, decision errors are skewed towards no regeneration because below-threshold results in the first test may be overruled by the second test, whereas above-threshold results are not retested. As CGN did not apply the retesting procedure from the start of its operations, retesting of below-threshold measurements has not been carried out for 65.5 % of the observed sub-standard seed lots. For these seed lots the probability of falling below the threshold level by chance is higher than the level presented in Fig. 5 as the results presented in this figure are based on retesting below-threshold measurements. The fraction sub-standard seed lots expected by chance can be estimated by sampling the error distribution under the assumption that viability levels remained stable during the first 25 years after regeneration. However, such an approach depends on reliable initial germination values. The fact that initial germination test results were affected by estimation errors as well prevented quantification of the effects of estimation error on the frequency of sub-standard seed lots.

In addition to factors influencing the quality of germination measurements, dormancy also plays an essential role in the measurement of seed viability (Pérez-García et al. 2007; 2009). Some species are more prone to dormancy than others, and the effects are generally found to be larger for crop wild relatives than for cultivated material. Swollen, non-germinating seeds that are hard and unaffected by fungal infection are generally scored as dormant seeds during germination testing. In the present study, 8.7 % of the observed sub-standard seed lots would no longer score below the threshold if germination data were corrected for observed dormancy. Unfortunately, dormancy data in CGN’s documentation system are rather anecdotal as quantification of dormancy has not been common practice from the start of CGN’s operations, and dormancy data have not been consequently registered. In fact, dormancy may have had a larger contribution to the scoring of samples as sub-standard seed lots than is apparent from the current analysis. For example, grain legumes are known for their hardseededness, which has been shown to affect germination negatively (Ellis and Roberts 1982; Ellis et al. 1987). In the present study, the grain legumes lupin, pea and faba bean displayed relatively high sub-standard seed lot frequencies of 16.1, 15.2 and 7.5 %, respectively. In all cases where dormancy data of these sub-standard seed lots were available (39 %), germination values corrected for dormancy no longer scored below-threshold. Therefore, it may be assumed that dormancy was also involved in part of the remaining 61 % of the observed sub-standard seed lots of these grain legumes. Moreover, dormancy and estimation errors may to some extent be related as the germination testing agencies are used to work with commercial cultivars rather than with various crop wild relatives and unknown landraces, samples that are often found to be more prone to dormancy. According to the germination reproducibility tests, indeed considerably higher error levels were found for crop wild relatives as compared to cultivated material (van Hintum and van Treuren 2012). Considering the relevance of dormancy data for the improvement of viability monitoring, recording and registration of dormancy is nowadays standard procedure at CGN.

A large part of the germination data used for the present study originated from cultivated lettuce (Lactuca sativa) and its crop wild relatives. Based on the seed viability equation of Ellis and Roberts (1980), the seed viability constants of Kraak and Vos (1987) and CGN’s seed management practices, the time for lettuce to fall to 50 % germination was estimated to be longer than 100 years, using Kew’s seed viability module (Flynn and Turner 2004). However, in various experimental studies, L. sativa has been classified as ‘bad keeper’ based on the short relative longevity observed in comparison with other crops (Walters et al. 2005; Nagel and Börner 2010). These findings are not supported by the present study as lettuce outperformed all other examined crops in terms of seed lot longevity (Table 3). If lettuce truly would be a ‘bad keeper’, then our results indicate that under optimal seed management procedures and storage conditions as practiced by CGN, i.e. dry seeds stored under vacuum at −20 °C, accessions of a wide variety of crops may expected to be maintained for at least 25 years without significant loss of viability. Whether this recommendation has general validity remains to be determined, as it cannot be ruled out that so far understudied species may turn out to be ‘very bad keepers’. Based on the present findings, CGN has decided to carry out first germination monitoring tests 25 years after regeneration, while keeping the option open to revise this policy if new relevant information becomes available.

Germination values of 77 % for cultivated and 57 % for wild material have been used by CGN as thresholds for decisions on accessing and regenerating seed lots, which are lower than the levels currently suggested in the draft updated genebank standards (CGRFA 2012). Deviations from the FAO recommendations are due to the fact that high germination levels are often difficult to reach, in particular for crop wild relatives. According to CGN’s quality management system (www.cgn.wur.nl/UK/CGN+General+Information/Quality+manual/), threshold levels of 80 % germination for cultivated and 60 % for wild material are used. However, due to misinterpretation of the statistical properties of CGN’s testing procedures in the past, a downward margin of 3 % has been applied. Following the present study, actual threshold levels have been brought in line with the aforementioned, intended levels.

In summary, the present findings support postponing the first germination monitoring test to 25 years after regeneration for a wide variety of crop species that are maintained under seed management procedures and storage conditions as practiced by CGN. Nevertheless, for the period thereafter, seed viability should be monitored because insufficient data are currently available for seed lots older than 25 years to predict their behaviour. Germination monitoring could be further optimized by sampling representatives of groups of similar accessions instead of testing each accession individually, for example on the basis of crop/species and regeneration year/location. Efficiency improvement would also be realized if the estimation error in germination testing could be reduced, an issue that is currently under consideration at CGN.

Acknowledgments

The work described in this study was part of the Programme for Statutory Research Tasks regarding Genetic Resources (WOT-03-436) and the Fundamental Research Programme on Sustainable Agriculture (KB-12-005.03-003) both funded by the Dutch Ministry of Economic Affairs, Agriculture and Innovation. The authors would like to thank Noor Bas, Roel Hoekstra, Frank Menting and Willem van Dooijeweert (CGN) for their contribution to the preparation of the study data. We are also grateful to Noor Bas, Willem van Dooijeweert, Steven Groot, Chris Kik, Bert Visser and two anonymous reviewers for their helpful comments to improve an earlier version of the manuscript.

Copyright information

© Springer Science+Business Media Dordrecht 2012