Growth stages were assessed regularly throughout the season. Straw height and ear length (cm) were measured in 50 randomly chosen stems per plot (BBCH 90) in order to evaluate morphological variation. Straw height was measured from the ground to the start of the ear, ear length was measured from the first full spikelet to the tip without awns.

Foliar diseases caused by fungal pathogens were assessed at BBCH stage 73/75. Non-green leaf area was estimated in % (1–100%). The three leaf levels of flag leaf (F), leaf below flag leaf (F-1) and leaf below F-1 (F-2) were assessed separately at six locations per plot.

To assess foot diseases (Fusarium spp., Pseudocercosporella herpotrichoides, Rhizoctonia cerealis), plant samples were taken at five to six points per plot (minimum 30 stems) with roots at BBCH 75. The lower stems were freed of soil and leaf sheaths and scored for foot rot symptoms based on the key of Bockmann [30] where 0 is healthy, 1 is symptoms on < 50% of the stem perimeter, 2 is symptoms on 50–100% of the stem perimeter, 3 is stem brittle/rotten (P. herpotrichoides only). Based on a pictorial key of symptoms [31] Fusarium root rot, P. herpotrichoides and R. cerealis were assessed separately.

Grain yield on a plot basis was measured in t/ha at 14% moisture content, additionally the thousand kernel weight (TKW) was measured in g at 14% moisture. Ear bearing tillers/m 2 were calculated from three rows of 1 m length. Plants were cut shortly before harvest in order to assess morphological traits.

Protein content (%) was calculated from the nitrogen content of the seeds (N [%] × 5.7), which was analysed in ripe seeds that were dried for 72 h at 60 ∘ C, milled (ultra-centrifugal mill, Retsch, Type ZM 2) and analysed in the elemental analyzer vario MAX CHN (Elementar Analysesysteme GmbH, Hanau, DE).

Hagberg falling number (HFN; sec.; ICC Method no. 107), sedimentation value (Zeleny; ml; ICC Method no. 116), and wet gluten (%; ICC Method no. 106/2) were analysed in the Aberham Laboratories, Großaitingen, DE. HFN was assessed in pooled samples in the first year of the trial and per plot in the second year. Sedimentation value and wet gluten were assessed in pooled samples from the four replications in both years.

Baking volume of test loaves (ml) was assessed using an internal method credited to Aberham Laboratories: test loaves were baked from wholemeal, no ascorbic acid was added but due to very high HFN of some samples the addition of malt flour was necessary to prevent the bread crust form liquefying. Baking volume was assessed per plot in the second trial year only. For a detailed rating system and its translation into a color code of the respective values see Table A1 in Appendix.

Foliar disease severity per plot was calculated as the means per leaf level. Means were weighed 4:3:3 for the flag (F) leaves, the F-1 and F-2 leaves, respectively to account for the greater contribution of the flag leaf to the total dry matter of ripe seeds compared to the lower leaves [32].

A foot disease severity index (DI) was calculated based on the severity classes as:

where x 1 …x 3 are the number of stems with disease scores 1 to 3, respectively, and n is the total number of stems assessed. The resulting index values fall between 0 and 100 and can be calculated for each of the three foot diseases separately or as an index of all three together.

The statistical calculations were performed using IBM SPSS Statistics (Version 22). Data were tested for normal distribution of residuals (Shapiro-Wilk-Test and Q-Q-plots) and for homogeneity of variance (Levene test) and transformed if required. When data were normally distributed and variance was homogeneous, a univariate ANOVA with subsequent Tukey-B-Test was calculated where appropriate to find significant differences between group means at p < 0.05.

Where normal distribution was the case but not homogeneity of variance, the Games-Howell post hoc test was used (foliar diseases in both trial years, total incidence of foot diseases in 2011/12, and ear length in both trial years). Linear contrasts were calculated to compare

Average temperature during the wheat growing season 2011/12 was 9.7 ∘ C, which is higher than the long-term average (2000–2013) of 9.3 ∘ C and the long-term average (1977–1994) of 7.9 ∘ C. During the growing season 2012/13 average temperature was between the two known long-term averages (8.5 ∘ C). Apart from two divergences and extremes in February/March 2012 and in February/March 2013, temperatures measured during the two growing seasons of the experiment from September 2011 to August 2013 roughly followed the 14-year trend from 2000 to 2013 (Figure 1).

The distribution pattern of the monthly precipitation, however, showed strong deviations from the long-term average. The average total annual precipitation from 1977 to 1994 was 619 mm, from 2000 to 2013 it was 684 and in 2012 and 2013 it was 792 and 657 mm, respectively. There were very dry periods in November 2011, February and March 2012 and in spring 2013, and some extremely wet months in winter 2011, summer 2012 and May 2013 (Figure 1).

The combination of extremes in winter 2011/12 exposed the plots to a severe winter. After two unusually mild and wet winter months temperatures suddenly dropped at the end of January 2012. Three weeks of black frost with minimum temperatures reaching down to - 15 ∘ C resulted in soil frozen to a depth of about 50cm. Although the number of frost days (= daily minimum temperature below 0 ∘ C) in February 2012 was not different than in other years, the number of days with daily maximum temperature below 0 ∘ C was higher in 2012 than it was in 2011 or 2013. Also average minimum and maximum temperatures ( - 9.9 ∘ C and - 5.7 ∘ C) were considerably lower in February 2012 than in the years before and after (Table 1). The lack of snow left the plants unprotected from these extremes.

In mid-February, temperatures increased again and March was warm (average monthly temperature 7.5 ∘ C which is 3.3 ∘ C above the 14-year trend of 4.2 ∘ C) and dry (precipitation was 15 mm, which is only 27% of the 14-year trend). These six relatively warm weeks of drought following the extreme cold worsened the effect of the cold and put surviving plants in the frozen soil under severe water stress. The CCP plots were noticeably damaged, but they recovered. However, most of the 20 parent varieties grown in 2011/12 next to the trial plots in two times replicated plots for seed multiplication, could not cope with the extreme climatic conditions and the severe winter resulted in winterkill in 16 out of the 20 varieties. On average only 33 plants/m 2 were left in the plots in April 2012 and only the four varieties ‘Bezostaya’, ‘Monopol’, ‘Renan’ and ‘Hereward’ survived with an average of more than 50 plants per m 2 (Figure 2). For winter wheat a density of 80 plants/m 2 or less is seen as an indicator for plowing the whole stand [33] and all plots of the parental varieties were abandoned.

Disease pressure in both years was low. In both years the dominant disease was Septoria tritici. In 2012, the average infestation of plants on the three top leaf levels was 14% (BBCH stage 73/75), in 2013 it was even lower (10%). In 2012, infestation rates ranged from 12% (CY I) to 17% (OY II), in 2013 disease ranged from 7% (‘Achat’) to 10% (CA I). There were no relevant differences among treatments in both years (data not shown).

For foot diseases, total incidence and disease severity indices (DI) were slightly higher in 2013 (2012: 13; 2013: 20). The contribution of the two high infection severity classes to DI was, however, low in both years (data not shown) and therefore, overall the plants could be considered almost healthy. In both years Fusarium spp. was the dominating foot disease (DI 11in 2012; DI 16 in 2013), followed by Pseudocercosporella herpotrichoides (DI 2 in 2012; DI 4 in 2013), and Rhizoctonia cerealis ranged last in both years (DI 1 in 2012; DI 0.4 in 2013). There were only small differences among populations and references. A statistically significant difference in overall DI and Fusarium infestation between the CCP o ⁢ r ⁢ g and CCP c ⁢ o ⁢ n ⁢ v is considered biologically not relevant and was disregarded (data not shown).

In 2012, overall straw length was considerably lower than in 2013 (77.2 cm vs. 90.5 cm, respectively). Overall, the CCPs were significantly shorter than the reference varieties in 2012 but not in 2013 and significantly taller than the mixture of the parental varieties in both years. The QCCPs were always significantly taller than the YCCPs (Table 2).

As expected, within-plot variation of straw length was in both years less for the references than for the CCPs and the mixture. As the references are pure line varieties, within-plot variation of plant height is very limited. The CCPs in contrast, originating from the intercrossing of several parental varieties of different height, show considerable variation in plant height. In 2012, the population CYQ II was tallest (85.0 cm), CY I was the shortest CCP (69.7 cm), and the mixture was even shorter (64.6 cm). CY I and CY II, although significantly taller than the mix of parents, were shorter than the other CCPs and references. All four YCCPs were shorter than the mean height of plants in the trial while all YQCCPs, QCCPs and the references were taller than the mean (Figure 3).

In 2013, ‘Capo’ was significantly tallest (99 cm), the mix of parental varieties was shortest (65 cm). The two other references were also very short (‘Achat’ and ‘Akteur’ with 86 and 87 cm respectively). While ‘Capo’ was tall or tallest in both years, ‘Achat’ and ‘Akteur’ changed in terms of their ranges in straw length values. While ‘Achat’ and ‘Capo’ were considerably shorter in 2012 than in the year after, absolute height of ‘Akteur’ changed only very little (83 cm in 2012 vs. 87 in 2013) and its change of position in the range of varieties and CCPs is only due to the overall taller plants in 2013.

In the group of CCPs, CY I was the shortest in 2013 (88 cm) as it was in 2012, followed by the three other YCCPs. Again, all YCCPs were shorter than the mean height of plants in the trial, forming a subgroup that was statistically distinguishable from the group of the taller YQCCPs and QCCPs (Figure 3, Table 2).

Variation in ear length of the references was similar to the variation in the CCPs. In 2012, ears varied between 8.1 cm (CQ II) and 9.9 cm (‘Akteur’), with a mean of 8.8 cm in the trial and no statistically significant differences (data not shown). In 2013, ear length varied between 8.7 cm (‘Capo’) and 10.2 cm (‘Akteur’), with a mean of 9.1 cm. In this year ‘Achat’ with 10.1 cm and less variance than ‘Akteur’ had the statistically longest ears. Overall, ear length of the references was significantly greater than that of the CCPs in both years (Table 2).

For the Hagberg falling number (HFN) values < 180 and > 280 are considered poor with values in between 240–280 good and 180–239 moderate. The other quality parameters (protein content, sedimentation value, wet gluten, baking volume) are usually assigned to three to six class values. Where the rating is done in three classes, values are grouped into the classes good, moderate and poor; based on these classes the cells in the overview table (Table 4) are color coded, with green indicating good, yellow indicating moderate and red indicating poor, in addition to listing the measured values. More detailed ratings can be done for some parameters with classes ranging from very good to inacceptable, these classes are described in Table A1 in Appendix.

HFN, which was done for pooled samples in 2012 and by replicate in 2013, was rather high in both years, with an average of 292 sec. in 2012 and 282 sec. in 2013. Sedimentation values were extremely good in 2012 (41 ml on average) and 32 ml in 2013, which is still good, although sedimentation values for several samples were lower (Table 4). Wet gluten was higher in 2012 (average: 28.5%; good) than in 2013 (average: 26.3%; satisfactory). The mean protein content [%] in the trial was medium in both years (12.1% in 2012 and 11.3% in 2013). Baking volume assessed in the second year of the trial was 383 ml on average, which is a good result for wholemeal test loaves. Volume ranged between 344 ml (OY II; satisfactory) and 428 ml (OQ I; very good; Table 4).

In general, it could be observed that in both years YCCPs were clearly separate from the other populations and varieties with the YCCPs ranging lowest for all baking quality parameters tested. The QCCPs were in both similar to the reference varieties, which is also consistent for all parameters except protein content in 2013, where QCCPs had a significantly higher protein content than the references. The YQCCPs ranged in both years between the other groups of populations and varieties regarding all values tested and also the finding that CCP o ⁢ r ⁢ g and CCP c ⁢ o ⁢ n ⁢ v did not differ is generally true for both years and all parameters tested (Table 5). The mixture of parents, which yielded very low in both years, showed much better results regarding baking quality parameters.

Values for protein content, HFN, baking volume, wet gluten as well as sedimentation value were close to the average of the trial in both years (Table 4).

When comparing groups (Table 5) the significantly higher baking volume of QCCPs was confirmed. YQCCPs ranged in the middle and YCCPs had the lowest baking volumes. Comparing the CCPs with the references, volume of references was significantly higher. A comparison of CCP o ⁢ r ⁢ g and CCP c ⁢ o ⁢ n ⁢ v yielded no relevant differences, also the difference between QCCPs and references is not significant.

For HFN in 2013, the comparison of CCP groups showed the statistically significant highest HFN for the group of QCCPs (average HFN of group 310 sec.) followed by YQCCPs (average HFN 262 sec.), followed by the significantly lowest group of YCCPs (average HFN 205 sec.). While an average HFN of 310 sec. is considered poor (too high), 262 sec. is good, and 205 sec. is moderate. The references had a significantly higher HFN (average HFN 370 sec.) than the CCPs, which is extremely high and thus poor.

For protein content the comparison of groups showed in 2012 a significantly higher protein content of the CCPs vs. references and higher protein content of QCCPs compared to YCCCPs, YCCPs, and references. In 2013, protein content of QCCPs was higher than the group of YCCPs and the group of references (Table 5).

Disease pressure was low and thus did not play a role for the performance of the CCPs or the references during the two experimental years. Overall, there was neither an influence of the choice of parents nor of the growing system visible. Parents were chosen with the focus on yield and baking quality and not in order to represent different disease resistances, therefore it is unlikely that the CCPs initially differed very much regarding their resistances. Disease pressure in the growing environment where the populations evolved was moderate and did not differ much between the organic and conventional growing area, this meant a strong differentiation of populations was not expected.

Higher disease pressure might have resulted in a different picture as the results of other experiments indicate. Observations of powdery mildew (Blumeria graminis f. sp. tritici) in wheat CCPs revealed that the frequency of B. graminis-resistance genes evolved differently according to the respective disease pressure [34, 35, 36] and Webster et al. [37] found that frequencies of Rhynchosporium secalis-resistance genes in a composite cross of barley changed between F 5 and F 45 in accordance with the respective disease pressure. In years when high pressure was recorded the frequency of the resistance genes rose, in years with low pressure, it fell.

Observations in stripe rust (Puccinia striiformis) in a wheat experimental population in France documented that the resistance gene Yr17, which provided complete resistance to stripe rust until 1997 and was thus suspected to be under strong selection, was indeed selected between generations 5 and 10 [38].

Since 2011, new races of stripe rust have made a dramatic appearance throughout Europe [39] and the main foliar pathogen observed since 2014 in the trial site is stripe rust. In comparison to the susceptible varieties ‘Akteur’ and ‘Naturastar’, disease severity on the CCPs has been very low [40].

The CCPs as well as the references could not reach their full height potential in the first year due to the extreme weather conditions. The same was reported from regional variety trials, where the average plant height of winter wheat grown without growth regulators in 2012 was reported to be only 87 cm [41].

The parents were equally short in both years as they were mostly dwarf types. In contrast, the CCPs were much taller indicating that the dwarfing genes have decreased in frequency. They might not have been eliminated completely though, as variation for this trait is still quite large. Nevertheless, the CCPs were within the normal height range; they were shorter than the references in the first experimental year and about the same height in the second year.

Findings of Goldringer et al. [27] and Le Boulc’h et al. [26] observing an increase in plant height cannot be confirmed. This could be due to the fact that the tallest plants ( > 130 cm) were removed from the populations in several successive years to limit their selective advantage. We conclude that the “good practice” of removing the tallest plants in an evolutionary population may improve their agronomic value. It might, however, have obscured any effects of natural selection on plant height.

Morphological characteristics of the parental varieties were documented in 2007 [42]. In that year, height of the yield parents was 87.5 cm while the quality parents were 97.1 cm tall on average. Thus, the significantly shorter straw length of the YCCPs compared to the other CCPs are founded in the original composition of the CCPs and should not be understood as divergent developments of the populations over time.

Measurements in the F 5 - F 9 also showed these differences in plant height of the CCPs [43]. Ear length has not previously been measured in the parental varieties. However, as the results show only marginal differences between ear length of references and populations, an influence of the parental varieties is unlikely. An influence of the two growing systems on straw height and ear length was not found.

Ear-bearing tillers were at the same low level for all CCPs, the mixture and the references without large variation in summer 2012, which shows that the winter conditions influenced all plots in a similar way resulting in overall low yields. Nevertheless, the resilience of the CCPs and reference varieties was remarkably higher than for most of the parents (Figure 2). Considering the poor survival of the parents in pure stands, the performance of the mixture in 2012 was impressive, demonstrating the general positive effects of mixtures over pure lines as has been shown on many occasions before [7, 44].

Based on previous year’s results [45] and because they were composed from high-yielding varieties, the YCCPs were expected to yield better that the other CCPs. However, in 2012 they yielded lowest of all CCPs. To explain this, the parental varieties used to create the CCPs have to be taken into account. Of the 20 parent varieties only the four varieties ‘Bezostaya’, ‘Monopol’, ‘Renan’ and ‘Hereward’ survived the winter reasonably well (Figure 2). As the CCPs were composed in the UK, 14 out of 20 parent varieties were of English origin and thus bred for a maritime climate. ‘Bezostaya’, however, is of Ukrainian origin, has high grain yield and quality, good frost resistance and is often used in crossing where winter hardiness is a desired trait[46]. ‘Monopol’ comes from Germany and ‘Renan’ is French [47], only ‘Hereward’ is an English variety.

A closer look at the pedigree reveals also here a German winter wheat variety—‘Disponent’—as a crossing partner [23] which most likely provided ‘Hereward’ with a certain degree of winter hardiness. Of these four varieties with good winterhardiness, only ‘Bezostaya’ was intercrossed into the YCCP, which most likely explains why the winter conditions affected the YCCPs more than the other populations. While it is possible that selection for greater winter hardiness occurred at the German site, this cannot be concretely concluded without direct comparison of early and late generations for this trait, or of populations that have undergone evolution in different climatic conditions.

The comparably good yield of ‘Achat’, ‘Akteur’, and ‘Capo’ in 2012 is most likely owed to their relatively good winter hardiness and to the fact that good winter hardiness was not one of the main traits in focus when establishing the CCPs. It remains to be seen if the CCPs respond better to freezing after having survived one especially cold winter. As we used the same seed in both years the winter effects did not affect the performance in the second year. Results from experiments investigating the effect of natural selection on the winter survival of barley CCPs indicate that natural selection did increase winter survival although not uniformly over different generations [48]. In bulk populations of winter oats an improvement in winter hardiness could only be found in populations with low initial survival levels [49, 50]. Also, apparent advances made in winter survival in one year can reverse in later generations due to a lack of competitive ability of the hardy types later in the growing season [49], when non-hardy types that were not eliminated resurface and restore themselves as major components in the population[48]. This shows that complex traits such as winter hardiness, that were not a main focus when establishing CCPs, are hard to achieve through natural selection only.

In 2013, yield of the YCCPs corresponded with expectations being 0.3 t/ha higher than the QCCPs and YQCCPs. These differences were, however, not statistically significant. Yield of the CCP o ⁢ r ⁢ g and CCP c ⁢ o ⁢ n ⁢ v varied minimally with no indication that their maintenance in different growing systems has led to strong variation between the two groups of populations regarding yield performance. Higher numbers of ears of the YCCPs was related to the higher yielding capacities of these populations. In contrast, the high yields of the references ‘Capo’ and ‘Akteur’ were due to high TKW and high number of seeds per head, respectively. This is in contrast to what was previously published by the seed producing industry. ‘Capo’ is known as a density type realizing yields through many tillers and ‘Akteur’ as a single ear type, forming many seeds per ear with high TKW [51].

A higher TKW was the only parameter that separated the CCP o ⁢ r ⁢ g from the CCP c ⁢ o ⁢ n ⁢ v in 2012. In the second year, absolute differences where at the same—low––level, the difference was, however, not statistically significant. Apart from this observation there was no field evidence that the differing environments of an organic and a conventional farming system could have shaped the CCPs in different ways. However, a study using hydroponics and bioassays to investigate early vigour and allelopathy in the F 6 and F 11 of the CCP o ⁢ r ⁢ g and CCP c ⁢ o ⁢ n ⁢ v , documented systems’ effects on the CCPs.

Characteristics for early vigour were improved after five years in the organically managed CCPs in comparison to the conventionally managed CCPs. The changes towards early vigour in the organic CCPs are thought to be due to the combined effects of selection for higher nitrogen uptake under low-input conditions, and increased competition for light and larger seeds, rather than a direct adaptation to higher weed pressure [52].

As baking tests are rather costly and time-consuming, various indirect parameters such as sedimentation value, wet gluten, protein content and falling number are often used to predict the baking properties of wheat flour. It has been assumed that protein and wet gluten content strongly correlate with the baking volume determined in the RMT. This is, however, not always the case [53]. In whole-meal-baking tests protein content, sedimentation value and wet gluten content often only have a very limited influence on the baking volume [54].

In our study indirect baking quality parameters were analyzed in both years while baking tests could only be conducted in 2013. The results for protein content and HFN in 2013 were in accordance with the outcomes of the baking tests while wet gluten and sedimentation value were less suitable to predict the baking test outcome. The results show a clear differentiation of groups based on the original composition of the populations for all parameters, except wet gluten.

Baking tests are usually done with the rapid mix test (RMT), which is the usual procedure when testing superfine flour. The RMT is, however, not optimized for the processing of organically produced wheat [55] and considering this, the baking test done in 2013 to assess baking volume of the CCPs was done with wholemeal test loaves.

For a wholemeal baking test the average volume of loaves of 383 ml is a good result. Baking with wholemeal flour, lower volumes are the norm and a volume of 400 ml or above is considered very good, 350 to 400 ml is good, below 350 ml is moderate and 330 ml and below is poor (pers. comm. Dr. R. Aberham).

In the test, all CCPs and references except OY II ranged above 350 ml. The strong differences between varieties that can be observed with white flour are less pronounced when testing with wholemeal flour [56]. In this way the results are more likely to correlate with results bakers producing organic bakery products would achieve. The high volumes of the QCCPs compared with YCCPs or YQCCPs indicate that the original choice of parental varieties still has an effect, while adaptation to the farming systems seems to have had no effect on baking volume. The same was true for protein content, falling number, and sedimentation values. In contrast, for wet gluten the influence of parents is not as clearly visible as for the other baking quality parameters. Overall, the QCCPs that were specifically created for good baking quality, are as good (baking volume) or better (protein content, HFN) than modern elite wheat varieties.

While yield is a trait that is subject to natural selection [15, 57], quality traits are not directly influenced by natural selection [15]. Without the genetic base of high-quality parents the breeding objective of high baking quality cannot be reached [58]. Including a parent with low baking quality in the setting up of a high quality CCP can be enough to counteract the high quality parents as some individuals with low quality will prevent the population as a whole from sustaining high quality [15]. Results from trials with variety mixtures show other patterns, however. In a mixture of two wheat varieties a higher total aerial biomass was achieved than was produced by each variety grown in a pure stand. This increase resulted in a grain yield similar to that one of the higher-yielding variety and an improved protein content was measured [59].

The crossing design of the CCPs developed by the John Innes Centre and Elm Farm Research Centre took it into account that quality traits are not subject to natural selection. As opposed to the early composite cross populations of wheat and barley [60, 61], which were established with the aim of representing the major wheat or barley growing areas of the world in order to assemble genotypes appropriate for each cultural practice in the respective agro-climatic zone [15], the focus was narrowed to yield or quality as key characteristics of the CCPs. The results show that the quality traits were successfully inherited and maintained over time and that acceptable yield levels were also achieved not only in the populations designed to be high-yielding, but also in the high-quality populations which were not much different in yield from the high-yielding populations in the second experimental year. By using seed of the same generation in both years, these genetic effects could be clearly separated from the lack of winter hardiness in the YCCP parentage.

Looking at the yield and quality achieved by the mixture of parents a contrast of low yield in both years, but good quality becomes visible. The CCPs out yielded the parental variety mixture in both years. Here the populations seem to have a clear advantage over the mixture. The overall higher diversity and/or natural selection and adaptation over time may be responsible for this. For the quality aspect natural selection played – as mentioned above – a minor role and QCCPs and parents continued to perform similarly after a decade of selection.