Breeding progress for pathogen resistance is a second major driver for yield increase in German winter wheat at contrasting N levels

Breeding has substantially increased the genetic yield potential, but fungal pathogens are still major constraints for wheat production. Therefore, breeding success for resistance and its impact on yield were analyzed on a large panel of winter wheat cultivars, representing breeding progress in Germany during the last decades, in large scale field trials under different fungicide and nitrogen treatments. Results revealed a highly significant effect of genotype (G) and year (Y) on resistances and G × Y interactions were significant for all pathogens tested, i.e. leaf rust, strip rust, powdery mildew and Fusarium head blight. N-fertilization significantly increased the susceptibility to biotrophic and hemibiotrophic pathogens. Resistance was significantly improved over time but at different rates for the pathogens. Although the average progress of resistance against each pathogen was higher at the elevated N level in absolute terms, it was very similar at both N levels on a relative basis. Grain yield was increased significantly over time under all treatments but was considerably higher without fungicides particularly at high N-input. Our results strongly indicate that wheat breeding resulted in a substantial increase of grain yield along with a constant improvement of resistance to fungal pathogens, thereby contributing to an environment-friendly and sustainable wheat production.

Grain yield and yield components. Mean grain yield (GRY) across all treatments was 6.4 ± 2.0 t ha −1 ( Table 2). Regarding yield components, mean thousand kernel weight (TKW) was 34.6 ± 5.9 g, kernels per spike (KPS) 38.7 ± 11.7, ears per square meter (ESM) 492 ± 84 ears m −2 , biomass (BM) 17.8 ± 3.6 t ATM ha −1 , and harvest index (HI) 0.34 ± 0.06 (Table 2). GRY, BM and all direct yield components showed substantial genetic variation (G, p ≤ 0.001). Treatments resulted in highly significant differences for GRY and most of the yield components except ESM as direct comparisons by orthogonal contrasts show (Tables 1, 2). Plant protection (PP, T1, T3 vs. T2, T4) significantly increased BM (+ 20.8%), GRY (+ 40.9%), and all yield components (TKW + 20.0%, KPS + 17.3%, ESM + 1.1%), as well as HI (+ 15.9%). Elevated N fertilization (T1, T2 vs. T3, T4) resulted in significantly higher BM (+ 4.3%) and ESM (+ 2.4%), whereas TKW (− 8.8%), HI (− 8.6%), and GRY (− 3.4%) were significantly decreased. KPS (+ 1.7%) was not significantly affected by N application. While a higher N supply together with plant protection results into higher GRY (T2 7.4 t ha −1 vs. T4 7.7 t ha −1 ), higher susceptibility to fungal pathogens at a high N rate resulted in lower GRY (T1 5.7 t ha −1 vs. T3 5.0 t ha −1 ). Increased GRY due to higher N under plant protection (T4 vs. T2) was much less than expected. While lodging (T4 > T3 > T2 > > T1) might have contributed to lower GRY at higher N the results can be explained by the influence of the year. Year (Y) significantly affected GRY, BM, and all yield components mainly due to annual differences in the intensity of drought periods during spring and summer. While GRY (7.8 t ha −1 ), KPS (45.6), BM (19.1 t ATM ha −1 ), and HI (0.37) were highest in 2016, the year 2015 with the strongest drought resulted in the lowest average performance (GRY 5.2 t ha −1 , BM 15.2 t ATM ha −1 , KPS 28.7, HI 0.30). ESM was significantly increased in 2015 (538 ears m −2 ) likely resulting from high N min and above average precipitation in winter 2014/15. N application was likely only partly successful in the drought stress years 2015 and 2017. Moreover, early N application might have been contributed to drought stress by higher tiller competition (ESM higher in T3 and T4 than in T1 and T2, c.f. Tab. 2). Higher N together with higher soil moisture in 2016 resulted in a more normal 14% increase in GRY in T4 vs. T2. Stressed conditions particularly at higher N rates in both drought years are also supported by harvest Impact of major diseases and genotypic resistance on yield and yield components. Results of correlation analyses between disease and yield traits at both N levels are summarized in Table 3 and Fig. 1. In general, factorial relation and correlation between parameters of diseases, grain yield and yield components were similar at both N levels. Factor PCA indicated that GRY, BM, HI and TKW formed a cluster at the first two principal components, which explain 39 to 44% (PC 1 at 110 and 220 kg N ha −1 ) and 23% or 22% (PC 2 at 110 and 220 kg N ha −1 ) of the phenotypic variance. Accordingly, GRY, KPS, BM, and HI correlate very strongly with each other at both N levels. In contrast, yield component ESM shows only a weak positive correlation to GRY, BM, and HI at both N levels but no correlation to KPS and a weak negative or no correlation to TKW. With regard to diseases significant positive correlations were observed between YR and FHB (110 kg N ha −1 r = 0.25, 220 kg N ha −1 r = 0.46) and between YR and PM (110 kg N ha −1 r = 0.42, 220 kg N ha −1 r = 0.22). Highly significant negative correlations were detected between YR and LR (110 kg N ha −1 r = − 0.40, 220 kg N ha −1 r = − 0. 45) indicating strong competition between the two rust pathogens. Correlations between LR and the other diseases as well as between LR with GRY and the yield components were relatively weak or not significant, likely due to the abundance of YR. Pairwise significant negative correlation coefficients between all fungal diseases taken together (Fungal sum) and GRY at both N rates (110 kg N ha −1 r = − 0.67, 220 kg N ha −1 r = − 0.76) clearly demonstrate the impact of fungal diseases on yield. In other words, approx. 50% of the differences in grain yield can be explained by the additive or combined effect of the fungal pathogens. Likewise, the cluster of GRY and yieldrelated parameters had a factor load opposite to YR at both N levels and additionally opposite to PM at low N and FHB at high N ( Fig. 1). In detail, highly significant negative correlations were observed primarily between GRY and YR (110 kg N ha −1 r = − 0.80, 220 kg N ha −1 r = − 0.86) and between GRY and FHB (110 kg N ha −1 r = − 0.42, 220 kg N ha −1 r = − 0.64). Correlations between GRY and PM were highly significant but lower, particularly at higher N rate (110 kg N ha −1 r = − 0.60, 220 kg N ha −1 r = − 0.36). Correlation between LR and GRY (r − 0.37) was obtained by partial correlation analysis taking into account those cultivars with low YR infestation (YR 1 < 4.0455%, low or no competition between YR and LR). Partial correlation coefficients between LR and GRY are in general declining in classes with higher YR infestation and are not significant in class YR 5 (> 13.242%) supporting the notion of a YR dominance effect in YR susceptible cultivars (cf. Tables 3, 4). Partial correlation coefficients between LR and GRY are in general declining in classes with higher YR infestation. www.nature.com/scientificreports/ All leaf diseases, including LR at 220 kg N ha −1 and low YR abundance, were also negatively correlated with the yield component KPS, with BM, and HI and to a lesser extend with TKW. In comparison, FHB shows higher correlations mainly with GRY, TKW, and BM at low N, but additionally also with KPS at high N supply. ESM is slightly negatively correlated with YR (110 kg N ha −1 r = − 0.33, 220 kg N ha −1 r = − 0.26), significantly but weaker with PM (110 kg N ha −1 r = − 0.19, 220 kg N ha −1 r = − 0.29), and not with LR (class with lowest YR infestation) and FHB. PM correlates only moderately negatively with TKW at low N, but no correlation was observed at high N. That is in contrast to the other fungal diseases, which show negative correlations with TKW at both levels of N input.
Bivariate analysis indicates substantial yield losses resulting from fungal diseases ( Table 4, Fig. 2). Using the slope of the regression line as an estimator, per basis point of visible pathogen symptoms (on leaves or spike) an average GRY loss of 102 kg ha −1 at 110 kg N ha −1 resp. 123 kg ha −1 at 220 kg N ha −1 is expected. More specifically, average GRY loss per % tissue damage caused by YR is 135 kg ha −1 (110 kg N ha −1 ) resp. 153 kg ha (220 kg N ha −1 ). LR (YR 1 ), based only on cultivars with low YR infestation results in a loss of 50 kg ha −1 (110 kg N ha −1 ), resp. 70 kg ha −1 (220 kg N ha −1 ). PM (172 kg ha −1 at 110 kg N ha −1 , 187 kg ha −1 at 220 kg N ha −1 ) and FHB (217 kg ha −1 at 110 kg N ha −1 and 305 kg ha −1 at 220 kg N ha −1 ) cause more severe yield losses than both rust diseases.
Data indicate a steady increase in yield gain due to fungicides (T4 minus T3; T2 minus T1) along with an increasing sum of disease scores (Fungal sum) at both N levels. Even the most resistant cultivars show a positive response to fungicides. If a linear dependency between yield gain to fungicides and the sum of disease scores is assumed (linear fit: T2-T1 = 0.642 + 0.0473*T1_fungal_sum; V4-V3 = 0.999 + 0.0712*V3_fungal_sum), a yield gain due to fungicides alone of + 0.64 t ha −1 at the low N rate and + 1.00 t/ha −1 can be inferred.

Discussion
It is well known that the rate of nitrogen supply can alter the susceptibility of crop plants to fungal pathogens, but the effects depend on the host, the type of pathogen (biotrophic vs. necrotrophic) 59,60 , and the time of N application 61 . As the impact of N supply on crop diseases is complex, it was argued that it should be investigated in a crop-and pathogen-specific manner 62 .
In the present study four major fungal pathogens, varying in their phenological niche and severity, were used to assess a panel of historically and agronomically important cultivars of German winter wheat at two levels of N supply, resembling relevant low and high input conditions. Resistances and yield traits were analyzed at each N level and in control treatments in which diseases were excluded by extensive fungicide spraying. Substantial variation was detected for all resistances and yield traits. It confirms previous findings that N fertilization tends to increase the number of ears per m 2 , total biomass, kernels per spike, and grain yield, but reduces TKW if fungal diseases are absent 58,63 . As ears per m 2 and biomass are elevated at higher N supply, it can be concluded that also the microclimate was more favorable to fungal pathogens. Accordingly, infestations by all four fungal pathogens analyzed were responsive to the N level (Table 1). Our results support the previous finding that biotrophic fungal pathogens (i.e. Blumeria graminis, Puccinia striiformis, P. triticina)) benefit from increased N supply 13,64,65 . The hemibiotrophic pathogen Fusarium culmorum (with a biotrophic and a necrotrophic phase) increases strongly with higher N as well, supporting the findings of 40 who reported increasing FHB severity with higher N fertilization rates from 0 to 160 kg N ha −1 (independent from the type of fertilizer). Although we found a broad genotypic variation among the cultivars ranging from nearly resistant to highly susceptible to all pathogens tested, G × N interaction was not detected for stripe rust (YR), leaf rust (LR), and powdery mildew (PM) infestation. That suggests that the cultivars analyzed respond similarly to increased N fertilization.
The abundance of YR was relatively high, which can be attributed to the fact that a mixture of relatively new and aggressive P. striiformis races such as 'Warrior' , 'Warrior (−)' , 'Triticale aggressive' and 'Oakley, v7/Kranich' have been used for inoculation. Most resistance genes deployed in the European breeding genepool are overcome by these highly virulent races, including Yr6 by all four races, Yr8 and Yr10 by 'Triticale aggressive' , Yr1, Yr2, Yr3, Yr7, Yr9, Yr17, Yr25 and Yr32 by 'Oakley, v7Kranich' and the Warrior races, and YrSd as well as YrSP by the Warrior races 21,66,67 . There was at least some YR infestation in even the most resistant cultivars, e.g. 'Desamo' ,  Table S3). As qualitative resistance should in general not result in infestation quantitative resistance most probably caused YR differences detected on the panel. This observation is in accordance with Hovmøller 66 who assumed that the qualitative resistance genes Yr5, Yr24, and Yr15, which are still functional, are very likely not deployed in German elite wheat varieties. Some genes are known to confer to quantitative resistance against YR and LR, namely Lr34, Lr46, Lr67, Lr68, and likely also Yr17. It is not known if German cultivars contain the less common quantitative resistance genes Lr67 and Lr68 as breeders often hide such information and differential sets to study Lr genes usually do not cover Table 5. Bivariate analysis between diseases (YR, LR including LR by YR classes, PM, FHB), and yield (GRY) with years of cultivars release based on LSmeans (observations treatment-specific, over years and repeats) assuming a linear model. Results contain phenotypic variance explained (R 2 ), significance levels of p values of each model tested, equations of linear fit to estimate breeding progress from year of release (YoR), and relative annual improvement of each parameter between 1965 and 2013 (estimated value of 1965 set to 100%). T1-T4, treatment 1 to 4; YR, stripe rust; LR, leaf rust; YR 1 -YR 5 , quantile classes of YR infestation (YR 1 < 4.0455%, YR 2 4.0455 to 6.1509%, YR 3 6.1509 to 9.1852%, YR 4 9.1852 to 13.242%, YR 5 13.242 to 31%); PM, powdery mildew; FHB, Fusarium head blight; GRY, grain yield; Signif. p, significance level of p value: *** < 0.001, ** < 0.01, * < 0.05, ns not significant.    together with further quantitative resistance loci/genes have been combined in the German elite cultivars as KWS released many successful cultivars with good overall resistance after the year 2000 and other breeders are allowed taking advantage from resistant cultivars through the breeder's privilege. Leaf tip necrosis (LTN) has been described to be associated with some of these quantitative resistance genes. LTN occurred at varying degree in most of the cultivars in 2015. It significantly increased in both protected treatments, T2 and T4, over the five decades along with the increase in disease resistance. Thus, increased LTN could result from increased disease resistance due to a combination of quantitative resistance genes. Higher N fertilization increases the susceptibility of wheat against YR supporting previous studies 13, 14 , but G × N interaction was not shown to be significant. In other words, the cultivars react very similar to higher N supply. Nevertheless, with respect to breeding progress varietal YR susceptibility was reduced considerably in the course of decades from an average of 11.4% leaf area infected to 5.9% at 110 kg N ha −1 and from 12.5% to 6.1% at 220 kg N ha −1 (cf. Fig. 3). Although the absolute reduction rate was steeper at the higher N level, the rate of relative improvement (starting in 1965 with the infected leaf areas each set to 100%) equals to a constant rate of 1.0% per year at both N levels. The significant decrease of infestation by YR despite continuous breakdown of previously introduced resistance genes during the observed time interval supports the hypothesis that at least some of the R genes, which mediate qualitative resistance, may contribute also to quantitative resistance already 21,68 . Additionally, accumulation of durable quantitative YR resistance loci may have contributed to improved resistance in the field 69 .
In comparison to YR a lower average leaf rust (LR) susceptibility was observed, but the yearly abundance of both rusts differed considerably (cf. Table 1). The increase of mean LR infestation due to higher N fertilization was pronounced which is in agreement with results of previous studies [22][23][24] . Being inoculated onto the same wheat plants LR mainly developed on cultivars with a low susceptibility to YR. As a result, LR and YR disease expressions showed a strong negative correlation. It may be hypothesized that LR infestation is competitively suppressed by YR as both foliar fungal pathogens rival for a very similar ecological niche where YR has the advantage of first occupation. Competition is supported by the fact that both diseases do not correlate with each other if YR abundance is low, and partial correlation coefficients decrease along with an increase of YR infestation.
Older cultivars in the panel were more often infested by YR and because of the negative correlation to LR do not allow an unbiased estimate of LR infestation. Hence, partial correlation analysis of cultivar classes with similar YR susceptibility (Fig. 4) shows that LR resistance has been strongly improved over the past five decades as well. Relative improvement of LR resistance at a rate of 1.1% per year is independent from N input and only slightly higher than the improvement against YR. Ahlemeyer and Friedt 51 found a lower relative annual raise of resistance at 0.6% for LR between 1966 and 2007 based on data from official trials by the German Plant Variety Office analyzed over that period. The difference might be attributable to the different trial types. Our rates of improvement are based on joint trials which subsume durable resistance progress and resistance which has been lost by breakdown of resistance genes 54 . The calculations of Ahlemeyer and Friedt 51 were based on data of continuing official German variety tests for cultivation and use (VCU 70 ) and reflect durable resistance progress only. If durable progress is assumed to be equal for both types of trials, subtracting the rate of durable resistance www.nature.com/scientificreports/ progress from the combined rate gives a rough annual estimate for the relative annual breakdown of resistance genes of 0.5% for LR. Susceptibility to powdery mildew (PM) was also considerably increased by higher N fertilization. As plant density and biomass was increased due to higher N, it strongly supports previous reports according to which dense sowing and high N supply can increase susceptibility to PM 23,[71][72][73][74] . Although no artificial inoculation with PM was carried out, natural infection occurred regularly with at least low abundancy in late fall of each year. The disease severely progressed under humid and warm weather conditions (e.g. in May and July 2017) when even the least susceptible cultivars such as 'Tabasco' , 'Zappa' or 'Memory' showed diseases symptoms. PM remains a constant threat to wheat despite the fact that resistance over the past five decades was improved strongly with a decrease of susceptibility from 8.8% to 2.5% leaf area infected at 110 kg N ha −1 and from 10.1% to 3.4% at 220 kg N ha −1 (cf. Fig. 3). That equals to a relative improvement of 1.49% a −1 and 1.39% a −1 , respectively. The relative rate of improvement in resistance according to Ahlemeyer and Friedt 51 was 1.46% a −1 , almost identical with the rate which we observed. Considering that the resistance progress levelled off in the mid 2000′s, recalculation gives a rough estimate of 0.7% (110 kg N ha −1 ) resp. 0.5% (220 kg N ha −1 ) for PM to compensate for the average annual rate of breakdown of resistance, similar to the value for LR. Our results support previous studies, that intense wheat resistance breeding achieved sustained resistance 4,52,58 .
The N application rate had a significant impact on the susceptibility to Fusarium head blight (FHB), as infestation at 220 kg N ha −1 was considerably higher than at the lower N supply. FHB symptoms developed rapidly after artificial inoculation with the highly virulent isolate Fc46 75 adjusted to the flowering time of cultivars and followed by two subsequent water sprayings. We found significant G × N interaction for FHB suggesting that cultivars react differently to the pathogen at different N levels. This is in contrast to previous studies, in which no significant effects of varying N rates on FHB susceptibility 43,44 or even contradicting data 42 were reported. As expected, only quantitative differences of susceptibility were detected. In contrast to YR, PM, and LR, were modern cultivars mostly belong to the least susceptible, breeding progress for FHB resistance is not obvious by years of cultivar's release of the best in comparison to the most susceptible cultivars. However, average infestation decreased slightly over time from 2.5% to 2.1% at 110 kg N ha −1 and from 3.3% to 2.7% at 220 kg N ha −1 (Fig. 3). If a linear trend is assumed, this results in a relative annual decrease in susceptibility of 0.33% a −1 at 110 kg N ha −1 and 0.37% a −1 220 kg N ha −1 . FHB resistance of European cultivars is predominantly quantitative, i.e. mediated by many loci with small effects 38,76 . It is generally argued that quantitative resistance is more durable and a (gradual) breakdown of such resistances is unlikely. If the latter would be assumed for the recalculation of the relative annual improvement, durable breeding progress for FHB resistance would be rather low.
The intensity of fungal diseases as a whole correlates significantly negatively with yield and single yield components ( Table 3). The negative effect of fungal pathogens on the different yield components corresponds to the seasonal occurrence of the diseases. Early occurring diseases such as PM and YR significantly decrease crop density (number of ears per square meter) while late occurring diseases usually have no pronounced or significant effect. The number of kernels per spike is the yield component most strongly affected by the foliar diseases. This can be explained by the impairment of photosynthesis and reduction of assimilates impacting flower development, fertilization, and the development of ovules and intact seeds. In contrast, toxins related to the ear disease FHB mainly affect later stages of ovule development and seed formation, potentially leading to the abortion of developing ovules. FHB, YR, and LR have been very abundant mainly during grain filling, which explains their negative correlation with TKW. On the contrary, PM usually occurs earlier but also more permanently and seems to reduce grain yield primarily via the reduction of kernels per spike, biomass and harvest index.
A high N supply tends to promote ears per square meter and all four fungal diseases. In turn, correlations between ears per square meter and each of these diseases are generally low and difficult to interpret. Higher N supply leads to stronger negative correlation between YR, LR and grain yield, as well as for FHB with TKW and grain yield in comparison to a lower N supply. This supports the assumption of the source-based cause of a higher N supply leading to stronger disease infestation. Stronger infestation by fungal pathogens uses the plants' assimilates and negatively affects photosynthesis, and organ development, which in turn damages yield components resulting in lower biomass formation and finally a reduction in grain yield. The expected loss of grain yield differs for each of the diseases but is in general increased at higher N supply (cf. Fig. 2). In the present study, the biotrophic rust diseases have caused yield losses of 50 up to 153 kg ha −1 per 1% tissue damage with lower losses by leaf rust occurring late in the season. Yield losses resulting from powdery mildew infection were higher (172-187 kg ha −1 per 1% tissue damage) likely due to a longer duration of PM infestation in comparison to the rusts. FHB has not only caused the most severe losses of GRY (217-305 kg ha −1 per % of damaged tissue) via a reduction in TKW, it also releases toxins such as Deoxynivalenol (DON) und Zearalenon (ZEA) which can cause a reduction of kernels per spike, particularly at an increased N level. As it might also lead to a total loss of a wheat harvest by toxic contamination, resistance against this disease should be of high priority in further wheat breeding.
As demonstrated, grain yield could be enhanced over the whole time period of five decades in all four production systems (Fig. 5) with an average of + 38.3 kg ha −1 a −1 . Increasing grain yield over time without chemical plant protection (T1, T3) results both from enhanced genetic grain yield potential (T2 + 34.1 kg ha −1 a −1 , T4 + 35.5 kg ha −1 a −1 ) as well as from positive and stabilizing effects of enhanced fungal resistance (Fig. 5). Based on the estimated slopes, improved fungal resistance accounts for 11% (+ 3.8 kg ha −1 a −1 ) at low N input and 28.5% (+ 10.1 kg ha −1 a −1 ) of the grain yield increase under intense N conditions. Thus, the strongest improvement of German wheat cultivars has been achieved by enhancing both, yield potential and fungal resistance at high input conditions without plant protection (T3). We found evidence for a positive yield response to fungicides at both N rates even in largely resistant cultivars. More precisely, susceptible varieties profit more from fungicides than resistant cultivars but even the most resistant cultivars show a positive response. Modern fungicides (without strobilurins which were not used) had a very positive effect on yield. Reasons might be protective effects of Scientific Reports | (2020) 10:20374 | https://doi.org/10.1038/s41598-020-77200-0 www.nature.com/scientificreports/ fungicides against not yet visible damage by fungal diseases (e.g. by hemibiotrophic pathogens) and an increase in wheat stress tolerance (reduction of ethylene and oxidative stress) which were described by Wu & Tiedemann 77 and Zhang et al. 78 . In our study we used an approach in which all cultivars were analyzed simultaneously in the same field trials; rather than the analysis of historical data from continued official variety trials (VCU trials). Genetic gains in trials of the former design tend to be overestimated in the variants without fungicide applications 79 , attributed to "variety ageing" which results from gradual resistance breakdown [79][80][81] . Laidig et al. 54 calculated the genetic gain of winter wheat in German official variety trials for the period from 1983 to 2012 at + 1.15% a −1 in intensity 1 (without fungicide) and + 0.66% a −1 for intensity 2 (with fungicide). The difference between the two intensities is interpreted by a non-genetic, agronomic trend plus an effect called 'variety ageing' of 0.36% a −1 . Earlier, in a similar study including cultivars released from 1966 to 2007 Ahlemeyer and Friedt 51 found a similar trend albeit lower values of 0.63% a −1 (intensity 1) and 0.46% a −1 (intensity 2), respectively. The respective trends observed in our trials range between a higher relative grain yield gain of + 1.32% a −1 in variant T3 (comparable with intensity 1 in VCU trials) and a similar 0.54% a −1 in T4 (comparable with intensity 2). ' Ageing' as calculated by Laidig et al. 54 assumes the predominant release of mostly resistant varieties which is unlikely as Ahlemeyer and Friedt 51 have shown. Thus, we have to consider 'ageing' effects by gradual breakdown of resistances and durable resistance gains both taken together explaining the difference between treatments with and without fungicides. Ahlemeyer and Friedt 51 used susceptibility data from historical field trials which reflect durable resistance gains whereas resistance gains in our trial represent both durable resistance gain plus the loss of resistance by breakdown of R genes (the latter equaling the ageing effect). Thus, the relatively strong difference in grain yield gain between T3 and T4 (and partly also between T2 and T1) may result from the combined effect of these two components of resistance improvement. Drought stress and severe biotic stress through fungal pathogen pressure by artificial inoculation may have contributed additionally to yield differences between treatments. We conclude that the genetic trend in our trial might be somewhat overestimated due to stress, comparable to a potential overestimation of the genetic trend at low input conditions as observed by previous authors 79,82,83 .
Genetic gains and 'variety ageing' (breakdown of resistance) can be estimated if both variants with and without fungicide treatment are carried out in one trial 81 . The effect of fungicide application then directly affects the outcome of comparative field trials 54 . If chemical plant protection is very strict and successful, genetic gains can be estimated reasonably. That was likely the case regarding YR, LR, and FHB in our trial. If by comparison the fungicide control is not fully effective, older cultivars are burdened stronger by diseases and genetic gains may be biased and over-estimated. Accordingly, average annual breakdown of resistance genes would be underrated which could to some extent be the case for PM in our experiments.
In summary, it can be stated that different fungal pathogens attack wheat at different time points causing varying degrees of crop damage and yield loss. Results allowed quantifying the impact of N fertilization on the infestation by the fungal pathogens and their effects on grain yield as well as yield components. Resistance against fungal pathogens has been improved significantly but at varying speed during the last decades. The slope of improvement of genetic resistance was higher against biotrophic pathogens than to the hemibiotrophic pathogen. Our analyses support the assumption of the accumulation of longer-lasting quantitative resistance loci along with the deployment of race-specific R genes against biotrophic pathogens. Improvement in resistance over time was steeper at the higher N level in absolute but very similar in relative terms. The results strongly indicate that a substantial increase of genetic yield potential has been achieved in addition to simultaneous efforts to improve resistance against different fungal pathogens. As a matter of fact, the trend to higher resistance against the different wheat pathogens does not show signs of slowing down. Therefore, current German elite winter wheat cultivars provide an excellent basis for further yield gain by cross-breeding ("combination breeding"). That will enhance a more environment-friendly, sustainable crop production in the future leading to wheat cultivars which are better suited for farmers' needs and consumers' preferences. Drought periods can reduce the effectiveness of N application. As weather extremes are predicted to increase due to climate change and drought periods might become a greater problem particularly in semiarid areas agronomic strategies including N application and seed rates surely need to be adapted accordingly.

Material and methods
Plant material. A total of 178 German winter wheat cultivars, released in Germany between 1965 and 2013 (Bundessortenamt, BSA) were selected for the study based on their superior economic and agronomic importance in wheat production. The panel therefore represents long-term wheat breeding activities in Germany. General cultivar information including breeders and year of official release are provided in Supplementary Table S1. The panel of cultivars subsumes the German cultivars investigated by Voss-Fels et al. 58 .
Experimental site and trial design. The  www.nature.com/scientificreports/ Field trials were conducted in yield plots (area 4.5 m 2 ) in three consecutive growing seasons from 2014-2015 to 2017-2018. In each of the three year × location environments all 178 cultivars were sown side by side in a full factorial combination of high or low nitrogen levels combined with the presence or absence of fungicides, respectively, each in two replicates. To ensure comparable pathogen pressure, artificial inoculation has been carried out in each year. In total, 4272 plots were analyzed. The wheat cultivars were grown in a randomized incomplete block design with four treatments (T1: 110 kg N ha −1 , no fungicide; T2: 110 kg N ha −1 + fungicide; T3: 220 kg N ha −1 , no fungicide; T4: 220 kg N ha −1 + fungicide) in two replications, each. All trials were sown at a density of 330 seeds m −2 and received an herbicide (Fenikan ® , Bayer AG, Germany) treatment shortly after emergence. All variants were fertilized by Calcium ammonium nitrate (CAN, adjusted for soil N min ) up to three times from March to May at growth stages BBCH23-25, BBCH24-31, BBCH33-45 (BBCH scale 84 ) until the intended target amount was reached. N application in 2015 and 2017 was time-delayed due to drought periods and, by extension, likely only partly effective. Plant height was controlled in all treatments by two applications of growth regulator (CCC 720 ® , Bayer AG, Germany; Medax Top ® , BASF AG, Germany) in April and May. Fungal pathogens have been successfully controlled (infestation ≤ 1%) in treatments 2 and 4 by three annual fungicide applications (Capalo ® , Adexar ® , both BASF AG, Germany; Prosaro ® , Bayer AG, Germany) between April and June. Fungicides, insecticides (Karate ® Zeon, Syngenta AG, Switzerland) and growth regulators have been applied due to standard recommendations in Germany. Specific management information of each year and trial including date and amount of application is given in Supplementary Table S2.
Fungicides (plant protection, PP) effectively prevented the development of fungal diseases within the noninoculated treatments (T2, T4). Symptoms were not scored plotwise but monitored to apply fungicides timely. Accordingly, FHB was never observed in the fungicide treatments. The diseased leaf area did not exceed 1% for YR and LR in all three years, and for PM in 2015 and 2016. Only PM occurred up to 4% on single susceptible cultivars in T2 and T4 during summer 2017.
Treatments T1 and T3 were artificially inoculated with mixtures of PS, and PT races (see above) once, and with FC two times, each at periods of favorable growing conditions of the specific pathogen (PS: mid-April, BBCH23-30 at 5 to 10 °C; PT: mid-May, BBCH33-40 at 15 to 25 °C; FC: flowering, BBCH63-69; each at humidity > 80%). For the rusts, all respective plots were inoculated with an oil-spore mixture using a hand-held spinning disc sprayer (ULVA +, Micron Sprayers, Herefordshire, UK). Suspensions with a ratio of 0.2 g rust spores per liter oil (Isopar M, ExxonMobil Chemical Company, Spring, TX, USA) were prepared, sufficient for 400 m 2 plot area per liter. FC was inoculated by an aqueous suspension adjusted to about 300,000 conidia ml −1 (25 m 2 plot area per liter). High abundance of BGT led to spontaneous natural infection.
Disease phenotyping and morphological and agronomical traits. Susceptibility of the cultivars to each disease was scored regularly four times (FHB two times) during each growing season starting when the first symptoms were visible. Percentage of leaf area infected was estimated as described by Moll et al. 88 . FHB infection percentage was quantified as a combination of disease severity and disease incidence, referred to as FHB index. The area under the disease progress curve (AUDPC) was calculated for each cultivar and disease. AUDPC values were then used to compute the average ordinate of infestation (AO 89 ) for each of the diseases (YR, LR, PM, FHB), where (N) is the number of observations, the disease level at the ith observation is coded by (y i ), time at the ith observation by (t i ), and the total monitoring period in days is coded by (t p ). Fungal sum was calculated by unweighted addition of AO values of all four diseases.
The number of ears per square meter (ESM) was calculated based on counting the ears in one running meter of one of the inner rows in each plot. A subsample consisting of all plants of a row length of 0.5 m from inside the plot was cut prior to harvest to determine the harvest index (HI). Each sample was oven-dried starting at 65° increasing to 105 °C overnight. HI was calculated plotwise as the ratio (%) of the grain to the aboveground biomass. Number of kernels per spike (KPS) was computed based on grain yield divided by the thousand kernel weight and number of spikes. Total grain yield (GRY) was measured by harvesting entire plots of each cultivar and replicate. Grain moisture was determined and grain yield was corrected to a standard moisture of 14%. About 500 random seeds were selected from all samples to determine thousand kernel weight (TKW) using a MARVIN Seed Analyzer (GTA Sensorik GmbH, Neubrandenburg, Germany). Aboveground biomass (BM) of plots was calculated based on GRY and HI. Statistical analysis. Descriptive and summary statistics, contrast statistic as well as analysis of variance (ANOVA) for each parameter were calculated using JMP 14.0.0 90 . The LSmeans contrast function as implemented in JMP 14.0.0 has been used to test orthogonal contrasts between treatments by means of F statistics in order to examine the effects of the fixed two N rates and two plant protection levels. ANOVA was computed applying a factorial design model including four treatments (two levels of N input times two levels of plant pro-

Scientific Reports
| (2020) 10:20374 | https://doi.org/10.1038/s41598-020-77200-0 www.nature.com/scientificreports/ tection) within three years, and 178 cultivars using the PROC MIXED procedure of SAS 9.4 91 as implemented in JMP 14.0.0. N input and plant protection levels were considered as fixed factors, cultivars (genotypes) and years as random. Treatment-specific adjusted means (LSmeans) of each cultivar were calculated prior to analyses of relationships among traits and between traits and year of cultivar release. PROC RANK via JMP was then used to build quantile groups (classes) of similar YR infestation (YR 1 < 4.0455%, YR 2 4.0455 to 6.1509%, YR 3 6.1509 to 9.1852%, YR 4 9.1852 to 13.242%, YR 5 13.242 to 31%). Pearson's correlation coefficients were calculated using the PROC CORR procedure of JMP 14.0.0/SAS 9.4 90,91 . Factorial analysis was carried out by means of principal component analysis (PCA) using the PROC FACTOR algorithm. Data were analyzed by means of bivariate analysis also using SAS 9.4 procedure in JMP 14.0.0 90,91 to explore the relationships between GRY and the fungal diseases based on treatment-specific adjusted means as input values. To quantify the breeding progress year of cultivar release, considered as continuous variable, and treatment-specific adjusted means have been used. Partial bivariate analysis with YR classes (quantile groups) was used to dissect the relation of both GRY to LR and LR to year of cultivar release due to the LR dependency of YR. Separate analyses of each class was conducted.

Data availability statement
The