Assessing genetic gain and diversity in Ethiopian bread wheat across six decades

Abstract

Assessing breeding progress and genetic diversity in released varieties is crucial for informing future crop improvement strategies. This study evaluated genetic gain and diversity in 49 Ethiopian wheat varieties released over six decades (1967–2021). Significant yield increases were observed, with ‘Shaki’ showing the highest gain. However, the average annual genetic gain was modest (0.90% under irrigated, 0.69% under rain-fed). The recent released bread wheat variety ″Shaki″ exhibited a 48.6% increase in grain yield under irrigated conditions and a 37.1% increase under rain-fed conditions over the oldest variety, “Lakech”. Genetic diversity analysis revealed a decline in heterozygosity and a skewed allele distribution, suggesting potential inbreeding within the breeding program. Population structure analysis indicated a shared genetic background across decades. The low genetic gain and declining diversity highlight the need for strategies to enhance genetic diversity and incorporate novel yield-associated traits to achieve sustainable genetic improvement in Ethiopian wheat breeding.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-20383-1.

Introduction

Wheat is a key food crop with a significant impact on food security. It ranks as the second most produced grain globally, following maize (FAOSTAT, 2022)¹ Despite being the third most important cereal crop in Ethiopia, with a production of 5.6 million metric tons in 2023/2024, the country’s average wheat yield remains significantly lower (3.0 t ha⁻¹ at rain-fed and 4.0 at irrigation condition) than other wheat-producing nations^2,4. This gap highlights the need for enhanced productivity to meet the projected 50–60% increase in food production required by 2050³.

Ethiopia’s wheat breeding programs have demonstrated notable success in increasing yields, with 88 varieties released since 1970, achieving an average annual yield increase of 40 kg ha⁻¹⁴. This progress has resulted in a significant increase in yield from 2 t ha⁻¹ to 6.5 t ha⁻¹ (at experimental site)⁴. This impressive growth highlights the success of breeding programs in enhancing wheat productivity and underscores the ongoing efforts to develop high-yielding varieties that meet the increasing demand for food. However, further genetic progress is hindered by the lack of varieties adapted to diverse and changing agro-climatic conditions. To address this challenge, breeding programs need to incorporate a wider range of environmental conditions during variety development. Understanding the historical genetic progress in major wheat-producing nations is crucial for realigning future breeding programs to achieve the desired genetic gain of 2.4%, which is necessary to meet the anticipated global food demand by 2050⁵.

Wheat production has witnessed significant yield gains globally, with many countries, including China, European Union, Russia and India, experiencing notable improvements, which is particularly pronounced due to advances in agricultural practices, such as the adoption of high-yielding wheat varieties and improved irrigation methods. Over the past several decades, annual yield increases have ranged from 23 to 44 t ha⁻¹, contributing to enhanced food security in the region^6–8. In Ethiopia, average annual genetic gain in grain yield has been observed at 23.04-40 t ha⁻¹ from 1974 to 2021^4,9. While historical data showcases substantial progress, understanding the genetic basis of these gains is paramount for future breeding strategies and ensuring continued improvement.

Wheat breeding programs rely on genetic diversity within germplasm to identify and select superior varieties for specific environments and target traits^10,11. However, a narrow genetic background can lead to reduced resistance to biotic and abiotic stresses, and low adoption of new technologies remains a challenge in wheat production^12,13. Genetic diversity analysis provides valuable insights for identifying alleles associated with desirable traits like high yield and stress resilience. Moreover, while numerous studies have assessed genetic gain using morphological traits over decades, few have effectively integrated SNP marker data with this historical phenotypic information. This gap restricts our understanding of the genetic basis of observed yield improvements and how the genetic architecture of yield has changed over time. Therefore, the objectives of this study were to (i) evaluate the genetic gain progress of grain yield and associated traits under irrigated and rain-fed conditions and (ii) assess the genetic diversity and population structure of bread wheat varieties released in Ethiopia over the six-decades (1967–2021).

Result

Analyses of variance

A combined ANOVA analysis under rain-fed conditions revealed significant differences among varieties for all traits except biological yield (Table 1). Under irrigated conditions, similar significant differences were observed among varieties for all traits except biological yield. Interactions between varieties and locations were significant only for thousand kernel weight, grain weight per spike, and number of grains per spike under rainfed conditions. Under irrigated conditions, plant height, number of fertile spikelets per spike, and grain weight per spike showed significant interactions. The interaction between growing conditions (irrigated and rainfed) demonstrated significant differences among varieties for all studied traits (Supplementary Table S1).

Table 1.

Combined analysis of variances of 49 bread wheat varieties released from 1967 to 2021 tested at four locations under rain-fed and irrigation conditions in 2021/2022.

Traits	Under rainfed condition
	Block	Var	Loc	Var: Loc	Resid.	Min	Max	Mean	CV
GY	1.8*	6.3**	0.5**	2.8ns	2.0	2.8 ± 0.2	2.4	3.8	7.41
TKW	206.2**	2924***	188.9***	1609.3**	147.7	46.73 ± 4.4	39	63	9.51
HI	1154**	3840***	3729***	1514ns	926	30.7 ± 5.1	23.2	44.1	16.5
GWSP	2.6ns	32.7***	3.27***	13.2**	4.2	2.85 ± 0.5	1.97	3.97	16.5
PH	1738ns	15,079*	2864***	9939ns	5668	83 ± 10.1	62.9	107.2	12.2
TNT	191.7**	331.5*	863.4***	3.9ns	126.3	8.00 ± 1.5	4.7	12.5	18.8
DH	351.7ns	1515.9**	1496.7***	511.9ns	467.4	74.4 ± 3.2	68	83	4.3
DPM	488***	1904***	4119***	434ns	293	114.8 ± 3.6	107.7	122.3	3.17
NSSP	3.76***	180.5***	2.7 ns	58.68ns	27.1	18.2 ± 1.3	15	21.1	8.61
NGSP	271 ns	6046***	7.0 ns	3314***	1122	65.4 ± 6.3	52.1	85	9.72
	Under irrigated Condition
GY	0.1ns	30.3***	0.45ns	3.8ns	4.3	3.47 ± 0.4	2.73	4.40	11.1
TKW	4586ns	64,804*	259ns	31576ns	108,186	44.1 ± 5.4	31.5	55.50	12.3
HI	482ns	4526**	1517***	1458ns	2733	41.6 ± 5.9	31.5	57.16	14.3
GWSP	4.2*	22.5**	16.7**	17.0 *	14.5	3.35 ± 0.5	2.36	4.51	14.7
PH	3125***	10,065**	5479***	8824**	4960	74.9 ± 9.7	63.3	91.2	13.0
TNT	181.9 ***	178ns	546.5***	213.6ns	203.4	6.7 ± 0.81	5	8.2	112
DH	114.7ns	1270.5**	7.4ns	548.3ns	702.3	69.9 ± 3.3	65	81	4.7
DPM	164.8ns	1388.8*	29.7ns	775.8ns	1004	106.5 ± 4	97	116	3.7
NSSP	54***	145.9**	83.1***	117.2***	61.9	17.8 ± 1.2	20.3	14.7	6.4
BY	8.3ns	68.3ns	0.1ns	73.7ns	143.4	8.57 ± 0.7	6.84	10.1	7.9
NGSP	0.2**	0.7*	0.004ns	0.4ns	0.73	55.5 ± 7.7	35.6	4.5	13.8

*** = p < 0.001, ** = p < 0.01, * = p < 0.05, ns = non-significant; GY = Grain yield, TKW = Thousand kernel weight, HI = Harvest index,.

GWSP = Grain weight per spike, PH = Plant height, TNT = Total tiller number, DH = Days to heading, DPM = Days to physiological maturity, NSSP = Number of fertile spikelets per spike, BY = Biological yield, NGSP = Grain number per spike; CV = Coefficient of variation, Max = Maximum, Min = Minimum.

Performance of grain yield, and yield related traits under rain-fed and irrigated conditions

The overall mean of grain yield under irrigation and rain-fed conditions was 3.47 t ha⁻¹ and 2.1 t ha⁻¹, respectively (Supplementary Table S2 and Figure S1). Grain yields ranged from 3.0 to 4.4 t ha⁻¹ under irrigation, while yields ranged from 2.47 to 3.48 t ha⁻¹ under rain-fed conditions. The recently released variety ″Shaki″ exhibited the highest grain yield under both rain-fed and irrigated conditions followed by ″Boru″ (3.25 t ha⁻¹) and ″Abay″ (4.21 t ha⁻¹) under rain-fed and irrigated conditions, respectively. The results of this study revealed mean values of 83 days (under rain-fed) and 81 days (under irrigation) for days to heading, and 122 days (under rain-fed) and 116 days (under irrigation) for days to physiological maturity under the rain-fed and irrigation environments (Table 1).

Evaluation of genetic gain in bread wheat varieties released in Ethiopia

The regression analysis of the mean grain yield over the year of release explained 25.69% and 12.69% of the variation under irrigation and rain-fed conditions, respectively (Fig. 1A & B). The regression analysis revealed an average yield increase of 14.91 kg ha⁻¹ per year under irrigation (Fig. 1A) and 5.42 kg ha⁻¹ per year under rain-fed conditions (Fig. 1B). The Kernels weigh per year inclement also estimated 0.13 g and 0.1 g under irrigated and rain-fed condition, respectively (Fig. 1C & D). Similarly, the harvest index also high under irrigated condition than rain-fed condition (Fig. 1E & F). The relative annual genetic gain was 0.59 and 0.95% per year when computed over the mean grain yield of the oldest varieties under rain-fed and irrigated conditions in the 54 years of the breeding program (Table 2). The variety ″Shaki″ exhibited a grain yield advantage of 1.44 t ha⁻¹ (48.58%) under irrigated and 0.92 t/ha (37%) under rain-fed conditions over the oldest variety ″Lakech″. Genetic improvements for thousand kernel weight and harvest index were 0.13 and 0.14 g per year under irrigated and 0.086 and 0.14 kg/ha/year under rain-fed conditions, respectively.

Fig. 1.

Genetic gain for grain yield and yield related traits in bread wheat released varieties from 1967–2021 A) grain yield under irrigation, B) grain yield under rain-fed, C) thousand grain weight rain-fed, D) thousand grain weight irrigation, E), harvest index under irrigation, and F) harvest index under rain-fed conditions.

Table 2.

Genetic progress of the Ethiopian bread wheat varieties for grain yield over six decades (1967 to 2021) under irrigated and rain-fed conditionins in a study conducted during in 2021/2022.

Rain-fed condition
Decades	Aver. of decade	Over genetic gain		Shaki vs. Lakech
		Inc. over 1 st decade t ha −1	ARGG%	Inc. t ha-1	%	RAGG%
1967–1976	2.9
1977–1986	3.34	0.45	15.7
1987–1996	3.16	0.28	8.27
1997–2006	3.32	0.43	13.6
2007–2016	3.37	0.48	14.6
2017–2021	3.48	0.59	20.55	0.92	37.1	0.69
Irrigated condition
1967–1976	3.11
1977–1986	3.28	0.17	5.50
1987–1996	3.34	0.23	7.12
1997–2006	3.28	0.18	5.28
2007–2016	3.48	0.38	11.46
2017–2021	4.053	0.95	30.30	0.144	48.58	0.90

ARGG%=Annual Relative Genetic Gain in the decade, RAGG% = the Relative Advantage in Grain Yield over 1 st decade (an increase in decade over the first decade).

Decade-wise analysis showed a relative advantage in grain yield (RAGG) of 5.5% (0.17 t ha⁻¹), 7.12% (0.23 t ha⁻¹), 5.3% (0.18 t ha⁻¹), 11.5% (0.38 t ha⁻¹), and 27.2% (0.95 t ha⁻¹) in the 2nd, 3rd, 4th, 5th, and 6th decades, respectively (Table 2), over the first decade under irrigated conditions. However, under rain-fed conditions, a RAGG of 15.7% (0.45 t/ha), 8.3% (0.28 t ha⁻¹), 13.6% (0.43 t ha⁻¹), 14.62% (0.48 t ha⁻¹), and 17.6% (0.59 t ha⁻¹) was achieved in the 2nd, 3rd, 4th, 5th, and 6th decades, respectively, over the 1 st decade (Table 2). Under rain-fed conditions, traits such as thousand kernels weight, harvest index, grain number per spike, spike weight, and grain weight per spike exhibited 11.95%, 29.27%, 20.88%, 5.80%, and 10.6% increases in the 6th decade over the first decade (1967–1974). Similarly, these traits showed an increase of 8.03%, 35.8%, 14%, 22.7%, and 6.12%, respectively, in the 6th decade over the first decade under irrigated conditions. The remain traits detailed in Supplementary Table S3 and Figure S2.

Genetic diversity and variation in bread wheat varieties released in Ethiopia over six decades

Of the 49 varieties used in the study, 41 contained genetic information, while the remaining eight did not. Consequently, the genetic analysis was based on the 41 informative varieties. The average frequency of heterozygous alleles (Ho = 0.23), high mean PIC value (0.56), moderate effective number of alleles (Nes, 2.8), and very low level of nucleotide diversity (π, 8.409688e-06) genetic variation across the varieties were observed. Across the decades’, observed heterozygosity is lower than expected (Table 3). Observed heterozygosity ranged from 0.047 to 0.106, with an average of 0.062. The expected heterozygosity also varied from 0.233 to 0.288 with average of 0.262 across the six decades. nucleotide diversity across the decays varied from 0.462 to 0.544 (averaging 0.500). The inbreeding coefficient (Fis) was negative ranged from − 0.125 to −0.109. Approximately similar values of polymorphic information content were found among each decade (mean value of 0.33). The values for major allele frequency (MaF) and minor allele frequency (MAF) ranged from 0.78 – 0.81 and 0.19 – 0.22 was exhibited across decades. The private number 20alleles across all six decades varied from 36 (3rd decade) to 2049 (5th decade). Private alleles varied from 30 (3rd decade) to 857 (5th decade) when cooperate to neighbour decades.

Table 3.

Summary of genetic indices among bread wheat released varieties within the six decades.

Decay	N	Ho	He	Fis	PIC	MAF	MaF	Pr. alle	PrP	N. div
1st	3	0.047	0.278	−0.116	0.34	0.22	0.78	218	130	0.495
2nd	5	0.056	0.280	−0.117	0.34	0.22	0.78	252	139	0.544
3rd	2	0.106	0.288	−0.112	0.33	0.21	0.79	36	30	0.485
4th	10	0.063	0.240	−0.109	0.31	0.19	0.81	418	226	0.492
5th	15	0.058	0.250	−0.118	0.32	0.20	0.80	2049	857	0.519
6th	6	0.055	0.233	−0.125	0.31	0.19	0.81	1041	545	0.462
Total/Average	41	0.064	0.262	−0.116	0.325	0.201	0.795	669.0	321.17	0.500

N = Number of varieties; Ho = observed heterozygosity; He = expected heterozygosity; Fis = inbreeding coefficient, PIC = polymorphic information content, MaF = major alleles frequency and MAF = minor allele frequency, Pr.alle.= privet alleles over population, PrP = private alleles compared to neighbour decades and n.div = nucleotide diversity.

Population structure and cluster analysis

A population structure analysis of the 41 bread wheat varieties for K ranging from 2 to 10 identified k = 4 as the optimum number of groups based on minimizing the cross-entropy of the data (Supplementary Figure S3). Similarly, discriminant analysis of principal components (DAPC) also clearly classified the varieties into four groups (Fig. 2A). However, the DAPC was not clearly differentiated the varieties based on the released decades (Fig. 2B). The LEA results largely supported the population structure analysis with 71% of the varieties assigned to one of the five groups with a higher than 0.60 ancestry membership coefficient (Fig. 3A). However, the graph depicting population structure suggested significant genetic admixture or a lack of distinct genetic differentiation within the release decades. The population structure across the released decades were highly admixture which none of the released decade continued unique group when grouping varies from k = 2–6 (Supplementary Figure S4). This pattern suggests that genetic traits used in the Ethiopian bread wheat breeding program to increase genetic gain across the six different periods (decades) were consistent (Fig. 3A). Similarly, the cluster analysis categorized the bread wheat released varieties into four distinct groups. Notably, the varieties within each cluster were not limited to those released within the same decade, emphasizing the genetic similarity observed in the clusters, regardless of the varieties’ release decades (Fig. 3B).

Fig. 2.

Discriminant analysis of principal components (DAPC) of 41 bread wheat varieties released in Ethiopia through 1967–2021 years (A) based on the genetic clustering & (B) based on the released decades: 1 st decade = 1967–1976, 2nd decade = 1977–1986, 3rd decade = 1987–1996, 4th decade = 1997–2006, 5th decade = 2007–2016 and 6th decade = 2017–2021.

Fig. 3.

Population structure among released bread wheat varieties within six decades in Ethiopia bread wheat (1967–2021), (A) based on optimum genetic group(K = 4) and (B) Dendrogram illustrating.

Differentiation within and among decades

The low fixation index values indicate that alleles are frequently exchanged between wheat genotypes released in different decades, reflecting genetic similarity. The pairwise F < sub > ST</sub > values, which range from 0.04 to 0.173, demonstrate moderate levels of genetic differentiation. The highest genetic differentiation was observed between the 1 st and 3rd decades (Fts = 0.173), while the lowest was found between the 4th and 5th decades (Fst = 0.040), as illustrated in Fig. 3. Edge thickness and color intensity in the network diagram represent the magnitude of genetic differentiation between decades (Fig. 4). Similarly, analysis of molecular variance over the six decades showed 1.6% variance (low Phi (Φ) value of 1.64%), suggesting minimal genetic progress or changes achieved in the allelic frequencies of the bread wheat varieties released over the six decades in Ethiopia (1976–2021). The majority of the genetic variances (98.36%) were within populations or decades, reflecting genetic variation among individuals within the same decades (Supplementary Table S4).

Fig. 4.

Network diagram of pairwise genetic differentiation among wheat genotypes across decades. Edge thickness and color intensity in the network diagram represent the magnitude of genetic differentiation between decades.

Discussion

This study assessed the genetic progress among the 49 historic Ethiopian bread wheat varieties released between 1967 and 2021 under irrigated and rain-fed growth conditions. The study consistently found significant differences among varieties for most traits, highlighting the potential for genetic improvement in bread wheat yield and quality. The significant interactions between varieties and locations for certain traits suggest that the performance of some varieties may be influenced by specific environmental conditions. Further investigation is needed to understand these interactions and identify varieties that are more suited to particular regions. This emphasize is importance to considering specific environmental conditions when selecting and recommending cultivars. These results align with the findings of Gemechu et al.¹⁴ reported significant variations among the varieties across all studied traits in bread wheat.

Grain yields were significantly higher under irrigated conditions compared to rain-fed conditions, emphasizing the importance of irrigation for improving wheat production in Ethiopia. The mean grain yield under irrigated conditions was 3.5 t ha⁻¹, while a mean grain yield of 2.8 t ha⁻¹ was obtained under rainfed conditions (Table 3). The interaction between growing conditions (irrigated and rain-fed) and varieties highlights the importance of considering both genetic and environmental factors in breeding programs. This suggests that breeding for adaptability to different growing conditions is crucial for sustainable wheat production.

The recently released variety “Shaki” demonstrated the highest grain yield under both rain-fed and irrigated conditions, indicating its adaptability to different environments. This variety’s consistently high yield performance indicates its adaptability to diverse agro-ecological conditions and its potential as a promising cultivar for Ethiopian farmers to enhance bread wheat production. “Boru” and “Abay” also performed well, suggesting promising genetic potential. The variety “Shaki” achieved the highest mean yield (4.4 t ha⁻¹) under irrigation, contrasting sharply with the lowest yield from the oldest variety, “Lakech” (2.26 t ha⁻¹). This represents a substantial yield gap, with “Shaki” outperforming “Lakech” by 48.58% and 37.05% under irrigated and rain-fed conditions, respectively. The substantial yield gap underscores the genetic progress achieved in wheat breeding over the years.

Similar to grain yield, other traits, thousand kernel weight and harvest index exhibited higher means under irrigated than rain-fed conditions. The observed differences in thousand kernel weight and harvest index between irrigated and rain-fed conditions highlight the significant influence of environmental factors on these key yield components. The consistently higher values under irrigation suggest the important role of irrigation in increasing the size and weight of wheat kernels, which contributes to the increased economic yield¹⁵. These findings underscore the potential contributions of irrigation management and breeding for yield related traits that are more responsive to favourable water conditions in improving grain quality. Moreover, the mean values for biological yield remained relatively consistent across varieties, suggesting that breeding efforts have not significantly altered the maturity cycle of Ethiopian bread wheat.

The findings of this study provide valuable insights into the genetic progress of Ethiopian bread wheat varieties over the past 54 years. Estimating genetic progress is vital for informing breeding strategies and enhancing productivity¹⁶. The regression analysis revealed a steady increase in grain yield over time, indicating significant genetic progress in Ethiopian bread wheat breeding programs. The decade-wise analysis highlights the accelerated genetic progress in later decades, particularly in the 5th and 6th decades, suggesting that breeding efforts have become more efficient over time. The genetic gain for grain yield explaining 25% and 16% of the variation under both irrigated and rain-fed conditions, respectively. The estimated genetic gain for grain yield was 5.95 kg/ha/year under rain-fed and 13.5 kg/ha/year under irrigated conditions, representing 37.1% and 48.6% improvements over the oldest variety, “Lakech”. These figures equate to annual genetic gains of 0.69% and 0.90% under rain-fed and irrigated conditions, respectively. This result revealed that grain yield gains were higher under irrigated conditions compared to rain-fed conditions, emphasizing the importance of irrigation for improving wheat production in the region.

These results surpass those reported by Bassi and Nachit¹⁷ under rain-fed conditions, they fall short of the gains achieved in Pakistan⁷ and other regions of Ethiopia^4,18. Similarly, the irrigated grain yield gain was lower than that reported for India⁸. Similarly, Kokhmetova et al.¹⁹ and Yadav et al.⁸ reported higher relative annual gains of 0.56% and 0.88%, respectively. Additionally, regional comparisons with Morocco, Sudan, and Egypt shows significantly higher progresses of 2.5%, 1.3%, and 2.3% under irrigation⁴. A domestic study by Girma et al.⁹ also indicated a higher relative annual genetic gain of 1.03% under rain-fed conditions. The rate of genetic gain is notably lower compared to those reported in other studies, highlighting the need for enhanced breeding strategies. The slower rate of genetic gain in the Ethiopian bread wheat varieties could be attributed to various factors, including limited genetic diversity in the breeding pool, the technical and research facility capacity of the breeding programs in the country, and adverse environmental conditions¹⁸. Moreover, the less pronounced trend under rain-fed conditions suggests that breeding efforts might need to focus more on improving yield stability and adaptation to drought stress. The positive trends in thousand kernel weight, harvest index, grain number, and weight per spike, particularly under irrigation, indicate that breeding programs have been successful in improving these yield components. These results suggest that continued selection for these traits, especially under irrigated conditions, could lead to further yield enhancement.

Genetic improvements in thousand kernel weight and harvest index were evident under both irrigated and rain-fed conditions, suggesting that breeding programs in Ethiopia have effectively prioritized these key yield components. Over the years, the selection criteria have gradually shifted to emphasize not only grain yield but also physiological traits like harvest index and Thousand kernels weight, which enhance resource use efficiency. The observed decrease in certain other traits could be attributed to this focus on harvest index and thousand kernels weight, potentially leading to reduced emphasis on traits such as total biomass production and overall plant growth. Additionally, changing agro-climatic conditions may have further constrained improvements in biological yield within the breeding program.

Genetic diversity and population structure analyses were performed on the 41 bread wheat varieties to assess and evaluate genetic parameter changes occurred on the released varieties over different decades of variety release. Regular assessment of genetic diversity is crucial in breeding high-performing wheat varieties adapted to diverse agro-ecological conditions and capable of withstanding evolving climates²⁰. Monitoring these genetic metrics Fis, MaF, and MAF, provides valuable insights into the dynamics of genetic progress and can guide breeding programs in designing strategies for effective genetic improvement in bread wheat.

The genetic analysis of these 41 Ethiopian bread wheat varieties revealed a moderate level of genetic diversity, as evidenced by the average heterozygosity (Ho = 0.23), polymorphic information content (PIC), and effective number of alleles (Nes). However, the nucleotide diversity (π) was relatively low, suggesting a limited genetic variation among the varieties (Table 3). These findings align closely with previous research on bread wheat in Turkey, where the average heterozygosity values were 0.201²¹. One probable reason for the moderate genetic diversity could be the historical emphasis on selecting for high-yielding traits, particularly focusing on Thousand Kernel Weight and Harvest Index. To address this limitation, we suggest considering a broader range of yield-related traits, including Thousand Kernel Weight and Harvest Index, alongside the utilization of modern breeding techniques such as genomic selection and marker-assisted selection. These approaches can help enhance genetic diversity while simultaneously improving breeding progress.

Over the six decades, observed heterozygosity was consistently lower than expected, indicating a potential loss of genetic diversity. This trend is further supported by the negative inbreeding coefficient (Fis), which suggests a degree of inbreeding within the breeding program. The major allele frequencies (MaF) remained relatively high, while minor allele frequencies (MAF) were correspondingly low. This suggests a skewed allele distribution, with a few alleles being more common than others. The number of private alleles (alleles unique to a particular decade) varied significantly across decades, with the 5th decade having the highest number of private alleles. This indicates that breeding programs have successfully introduced new genetic variation into the gene pool, especially in recent decades. However, this decade contained large number of varieties compared to other decades.

The observed (Ho) and expected (He) heterozygosity increased up to the third decade and then decreased until the sixth decade (Table 3). These findings indicate a potential decline in genetic diversity over time, which warrants further investigation to understand its implications for breeding programs. Lower heterozygosity indicates a reduction in the number of different alleles present in the population. This limits the genetic pool available for breeding and adaptation. A decrease in heterozygosity leads to an increase in homozygosity, which can expose deleterious recessive alleles and reduce plant vigour. This may also limit the potential of a germplasm for genetic gain: as a reduced genetic base hinders the development of new, superior varieties through breeding efforts.

The relatively stable levels of observed heterozygosity and inbreeding coefficient across decades, along with the modest genetic gains achieved, point toward a potential slowdown in breeding progress. This trend may be due to several factors, such as a narrow genetic base in elite germplasm, continued directional selection that limits diversity, and limited incorporation of new alleles from landraces or wild relatives. Addressing this challenge calls for more dynamic and forward-looking strategies. For instance, integrating genomic selection can help accelerate breeding cycles, while targeted introgression of diverse genetic material can broaden the genetic base. Similarly, using multi-trait selection indices can improve the efficiency of selection without compromising key traits. Importantly, any approach should also consider maintaining a healthy level of genetic diversity and managing inbreeding to ensure the long-term success and sustainability of wheat breeding programs²².

The consistently negative inbreeding coefficient (Fis) values observed across all datasets, ranging from − 0.109 to −0.125 (Table 3), indicate a notable trend in the genetic diversity of the studied bread wheat varieties. This uniform negative Fis trend suggests that the varieties are relatively well-mixed genetically across the six decades. This observation is further supported by the population structure analyses that demonstrated a shared genetic background among the varieties (Supplementary Figure S4). When the population structure run from K = 2–6, non-off the released decades clearly subset into one group. Similarly, the DAPC (Fig. 2A) and cluster analysis (Fig. 2B) classified the population into four distinct groups without regardless of the released decades. The results from the population structure analysis reaffirm the gene flow (genetic mixing) across the six decades. Significant allelic shifts were observed between varieties released in the first decade and those from the other five decades. Similarly in USA, a higher inbreeding coefficient value (inbreeding coefficient, −0.67) was reported in a study of 185 bread wheat cultivars released in the United States from 1943 to 2013, which also depicted the sharing of a common genetic background across breeding periods²³.

Similar to other genetic diversity parameter, the decreasing PIC (from 0.34 to 0.31) and MAF suggests that the breeding program is experiencing a decline in genetic diversity. This decline can hinder genetic progress and reduce the adaptability of varieties to environmental changes²⁴. This implies a decrease in the number of different alleles present in the population, which can limit the genetic progress available for breeding. This increase in homozygosity can expose deleterious recessive alleles, and hence decrease plant vigour and adaptability²⁵. Moreover, it limits the potential for genetic gain due to a reduced genetic base, which hinders the development of new, superior varieties in the future breeding efforts. Rare alleles often carry unique genetic information that can benefit the future generations. Their loss can limit the potential for genetic improvement. Reduced genetic diversity can also make a population more vulnerable to abiotic stress conditions. Contrasting with the considerably lower PIC value of 0.158 reported for Pakistani bread wheat²⁶. While other studies have reported comparable PIC values^27,28 reported moderately high PIC values in Australian and Belgian wheat. Moreover, the results underscore the importance of continuously monitoring these metrics to effectively guide breeding decisions and maintain genetic health within the breeding population. This proactive approach will help ensure the breeding program adapt to future challenges and sustain its effectiveness over time.

Analysis of molecular variance indicated that 98.4% of genetic variation in bread wheat populations was within rather than among decades. This has significant implications for the crop’s genetic structure and potential for future genetic improvement. This suggests that individual bread wheat varieties released within a decade are genetically quite different from each other. This high-level within-decade diversity also indicates the breeding program used a highly valuable germplasm that served as sources of a wider pool of alleles for selecting desirable traits. The small percentage of genetic variation among decades implies that there are relatively low genetic differentiations among the different bread wheat populations across the different decades. This might be due to factors such as gene flow (pollen and seed dispersal), historical bottlenecks, using similarly varieties for improvement or recent population expansions. A pairwise Fis range of 0.04 to 0.17 indicates moderate to relatively high levels of inbreeding among the released decades studied. Similar to genetic diversity parameters, this low inbreeding can lead to a reduction in genetic diversity. A population study on bread wheat under drought conditions conducted by Mdluli et al.²⁹ reported a mean Fis value of 0.4, with a maximum value of 0.96. Hussain et al.²⁶ reported a higher level of population differentiation (20%) in Pakistani bread wheat, potentially due to increased germplasm exchange and selection for specific traits³⁰. Consequently, it may be essential for breeders to incorporate new genetic material and reassess their strategies to ensure that the breeding program remains effective and resilient in the face of changing agro-climatic conditions.

In conclusion, the findings of this study emphasize the need for a balanced approach to breeding Ethiopian bread wheat. While maintaining genetic diversity is crucial, it is equally important to focus on developing varieties with desirable traits that can withstand evolving environmental conditions. By carefully managing genetic resources and adopting effective breeding strategies, Ethiopian bread wheat programs can ensure the continued production of high-yielding and resilient varieties.

Methods

Description of the study area

The field experiments were conducted at four locations in Jimma Zone, Southwest Ethiopia, during the 2021–2022 main cropping seasons (Supplementary Table S5). Two of these sites, Dedo and Gomma, were used for the rain-fed experiments, while Gomma and Gummay were used for the irrigated environment.

Study materials, experimental design and procedure

The study comprised of 49 bread wheat varieties, developed over a 54-years period (1967–2021) by the wheat breeding program in Ethiopia (Supplementary Table S6). These varieties were collected from Kulumsa agriculture research centre, Ethiopia. The experiment was laid out in a randomized complete block design with two replications under both irrigated and rain-fed conditions. The spacing between blocks was maintained at 80 cm, while the spacing between plots within a block was 60 cm, and between rows within a plot was 20 cm for all varieties. Blocks were spaced 1 m apart and Seeds were sown manually at a rate of 152 kg ha⁻¹. Recommended fertilizer (NPS; nitrogen, phosphorus, and sulfur in a ratio of 19% N, 38% P₂O₅, and 7% S) was applied uniformly at 100 kg ha⁻¹. All NPS and one-third of the urea were applied at planting, while the remaining two-thirds of the urea was applied at the tillering stage. Irrigation was applied every 4–5 days during the first three weeks, and then every ten days later in the crop cycle. Three rounds of hand weeding and fungicide applications were carried out at all experimental sites for weed and disease management.

Data collection

Phonological and agro-morphological data were collected using the wheat descriptors of the International Board for Plant Genetic Resources³¹. Days to 50% heading was recorded at the anthesis stage. Days to 75% physiological maturity, total number of tillers per plant and fertile tillers per spike were measured when 75% of the glumes on the main spike turn yellow, and the leaves are beginning to turn yellow. After harvest, number of fertile spikelets per spike (excluding the extreme spikelets at the top and bottom) and grain number per spike was counted. Thousand kernels weight (g) was measured from a randomly taken 1000 kernels after adjusting their moisture content to 12%. Biological yield (t ha⁻¹) was obtained from the harvested two middle rows of a plot and converted to tone per hector (t ha⁻¹. Grain yield (t h⁻¹) was recorded from the two middle rows of a plot, adjusted to 12% moisture content using smart moisture meter, and expressed in ton per hectare using the following formula;

where: Measured yield is the actual yield recorded at harvest moisture, measured moisture is the moisture content at harvest (%), and Standard Moisture is the target moisture content for comparison (12%).

Data analysis

To address spatial heterogeneity within each environment or trial arising from the large field size, a linear mixed-effects model was fitted using the lme4 package in R. The model included row and column positions as fixed effects to account for spatial trends across the field, while block was modeled as a random effect to capture local variation. This spatial adjustment improved the accuracy of genotype performance estimates by reducing the confounding influence of environmental gradients within each trial. The differences among varieties were assessed using analysis of variance for all the variables with agricolae v1.3.6 package³² in R version 4.3.1. Each recorded trait was analyzed independently for each environment. Combined analyses of variance for each of the irrigated versus rain-fed conditions were conducted following the guidelines outlined by Gomez and Gomez (1984) after testing the homogeneity of error variances using the F-test value of the MSE ratio of the locations that will be combined. The combined analysis model was:

Yijk refers to the observed value of variety i in block k of location j; m = grand mean, G = effect of genotype i, L_j = effect of location j, GL_ij = the interaction effect of genotype i with location j; RepL (_j) = effect of replication k in location j, e_ijk = random error (residual) effect of genotype i, location j and k replications.

A post-hoc Fisher’s (least significant difference, LSD) at 5% level of significance was performed using the LSD.test function in the agricolae package for mean separation³². To address spatial heterogeneity due to the large field size, a linear mixed-effects model was also applied using the lme4 package in R. This model incorporated plot row and column positions as fixed covariates and treated blocks as random effects, allowing adjustment for environmental gradients across the site.

To assess grain yield trends over time, the varieties were grouped into six decades based on their year of release: 1967–1974 (1st decade), 1975–1984 (2nd decade), 1985–1994 (3rd decade), 1995–2004 (4th decade), 2005–2014 (5th decade), and 2015–2021 (6th decade). To estimate the linear relationship between yield and its related traits, and to determine the unit improvement over the years, the following linear equation from Gomez and Gomez³³ was employed. Genetic gain and coefficient of determination were derived from separate regression analyses conducted for the rain-fed and irrigated conditions, using the following equation:

Where Y = the value of the dependent variable, X = the independent variable, β0 = the intercept of the regression line, β1 = the regression coefficient, or the changes in Y per unit change in X. To estimate the improvement in each decade, the average grain yield and related traits of varieties within each decade were computed relative to the average grain yield of varieties in the first decade (1967–1974). The relative annual gain over the 54-year period (1967–2021) was calculated by subtracting the mean grain yield of the oldest variety (1967) from the mean grain yield of the most recently released variety (2021). This difference was then divided by the mean grain yield of the oldest variety. Finally, RAGG% (relative annual genetic gain in %), the quotient obtained in the previous step, was divided by the total number of years in the breeding period and expressed as a percentage using Microsoft Office Excel (2010).

Genetic analysis

A total of 220 bread wheat accessions were genotyped using DArTseq^® sequencing technology by Diversity Arrays Technology Pty Ltd (DArT P/L), a sequencing-based genotyping service provider based in Australia (10.14264/54322). For genetic analysis in this study, 41 bread wheat varieties were extracted with 15,139 SNP markers contain less than 20% missing values and > 0.05 minor allele frequency. Genetic diversity parameters such as allele frequencies, minor allele frequency (MAF), heterozygosity (He), private alleles, nucleotide diversity and polymorphism information content (PIC) were determined using adegenet Package in R version³⁴. Analysis of Molecular Variance (AMOVA) and pairwise fixation index (Fst) were performed using the AMOVA and genet.dist functions, respectively from the pegas package in R³⁵. An unweighted pair group method with arithmetic mean dendrogram was created using cluster and factoextra R packages³⁶. The genetic distance was computed using the Euclidean method and the dendrogram was constructed using ward.D² method. Bayesian model-based software, STRUCTURE version 2.3.4³⁷ (URL: https://web.stanford.edu/group/pritchardlab/structure.html), was used to analyze population structure of the bread wheat varieties using a burn-in of 10,000, a run length of 100,000 using the LEA R package (version 3.10.0)³⁸ (https://github.com/bcm-uga/LEA) in R version 4.3.2 (https://www.r-project.org/).

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Sefawdin Berta: Conceptualization; data curation; formal analysis; methodology, writing original draft; writing review and editing. Temesgen M. Menamo: Conceptualization; formal analysis; methodology; supervision; writing original draft; writing—review and editing. Abush Tesfaye: Conceptualization; supervision; writing review and editing. Zerihun Tadesse: Conceptualization; resources; supervision; validation; writing review and editing.: Conceptualization; supervision; validation; writing original draft; writing review and editing. Techale Birhan: Conceptualization; supervision; validation; writing original draft; writing review and editing.

Data availability

The datasets supporting the findings of this study are publicly available at [https://doi.org/10.25502/n5v8-3784/d] Accession number n5v8-3784.

Declarations

Competing interests

The authors declare no competing interests.