A candidate gene from the family CaZF-CCHC in chickpea with a potential role in tolerance to abiotic stresses was identified in the NCBI database for SNP in C. arietinum L. (https://www.ncbi.nlm.nih.gov/snp). The gene family with full-length nucleotide sequences and corresponding polypeptide sequence was retrieved from the same database and used for both BLASTN and BLASTP in databases NCBI and GenomeNet, Kyoto University, Japan (https://www.genome.jp/tools/blast). After the identified CaZF-CCHC gene family was studied, based on differences in SNP databases, two genes, Ca04468 and Ca07571, with a specific combination of Zn-finger domains, were selected for further analysis as the most suitable candidates for plant response to drought and dehydration. All sequences of genes and encoded proteins in chickpea and other legume species were identified and downloaded from GenomeNet and NCBI databases. Chromosome locations, positions on the physical map, and gene identification for chickpea was found in the Legume Information System database (LIS) using BLAST, cv. Frontier v1.0 assembly (https://www.legumeinfo.org/taxa/cicer).

The molecular–phylogenetic dendrogram was constructed using CLUSTALW Multiple Sequence Alignment at GenomeNet Database Resources (https://www.genome.jp/tools/blast) with the ETE3 FastTree program. The result file was converted into a “nex” file for further use in SplitsTree4 version 4.14.4 from algorithms in the bioinformatics website at the University of Tübingen, Germany (https://uni-tuebingen.de).

Initially, six chickpea germplasm accessions were studied, including three Kabuli-type cultivars, Kamila, Krasnokutsky-123, and Looch, and three Desi-type accessions, ICC-1083, ICC-10945, and ICC-12654. Additionally, accession ICC-4958 was used as a reference genotype. Seeds of the Kabuli-type cultivars were obtained from the collection of S.Seifullin Kazakh AgroTechnical Research University, Astana, Kazakhstan, and included locally adapted genotypes originating from both Russia and Kazakhstan. The remaining chickpea accessions were obtained from the ICRISAT Reference set collection, India, and distributed for research purposes in Australia.

Selected chickpea accessions were crossed using manual emasculation and controlled pollination. The true hybrid F₁ plants were confirmed by flower color, with maternal and paternal parents bearing white and colored flowers, as well as PCR analysis of the marker KATU-C22 developed earlier (Khassanova et al., 2019). Each F₂ plant was used as the progenitor of breeding line families, and F₆ breeding lines were used in the analysis. For the purpose of this study, two hybrid populations were studied as follows: (1) ♀ Krasnokutsky-123 × ♂ ICC-12654 and (2) ♀ ICC-10945 and ♂ Looch. Cultivars Krasnokutsky-123 (with dark seeds) and Looch (with light seeds), Kabuli-type, originated from Russia and Kazakhstan, respectively. Chickpea ICRISAT accessions, ICC-12654 and ICC-10945, both with dark seeds, Desi-type, originated from Ethiopia and India, respectively. In hybrid populations 1 and 2, originally 12 F₆ breeding lines were developed in each population with subsequent selection of six and seven, respectively, for DArT analysis, and three and four for qPCR and field trial testing, respectively.

Seeds were sown in 6-L plastic pots filled with BioGro, Australia, potting media with the addition of “NoduleN” chickpea legume inoculant, New Edge Microbial, Australia. For each genotype, four plants per pot and two pots per treatment were grown in a greenhouse with 26°C/18°C day/night, with natural light and, on cloudy days, supplemented with LED with a photon flux density of 800 µmol m⁻²·s⁻¹. The relative humidity was controlled at 40%, and pots were watered with tap water twice per week. Placement of plants in the greenhouse was fully randomized.

For slowly developed drought, plants were grown for 1 month in well-watered conditions. Four time-points were designated for sampling starting from “day 0,” before the start of drought treatment, and 5, 7, and 9 days after watering was withdrawn. Leaf wilting symptoms in stress-treated plants were observed gradually. Watering was continued unchanged for control plants. Leaf samples were collected from three randomly selected plants (three biological replicates) for each time-point, genotype, and treatment, then frozen in liquid nitrogen and stored at −80°C for further analyses.

For rapid dehydration of detached leaves, a separate set of 1-month-old well-watered plants was used. For each genotype, leaf samples were split into four batches, designated as 0, 2, 4, and 6 h, with three biological replicates in each batch. The first group of samples (0 h) was frozen immediately in liquid nitrogen, whereas other leaf samples were exposed to dehydration on paper towel on the lab bench at room temperature (22°C). Consequently, after 2, 4, and 6 h of exposure to dehydration, leaf samples were frozen in liquid nitrogen and stored for further analyses.

DNA was extracted as described earlier (Weining and Langridge, 1991; Khassanova et al., 2019) with the following adjustments. Leaf samples were collected in 10-ml tubes from individual plants and frozen in liquid nitrogen. They were ground with two 8-mm stainless ball bearings using a Vortex mixer. One microliter of DNA was checked on a 1% agarose gel to assess quality, and concentration was measured by Nano-Drop (Thermo Fisher Scientific, USA).

To identify SNPs in the identified gene fragments, they were compared with reference genotypes of cv. Frontier and accession ICC-4958, then primers were designed in coding regions and promoters ( Supplementary Material S1 ) covering approximately 1 kb of amplification fragments. PCR was performed in 60-µl-volume reactions containing 6 µl of template DNA adjusted to 50 ng/ml and with the following components in their final concentrations as listed: 1× PCR buffer, 2 mM of MgCl₂, 0.2 mM each of dNTPs, 0.25 mM of each primer, and 4.0 U of GoTaq Flexi DNA polymerase (Promega, USA) in each reaction. PCR was conducted on a Thermal iCycler (Bio-Rad, USA) using a program with the following steps: initial denaturation, 94°C, 2 min; 35 cycles of 94°C for 15 s, 55°C for 15 s, 72°C for 1 min, and a final extension of 72°C for 3 min. Single bands of the expected size were confirmed after visualization of 5 µl of the PCR product in 1% agarose gel. The remaining PCR reaction volume (55 µl) was purified using a FavorPrep PCR Purification kit (Favorgene Biotec Corp., Taiwan) following the manufacturer’s protocol. The concentrations of purified PCR products were measured using NanoDrop (Thermo Fisher Scientific, USA) and later used as a template for Sanger sequencing at the Australian Genome Research Facility (AGRF), Adelaide, Australia. SNPs were visualized using the Chromas computer software program version 2.0 with manual comparison and SNP identification.

The ASQ method was used for plant genotyping following the protocol described recently (Amangeldiyeva et al., 2023). In brief, option A with a short 4-bp tag was used, and two allele-specific forward primers and one reverse primer were developed targeting two SNPs, Ca04468-SNP1 and Ca07571-SNP4. The sequence of the primers and three universal probes with attached FAM, HEX, and quencher Dabcyl are presented in Supplementary Material S1 . Primers and universal probes were ordered from Sigma-Merck (Australia). Master-mix preparation was the same as described by Amangeldiyeva et al. (2023). Each reaction had a 10-µl cocktail in total and was loaded in a 96-well BioRad microplate sealed with clear tape prior to amplification in a CFX96 Real-Time PCR Detection System (Bio-Rad, USA) with automatically recorded fluorescence using the described protocol (Amangeldiyeva et al., 2023). Amplification of FAM and HEX was checked and controlled, whereas genotyping results were determined in a post-run step automatically using the software CFX Manager accompanying the qPCR instrument.

The CAPS method was used for verification of Ca07571-SNP4. Primers were re-designed to make the amplicon shorter to avoid multiple cutting. Sequences of the primers are present in Supplementary Material S1 . The PCR reaction mix was as described above for sequencing but reduced 2× proportionally for all components with a total PCR volume of 30 µl. The PCR amplification protocol was also adjusted for shorter steps but maintained the same temperatures as follows: initial denaturation, 94°C, 2 min; 35 cycles of 94°C for 10 s, 55°C for 10 s, 72°C for 25 s, and final extension, 72°C for 1 min. After amplification, the entire 30-µl reaction was subjected to digestion with 4 µl (20 U of enzyme activity) of MnlI (NEBiolab, England), 4 µl of supplied 10× CutSmart buffer, and 2 µl of sterile water making 40 µl of total digestion mix volume. After 2 h of digestion in an incubator at 37°C, digested PCR products were separated by running in 12% polyacrylamide gel, visualized with ethidium bromide using a GelDoc imaging system (BioRad, USA). In both ASQ and CASP methods, the sources of potential error could be related to the accuracy of instruments, “human error,” and reproducibility of the received results. All experiments were carried out with three biological replicates (plants) and two technical replicates (repeated runs) for each studied sample, so that all potential errors are eliminated in the preliminary steps.

DNA extracted from leaves of individual plants as described above was adjusted to 100 ng/µl and aliquoted into 50-µl volumes and submitted to Diversity Array Technology Co., Canberra, Australia (https://www.diversityarrays.com) for genotyping using chickpea DArTseq (1.0) with 6K DArT clones. Results were presented in two major files with the Silico-DArT and SNP map used for further analysis.

Leaf samples were collected from individual plants, frozen, and ground as described for DNA extractions. TRIzol-like reagent was used for RNA extraction following the protocol developed earlier (Shavrukov et al., 2013), then cDNA synthesis and RT-qPCR analysis as described previously (Sweetman et al., 2020). Briefly, after 1 µl of DNase treatment (NEBiolab, England) and reverse transcription with 2 μg of RNA using the Protoscript Reverse Transcriptase kit (NEBiolab, England), cDNA samples were diluted with sterile water (1:10), resulting in a DNA quantity of around 10 ng of cDNA, and used for RT-qPCR analysis. KAPA SYBR Fast Universal Mix (KAPA Biosystems, USA), was used in 10-µl total volume containing 0.5 µM primers and 3 µl of cDNA, and run in a CFX96 Real-Time qPCR system (BioRad, USA). Thermal cycling conditions involved an initial melt at 95°C for 3 min, followed by 40 cycles of 95°C, 5 s, and 60°C, 20 s, with post-PCR melt curve from 60°C to 95°C increasing by 0.5°C increments every 5 s. Expression levels of target genes were normalized relative to the geometric average of two reference gene transcript levels (Bustin et al., 2009): CaELF1α, elongation factor 1-alpha (AJ004960) and CaHsp90, and heat shock protein 90 (GR406804) (Garg et al., 2010). Sequences of all gene-specific and reference primers are present in Supplementary Material S1 . At least three biological replicates (individual plants) and two technical repeats were used for each genotype and treatment.

For field experiments, the same three F₆ breeding lines and parents were used from hybrid population 1 (♀ Krasnokutsky-123 × ♂ ICC-12654) and four F₆ breeding lines and parents from hybrid population 2 (♀ ICC-10945 × ♂ Looch). Field experiments were carried out in 2021 and 2022 in the Akmola region (Kazakhstan) in a research field of S.Seifullin Kazakh AgroTechnical Research University. Volumetric water content (VWC) in the soil was measured using a portable Moisture meter (Model CS616, Campbell Scientific, Australia). At the stage of fully developed plants (45 days since seed sowing), VWC value was decreased from 80% field capacity (sowing time) to 48% (mild drought) in 2021 and to 35% (strong drought) in 2022, respectively. These mild and strong drought conditions corresponded to 10% and 30% less precipitation during the crop growth period recorded in 2021 and 2022 compared to the average for many previous years.

Chickpea seeds of the selected genotypes were included in the field test and sown at the regular scheduled time, in 10–13 May. Plots were 1 m² with a density of 10 plants per m². Each genotype had three replicated plots in a randomized block design. Seeds sown manually were watered once immediately after sowing with no further watering. Seed-related traits were measured individually in each plant after harvesting, as used in other chickpea studies (Purdy et al., 2023; Tiwari et al., 2023). For the purpose of this study, only two traits were considered as follows: seed weight per plant (SWP) and hundred seed weight (HSW), which was measured and calculated directly for the seeds of each plant.

Excel 365 (Microsoft) and SPSS 25.0.0.0 (IBM) packages were used to calculate and analyze means, standard errors, and significance levels using unpaired t-test, ANOVA, and post-hoc Tukey test. Three biological and two technical replicates were used for RT-qPCR experiments, whereas two repeats were used for seed yield traits and field trials.

Based on bioinformatic analysis for chickpea, 21 zinc finger proteins with a CCHC domain were identified ( Supplementary Material S2 ). The molecular-phylogenetic tree was constructed for chickpea and several legume species, generating six clades, A-E ( Figure 1 ). Clade A represents a diverse group among legumes but almost all proteins contained six CCHC domains and are annotated as ZF-CCHC protein 7. In chickpea, two genes were identified as belonging to Clade A. The first protein, Ca10268, contains seven CCHC domains while the more typical second protein, Ca07571, had six CCHC domains.

Stronger similarity of ZF-CCHC proteins among legumes was found in Clade B, with all of them containing five CCHC domains. However, the annotations of these proteins (e.g. NCBI, https://www.ncbi.nlm.nih.gov) were quite variable. In chickpea, only a single gene, Ca4468, was found in Clade B encoding ZF-CCHC domain protein 9. In contrast, Clade C contained five loosely related chickpea proteins, together with two proteins from M. truncatula. Two genes in Clade C, Ca11100 and Ca25010, encoded zinc knuckle proteins with a single CCHC domain, but other genes with three CCHC domains had different annotations and variable functions. Clade D appears isolated, where all proteins in legumes clustered very tightly, including a single chickpea gene Ca04752 encoding a ZF-protein with nine CCHC domains. All genes in Clade D were annotated as GIS2, Glucose inhibition of gluconeogenic growth suppressor 2, and are involved in interaction with DNA.

Clade E contained the largest number of proteins, including several from chickpea, and these were annotated as CSDP, cold shock domain proteins, and GRP, Glycine-rich RNA-binding proteins RZ1. The proteins in this Clade contained a diverse number of CCHC domains with as many as 11 domains in Ca17110 to the more typical five domains in Ca09965, and the rest contained only one or two CCHC domains. The last Clade F contained only two chickpea proteins, Ca24838 and Ca26496, and was identified as a typical Zinc knuckle family protein with two CCHC domains but unrelated to CSP or GRP.

For this study, two chickpea genes, Ca07571 and Ca04468, from Clades A and B, respectively, different in a combination of zinc finger domains, were selected for further analysis and functional characterization representing genes potentially responsive to drought and dehydration ( Figure 1 ; Supplementary Material S2 ). The first gene, Ca07571 (accession: XM_012716319), encoded “Zinc finger CCHC domain-containing protein 7” with six CCHC motifs, 528 aa (accession: XP_004502023), UniProt: A0A1S2YAP2. The gene Ca07571 was located in the annotated reference genome of chickpea cv. Frontier, Ca5: 40,504,781–40,507,811, and has similarity with two genes in A. thaliana, At3g43590 and At5g36240, annotated as “Zinc knuckle (CCHC-type) family protein”.

The second selected gene, Ca04468 (accession: XM_004496505), encoded “Zinc finger CCHC domain-containing protein 9” with five CCHC motifs, 262 aa (accession: XP_004496562), UniProt: A0A1S2XZY1. The gene was located in the reference genome of chickpea cv. Frontier, Ca4: 12,446,515–12,449,051, and has a similarity with the A. thaliana gene, At5g52380, annotated as “Vascular-related NAC-domain protein 6”.

Sanger sequencing results for open reading frames (ORF) of both genes, Ca04468 and Ca07571, in six chickpea germplasm accessions as follows: Kamila, Krasnokytsky-123, and Looch (Kabuli type) and ICC-1083, ICC-10945, and ICC-12654 (Desi type) revealed strong conservation, and no SNPs were identified (data not shown). In contrast, promoter regions of both genes contained four SNPs each ( Figure 2 ; Supplementary Material S3 ). The distribution of the identified SNP was different in each gene among chickpea accessions in comparison to two reference chickpea genomes cv. Frontier (Kabuli type) and accession ICC-4958 (Desi type).

For gene Ca04468, in chromosome Ca4, all four SNP were present in the reference genomes of cv. Frontier and accession ICC-4958 ( Figure 2A ). SNP1 [C/T] was widely distributed among the accessions, while SNP2 [T/C], SNP3 and SNP4 [both A/T] also represented haplotypes for Frontier and ICC-4958, respectively.

For gene Ca07571, in chromosome Ca5 ( Figure 2B ), the first three SNPs, at the greatest distance from the Start codon, represented rare alleles and occurred in a single accession only, i.e., SNP1 [A/T] and SNP3 [A/T] in cv. Looch, while a “doubled” SNP2 (two very closely located SNP with just five bp between them), both [T/C], was found only in cv. Kamila. These SNPs were not found in either reference chickpea genome. In contrast, SNP4 [G/A], the most proximally located to the ORF (–261 bp) was equally distributed among six accessions: three Kabuli-type cultivars (Kamila, Krasnokutsky-123, and Looch) had SNP4 [G] similar to reference cv. Frontier, while SNP4 [A], as in reference genome of ICC-4958, was also identified in three Desi-type accessions (ICC-1083, ICC-10945, and ICC-12654). Additionally, SNP5 [T/C] was common between the reference genomes of cv. Frontier and accession ICC-4958, but all six studied chickpea germplasms were monomorphic for SNP5 [C] ( Figure 2B ).

Finally, one SNP was selected for each gene and confirmed in the parents of two hybrid populations as follows: Ca04468-SNP1 [C/T] was suitable for parents, ♀ Krasnokutsky-123 and ♂ ICC-12654, labeled as hybrid population 1, while Ca07571-SNP4 [A/G] was selected for the other parents, ♀ ICC-10945 and ♂ Looch, hybrid population 2 ( Figure 3 ).

Genotyping of chickpea plants was carried out based on SNP in each targeted gene, Ca04468-SNP1 and Ca07571-SNP4, as described above. Parents and breeding lines developed from the progenies were examined. For Ca04468-SNP1, hybrid population 1 [♀ Krasnokutsky-123 and ♂ ICC-12654] was used, while hybrid population 2 [♀ ICC-10945 × ♂ Looch] was selected for Ca07571-SNP4 genotyping. The parents of both hybrid populations were used as reference genotypes for comparison with their breeding lines ( Figure 3 ).

Genotyping was based on the ASQ method and examples of clear amplification of FAM or HEX fluorescence in parents Krasnokutsky-123 or ICC-12654 for Ca04468 gene are shown ( Figures 4A, B ). Examples of allele discrimination are shown in Figures 4C, D for genes Ca04468 and Ca07571, respectively. Results of SNP genotyping for Ca04468 and Ca07571 genes are present in Supplementary Material S4 .

The results from SNP genotyping via the ASQ method were confirmed using CAPS, but only for SNP4 in the Ca07571 gene, because there was no suitable restriction enzyme site targeting the SNP1 fragment in the Ca04468 gene.

For CAPS marker Ca07571-SNP4, a MnlI restriction site was suitable, with the recognition sequence “CCTC” and “GGAG” (in the complemented strain). Due to the presence of several recognition sites in the region surrounding the SNP, a shorter amplification fragment was designed producing a 77-bp digested fragment easily visible for allele [G] in Kabuli-type chickpea (Kamila, Krasnokutsky-123, and Looch). In contrast, there was no such fragment in genotypes with allele [A] in Desi-type chickpea (ICC-1083, ICC-10945, and ICC-12654).

Results of CAPS marker analysis of ICC-10945 and Looch, parents of hybrid population 2, and eight of their breeding lines are shown in Figure 5 after separation either in agarose or in polyacrylamide gels. The paternal genotype Looch and three breeding lines confirmed the [G] allele in Ca07571-SNP4.

DArT-seq analysis was applied to study genetic polymorphism and variability among a set of chickpea accessions and several hybrid populations, including hybrid populations 1 and 2, described above. After an initial filtration of the 6,000 DArT markers, 3,600 of them were polymorphic, and following that, 1,600 DArT markers had known mapping locations in the chickpea genome. For this study, DArT markers identified in the physical map of chromosomes Ca4 and Ca5 cv. Frontier were used for further analysis ( Figure 6 ). Sequences and genetic positions of the used DArT markers and raw-data for DArT microarray analysis are present in Supplementary Materials S5 , S6 , respectively.

Among 298 DArT markers in chromosome Ca4 with approximately 59M bp in length, five major genetic blocks were found indicating differences in the haplotypes between parents of hybrid population 1, Krasnokutsky-123, and ICC-12654. These DArT haplotype blocks varied in size from 0.22M bp (Block 2) to 5.9M bp (Block 3), and blocks 1–3 and 4–5 were located in opposite arms of chromosome Ca4 ( Figure 6A ).

For chromosome Ca5, which is approximately 69M bp in length, only 176 DArT markers were suitable for the analysis. Haplotypes of ICC-10945 and Looch, parents of hybrid population 2, were also characterized by five major genetic blocks. Blocks 5 and 2 were the shortest and longest DArT haplotype blocks and accounted for about 1M and 5.3M bp, respectively. The five haplotype blocks were distributed relatively similarly along chromosome Ca5 and without a tendency for concentration on the chromosome arms ( Figure 6B ). The density of DArT markers in chromosome Ca5 was fewer by 1.7-fold, compared to chromosome Ca4, and resulted in less identified SNPs in each haplotype block in chromosome Ca5 compared to chromosome Ca4, respectively.

A more detailed zoom of the genetic regions is shown in Figure 6C . For gene Ca04468 in chromosome Ca4, the region between DArT markers 13146196 and 10263426 covered approximately a 1.8M-bp fragment of the chromosome. DArT markers 50759538 and 5824740 were identified as flanking the target gene Ca04468. Based on the flanking genetic region, four out of six hybrid breeding lines had a DArT haplotype identical to the maternal parent cv. Krasnokutsky-123. The haplotypes of two breeding lines H18-1 and H18-5 were identical to the paternal parent ICC-12654. All recombination events in the entire genetic region took place in four breeding lines, but all in the proximal part, i.e., between DArT markers 5825634 and 10263417 in lines H18-1 and H18-4, and between DArT markers 5824874 and 5826066 in lines H18-5 and H18-6 ( Figure 6C ).

Similarly, for gene Ca07571 in chromosome Ca5, the region between DArT markers 10269623 and 35486562 covered a chromosome fragment of approximately 1.9M bp. The flanking DArT markers 23889105 and 23886128 surrounded the target gene Ca07571. Three hybrid breeding lines (H35-1, H35-5, and H35-7) had a DArT haplotype identical to the maternal parent ICC-10945. The haplotype of the paternal parent cv. Looch was found in the remaining four hybrid breeding lines. Four breeding lines, H35-2, H35-4, H35-6, and H35-7, have undergone various recombination events in the entire genetic region, both in proximal and distal parts from the target gene Ca07571 ( Figure 6C ).

For further studies on gene expression, the following genotypes were selected: (1) the first three breeding lines (H18-1, H18-2, and H18-3) from hybrid population 1 together with their parents, Krasnokutsky-123 and ICC-12654; and (2) the first four breeding lines (H35-1, H35-2, H35-3, and H35-4) from hybrid population 2 and their parents, ICC-10945, and Looch.

Under drought stress, chickpea plants from the parental accessions and breeding lines from hybrid populations 1 and 2 showed both some similarities and differences. The expression of the Ca04468 gene was diverse among genotypes in each hybrid population ( Figure 7A ). For example, in parent Krasnokutsky-123, the expression level of Ca04468 was significantly higher at all time-points of drought treatment compared to Controls. A similar pattern was observed in breeding line H18-3, and it was very high in line H18-2 especially after 7 days of drought stress. In contrast, no changes in Ca04468 expression under drought was found in parent ICC-12654 and breeding line H18-1. A different situation was found in hybrid population 2, where both parents, ICC-10945 and Looch, as well as two breeding lines, H35-2 and H35-3, showed increased mRNA level and corresponding gene expression. Unexpectedly, no changes in Ca04468 expression were found in line H35-1, while only delayed increase in expression was recorded in line H35-4 after 9 days of drought treatment. The reference genotype ICC-4958 with a high level of mRNA production was closer to genotypes of Looch and ICC-10945 rather than to ICC-12654 ( Figure 7A ).

The situation was dramatically different when the genes were studied in plants after dehydration treatment ( Figure 7B ). For the Ca07571 gene, the parents and breeding lines of hybrid population 1 showed constantly low expression and downregulation in some cases. However, the Ca07571 mRNA level was quite different in the parents and breeding lines from hybrid population 2. The parent ICC-10945 had no change in Ca07571 expression, while parent Looch and three breeding lines, in contrast, had a high increase in gene expression. The exception is in breeding line H35-1, which showed an increased expression level after 4 h of dehydration, but it was still much smaller than the other three breeding lines in hybrid population 2. The reference genotype ICC-4958 showed no changes in Ca07571 expression under dehydration, similar to the majority of the studied genotypes, but differing from Looch ( Figure 7B ).

The parents and selected breeding lines from hybrid populations 1 and 2 were evaluated for seed-related traits (seed weight per plant, SWP; and 100-seed weight, HSW) in field experiments in Northern Kazakhstan for 2 years under mild and strong drought conditions ( Figure 8 ).

SWP did not vary significantly among genotypes during mild drought stress in 2021 regardless of their alleles of both Ca04468 and Ca07571 ( Figure 8A ). In contrast, clear discrimination for SWP was found between genotypes in conditions of strong drought in the field study in 2022. Parent ICC-12654 and breeding line H18-1 showed a significantly lower SWP, and they also shared the same allele Ca04468-SNP1, which can be described as “unfavorable” for this trait. Other breeding lines H18-2 and H18-3 with Ca04468-SNP1 allele identical to the maternal parent Krasnokutsky-123 showed significantly higher SWP. Similarly, in hybrid population 2, a significantly lower SWP was recorded in maternal parent ICC-10945 and one breeding line H35-1. They have the same allele Ca07571-SNP4, which is also unfavorable for this trait. Another allele is present in the paternal parent Looch and the remaining three breeding lines H35 with a significantly higher SWP ( Figure 8A ).

Different results were found for HSW, where alleles of Ca04468 and Ca07571 were not associated with this trait ( Figure 8B ). In hybrid population 1, the seeds of paternal parent ICC-12654 and breeding line H18-2 showed a significantly smaller HSW, while other genotypes had a significantly higher HSW. In hybrid population 2, no significant differences were recorded for HSW among all genotypes, while the maternal parent ICC-10945 recorded a slightly lower HSW, probably the result of smaller seeds, but not significantly lower compared to other genotypes. The results indicate that alleles of Ca04468 and Ca07571 did not affect the HSW trait in chickpea genotypes under conditions of mild and strong drought ( Figure 8B ).