An insight into tissue culture-induced variation origin shared between anther culture-derived triticale regenerants

Background The development of the plant in vitro techniques has brought about the variation identified in regenerants known as somaclonal or tissue culture-induced variation (TCIV). S-adenosyl-L-methionine (SAM), glutathione (GSH), low methylated pectins (LMP), and Cu(II) ions may be implicated in green plant regeneration efficiency (GPRE) and TCIV, according to studies in barley (Hordeum vulgare L.) and partially in triticale (× Triticosecale spp. Wittmack ex A. Camus 1927). Using structural equation models (SEM), these metabolites have been connected to the metabolic pathways (Krebs and Yang cycles, glycolysis, transsulfuration), but not for triticale. Using metabolomic and (epi)genetic data, the study sought to develop a triticale regeneration efficiency statistical model. The culture’s induction medium was supplemented with various quantities of Cu(II) and Ag(I) ions for regeneration. The period of plant regeneration has also changed. The donor plant, anther-derived regenerants, and metAFLP were utilized to analyze TCIV concerning DNA in symmetric (CG, CHG) and asymmetric (CHH) sequence contexts. Attenuated Total Reflectance–Fourier Transfer Infrared (ATR-FTIR) spectroscopy was used to gather the metabolomic information on LMP, SAM, and GSH. To frame the data, a structural equation model was employed. Results According to metAFLP analysis, the average sequence change in the CHH context was 8.65%, and 0.58% was de novo methylation. Absorbances of FTIR spectra in regions specific for LMP, SAM, and GSH were used as variables values introduced to the SEM model. The average number of green regenerants per 100 plated anthers was 2.55. Conclusions The amounts of pectin demethylation, SAM, de novo methylation, and GSH are connected in the model to explain GPRE. By altering the concentration of Cu(II) ions in the medium, which influences the amount of pectin, triticale’s GPRE can be increased. Supplementary Information The online version contains supplementary material available at 10.1186/s12870-023-04679-w.


Background
With the development of in vitro tissue culture techniques, morphologically atypical regenerants representing somaclonal variation (or tissue culture-induced variation, TCIV) were noticed [1].Some authors have questioned the phenomenon, claiming that the variation was caused by the so-called preexisting variation related to differences between individual cells of the donor plant [2].It was simple to eliminate unusual regenerants that displayed unexpected morphological variation during culture, but somaclonal variation was contentious and has been the subject of ongoing debate [3][4][5][6].DNA marker-based studies have shown that TCIV is linked to mutations [7], but some have indicated that the link is not clear [8] or was not found [9].Some data suggested the significance of DNA methylation changes [10].Furthermore, the relationship between components (sequence variation, DNA demethylation and de novo DNA methylation) of TCIV and green plant regeneration efficiency (GPRE) from anther cultures was proposed [11].One of the biggest problems with TCIV and GPRE was that there needed to be more tightly defined biological materials to eliminate or lessen the effect of the preexisting variation [12].Moreover, to reduce the amount of variation induced during doubled haploid donor plant regeneration it was suggested utilizing the generative progeny of double-haploid donor plants [13].However, in some species, more than a single generative cycle following doubled haploid production may be required [10,14].It was also hard to find a DNA-based marker system that could quantify both changes in DNA methylation patterns and differences in sequences between regenerants and the donor plant simultaneously.The restriction was successfully overcome with the creation of the metAFLP methodology [15].
With the growing number of studies, it became apparent that the TCIV is a complex phenomenon involving DNA methylation within a variety of sequence contexts, including symmetric (CG, CHG) and asymmetric (CHH) sequences, and that the methylation of these sites differs in their epigenetic background [16][17][18].It was shown that the DNA methylation pattern and the level of sequence variation induced de novo during tissue culture plant regeneration (androgenesis [13,19] and embryogenesis [20,21]) depend on species.It was also suggested that methylation may participate in GPRE [22].Unfortunately, studies of TCIV failed to explain the nature of the phenomenon (for more information on this, see the reviews [23,24]).Therefore, it was hypothesized that it might have a biochemical background.A series of studies conducted on barley and triticale demonstrated that S-adenosyl-L-methionine (SAM) [25], glutathione (GSH) [26], b-glucans [27]/pectins [28], and Cu(II) ions [22] are involved in GPRE.The metabolites, via structural equation models [28], were linked to the Krebs and Yang cycles and biochemical pathways, including glycolysis or transsulfuration.However, no generalized scheme linking the metabolome, TCIV and GPRE has been evaluated, including triticale.
We hypothesize that GPRE and TCIV have biochemical backgrounds linked to glycolysis, the Krebs and Yang cycles, and the transsulfuration pathway that influences GSH production and affects GPRE.Propper functioning of glycolysis during plant regeneration via anther culture may require copper ions as they act as cofactors of many enzymatic reactions.However, the relationships between metabolome, TCIV, and GPRE in triticale must be made apparent and should be investigated.
The study aimed to evaluate a statistical model of anther culture plant regeneration efficiency utilizing metabolomic data reflecting biochemical processes linked to DNA sequence and methylation changes induced via in vitro cultures of triticale.Such a model is required to progress in theoretical understanding and future practical knowledge utilization to increase GPRE in cereals.

Results
A single double haploid plant's generative progeny was chosen randomly, providing the explant tissue.Thirtyseven morphologically consistent regenerants that were identical to the donor plant were created by the eight experiments (A-H).The Cu(II), and Ag(I) ion concentrations in the induction medium (IM) and the length of the in vitro anther cultures differed between these eight investigations.There were anywhere between three and ten regenerants in each trial.Trial C had the lowest GPRE value, whereas trial H had the highest (Additional file 1; Tab S1).The fourteen-days-old leaf of a donor and leaves from the 37 regenerants were employed to extract the genomic DNA of sufficient quality and quantity for the metAFLP analysis.Quantitative metAFLP analysis of banding patterns shared by the donor plant and its regenerants revealed that the presence of sequence variation within the CHH asymmetric sequence context ranged from 8.48 to 8.91%, with the mean value equaling 8.65%.The CHH context's de novo methylation ranged from 0.36 to 0.76% (mean value 0.58%) (Additional file 1; Tab S1).
The FTIR analysis delivered spectra previously assigned to metabolites (GSH: 2550 − 2540 cm −1 [26], SAM: 1630…1470 cm −1 [25]), representing biochemical pathways and cycles, and proved to be essential for partial structural equation models.The mean absorbance values of low methylated pectins (LMP), SAM, and GSH across all samples encompassing all trials were 0.5512, 3.9617, and 0.00515, respectively.All trials' mean green plant regeneration efficiency was 2.55 and varied from 0.71 to 6.06.Minor standard errors, standard deviations, and variances were calculated for all variables across all samples (Additional file 1: Tab S1).
Based on 37 regenerants from eight experimental trials, the structural equation modeling analysis (Table 1) revealed a minor departure from the normal distribution, as shown by skewness and kurtosis values.All of the variables were quantitative and met the requirements of the Lindeberg-Lévy theorem [29], stating that if the formation of a variable of natural origin is influenced by a large number of primarily random factors, then the distribution of this variable is asymptotically convergent to the normal distribution.Thus, the variables' asymptotic distribution is anticipated to approach their theoretical normal distribution.
Pearson's linear correlation coefficients (Table 2) showed that Cu (II) was positively correlated with GPRE and FTIR GSH range, while a negative correlation was evaluated for CHH_SV.Ag(I) was negatively related to CHH_DNMV, whereas the time of anther culture was positively related to CHH_SV and negatively related to FTIR LMP range.CHH_DNMV was positively correlated with the FTIR SAM range.A positive correlation was also evaluated between FTIR SAM and FTIR GSH range.Furthermore, the FTIR GSH range was positively correlated with GPRE.The other correlations were insignificant.
Data from previously published partial structural models of the relationships between CHH_SV, CHH_DNMV, Cu(II), Ag(I), time of anther cultures, β-glucans/pectins, GSH, and GPRE were employed (Ag + [30], GSH [26], SAM [25], GSH and SAM [28]).Five endogenous variables (described by the model; CHH_DNMV, CHH_SV, FTIR absorbance in the region assigned to SAM, FTIR absorbance in the region assigned to GSH and GPRE) and two exogenous (not described within the model) (FTIR absorbance in the region assigned to LMP, Cu(II)) variables make up the proposed model.The relationships

Table 1 Descriptive statistics of the variables present in the postulated models
The 990.950 cm −1 reflects continuous FTIR spectrum starting from 990 and ending at 950 and encompasses values calculated every ten units, the range can be assigned to low methylated pectins (LMP); similarly the 2550 − 2540 cm −1 FTIR reflects data related only to the 10 units range and is assigned to glutathione (GSH), whereas the 1630…1470 cm −1 FTIR range encompasses combined spectra from the given range but the data is not continuous (some ranges were not included into combined variable) [28], the range is assigned to S-adenosyl-L-methionine (SAM)  had a single covariance.All paths were non-recursive.There were five residuals in the model (Fig. 1).Chi-square statistics analysis of the model fit revealed that the model was not significant (Table 3).A low χ 2 value indicates a well-fitting model when the sample size is big.However, the test may produce inaccurate results when the sample size is limited (as in the case of our data) [31].In such a situation, it is advised to employ the chi-square test as the sole information criterion [32].Thus, other models' descriptive goodness-of-fit measures were applied.The model fits perfectly the experimental data, as evidenced by the χ 2 /df being less than 3, the Root Mean Square Residuals (RMR) being under 0.05, and the Root Mean Square Error of Approximation (RMSEA) being below 0.05.The RMSEA was also within the 90% confidence interval.The Standardized RMR (SRMR) was < 0.08, indicating that the model was well-fitted.Except for the Adjusted Goodness-of-Fit Index (AGFI) and Relative Fit Index (RFI), which were slightly below the 0.9 cut-off values [33], nearly all goodness-of-fit measures appeared to match the proposed requirements (above 0.95 but not below 0.9).The fact that the Parsimonious Normed Fit Index (PNFI) and Parsimonious Comparative Fit Index (PCFI) were so modest suggests model complexity.The Expected Cross-Validation Index (ECVI) is within the 90% confidence interval, indicating that the data should be repeatable using the same sample size.Last but not least, Hoelter's 0.05 index was 67, indicating that at least 67 samples needed to be utilized in the SEM design in order to create a model with a p = 0.05 significance (Table 3).
All the paths' (β) coefficients (except CHH_SV→GPRE) of the hypothesized model were significant (Table 4).The effect of Cu(II) on GPRE was the highest and most positive, followed by the FTIR SAM range on the FTIR GSH range as indicated by standardized estimates.The effect of Cu(II) on CHH_SV was the next one but negative.The effects between CHH_DNMV and CHH_SV, FTIR GSH range and GPRE, FTIR LMP range and FTIR SAM range, and Cu (II) and FTIR GSH range were also strong and positive.The effects of the FTIR SAM range and CHH_ DNMV on GPRE were moderately strong and negative.The only covariate relationship between Cu(II) and FTIR LMP range was insignificant.
The hypothesized model includes direct, indirect, and total effects (Table 5).Direct effects indicate the direct impact of variables on each other.Indirect effects, however, indicate the mediation effect of variables mediating the described relationship and may also indicate relationships of variables for which no direct effects were observed.The GPRE variable showed the most significant dependence on Cu (II) (total β = 0.8226; including the direct effect as β = 0.8252 and negative and small indirect effect).The FTIR GSH range was the second and the most influential variable with a positive direct effect on GPRE (β = 0.3908).The CHH_DNMV, and FTIR SAM range affected GPRE preferentially via negative direct effects; however, indirect effects were positive.Finally, the FTIR LMP range affected GPRE via small, negative and indirect effects.FTIR GSH range was affected mainly by the FTIR SAM range with a positive direct effect (β = 0.6339).It was also directly and positively controlled by Cu(II) and indirectly by LMP (β = 0.3437 and β = 0.2462, respectively).CHH_SV was negatively dependent on Cu (II) (primarily via direct (β = -0.5158)effect with total equal β = -0.3955)but positively dependent on CHH_DNMV (via direct and total effect (β = -0.4082).CHH_DNMV depended on the FTIR SAM range via direct effect (β = 0.3697) followed by Cu(II) (β = 0.2947) and by FTIR LMP range via positive indirect effect (β = 0.1436).Lastly, the FTIR SAM range was directly affected by the level of pectin demethylation (β = 0.3884).The effect was positive.
The Ag(I) failed to show significant effects when GSH was assumed (the characteristic was not included in the model), despite being significant in our previous model describing similar relationships but without GSH [30].We also failed to include the Time of another culture in the model.

Discussion
A complex statistical method known as "structural equation modeling" was used to test the working hypotheses that in vitro tissue culture conditions affect anther tissue culture plant regeneration and biochemical pathways and cycles and that the biochemical background regulates GPRE, which is under Cu(II) ions' control acting as a cofactor in enzymatic reactions.In biology, the reflection of theoretical assumptions is often known, and the effects of specific variables in a model are usually big.Because of this, the sample size is less critical than in psychology, where only minor effects are usually assumed.Such a situation was present in the current study, where the model was constructed based on 37 samples and included many variables.Different aspects of the final model were tested to overcome obvious statistical limitations.
We recently used the SEM method to build models that explain TCIV or GPRE.It was shown that sequence variation is linked to DNA methylation status change [22] and that S-adenosyl-L-methionine, a compound made from ATP and L-methionine, is involved in the process.ATP synthesis requires copper ions in the Krebs cycle's electron transfer chain complex IV, which links the TCA cycle with the Yang cycle.SAM is a crucial molecule the cell uses for methylating other cellular compounds [34].So, the model showed that SAM is involved in TCIV.Furthermore, it shows that changes in the CHH sequence context of DNA methylation play a role in sequence variation and GPRE.Evidence demonstrates that GSH may be critical in green plant regeneration efficiency [26].Its importance was demonstrated experimentally in the rye (Secale cereale L.) and triticale [35,36].Using SEM, we discovered close relationships between GSH and GPRE, allowing us to include metabolite variables in the SEM model, demonstrating that the Yang cycle disturbances influence GSH synthesis via the transsulfuration pathway [34,37].Furthermore, we have also suggested that the Krebs cycle may also be affected by in vitro tissue culture manipulation [28].It was demonstrated that glycolysis could use β-glucans as a carbon source to pump the barley's TCA cycle [38].A comparable study in triticale yielded inconclusive results [28].Nevertheless, the study suggested that pectin demethylation level could indirectly affect TCA via acetyl-CoA.Pectins partly affect ascorbic acid (AA) synthesis, and AA turns on pyruvate dehydrogenase, which is part of the pyruvate dehydrogenase complex.The latter is responsible for pyruvate's oxidative decarboxylation, leading to acetyl-CoA used by the Krebs cycle.Following the reasoning, we have tested the relationships between the FTIR signal from low methylated pectins, the TCA, and the Yang cycles.Thus, distinct but closely linked biochemical background aspects of the TCIV and GPRE work in concert [28].
The study's main finding was the importance of Cu(II) in affecting GPRE and sequence variation.Cu(II) can cause DNA modifications via oxidizing methylated cytosines, resulting point mutations [39].Cu(II) may also work with Zn/Cu-dependent dismutase to remove excess ROS [40] that can cause DNA mutations.Since several SEMs showed links between different parts of TCIV and GPRE, we found combining all models into a single scheme justified, considering that tissue culture experiments only looked at a small number of regenerants.
The fact that CHH_DNMV adversely influences GPRE may indicate that gene expression regulation affects the phenomenon and that CHH_DNMV may even block GPRE.The concept is consistent with earlier research showing that sequence variation impacting the CHH context may have had little to no effect on GPRE.Contrarily, the impact of the CHH on GPRE comes from the interaction of methylation cytosines with Cu (II).Pectin's demethylation level indirectly affects SAM modulating DNA de novo methylation, which is consistent with the model's biochemical background.Pectins are also linked to GSH via the ascorbate-glutathione cycle.However, we did not test whether or how much ascorbic acid (AA) synthesis is affected by the in vitro culture conditions.Still, considering the role of GSH and the Krebs cycle involved in the GPRE and TCIV, AA is an integral part of the phenomenon that links oxidative processes.The model demonstrates that SV is only one of several minor issues affecting GPRE.However, DNA methylation changes should still be considered because they may control how much a gene is expressed.However, the fact that changes in DNA methylation have a negligible effect on GPRE may mean that problems with biochemical cycles and pathways are more important.The provided model has practical ramifications in addition to its scientific significance.It shows that adjusting the concentration of copper ions in induction media (IM) can modify cell activity and increase GPRE.
Our previous studies have demonstrated that SAM and GSH impact GPRE.SAM achieves this by methylating the asymmetric CHH sequence, which is related to de novo methylation [25].At the same time, GSH does so through an epigenetic process that is not fully understood [26].β-glucans may be the primary source of glucose during tissue culture carbon shortages [27] and may, therefore, supply energy to glycolysis, thereby affecting the Krebs cycle.Similarly, pectins are involved in the production of ascorbic acid and may indirectly affect the Krebs cycle via acetyl-CoA.Moreover, pectins are the most likely metabolites involved in the interactions between SAM, GSH, Cu(II), and GPRE [28].Lastly, we cannot ignore the positive impact of copper ions, which help elevate GPRE levels [28].The current model includes LMP, and its significance is based on the significance of the former SEMs.Unfortunately, with a limited sample size (limited number of regenerants), we could not build such an advanced model without demonstrating that all preceding SEMs were significant.Still, our results show that the model explaining GPRE is not complete.We hypothesize that some unrecognized or unconsidered enzymatic processes or epigenetic regulation of gene expression that result in GPRE possibly involving Cu(II) as a cofactor should be taken into account.The vast and robust impact of GSH on GPRE and the effect of SAM on GSH lend some support to this idea.

Conclusions
The structural equation model shown is based on biochemical cycles and pathways involving the metabolites involved in TCIV, GPRE, and epigenetic mechanisms.The model explains GPRE by connecting LMP, SAM,

Methods
The current study was based on former experiments involving specially designed biological materials [13], the metAFLP approach [19], FTIR data concerning β-glucans/ pectins [27,28], SAM [25], and GSH [26].A structural equation model [30] was used to frame the data.A donor plant is a generative offspring of a double haploid in the biological system.In order to regenerate new plants under various Cu(II), Ag(I) ion concentrations in the induction medium (IM) and time of anther culture (covers period starting from the plating of anthers on the IM to the collection of calli, embryolike structures, or somatic embryos, and their transfer on regeneration media), the donor plant functioned as a source of explant tissues (anthers) (Additional file 1: Tab S1).Eight different A-H trial conditions were prepared to test the effect of Cu(II), Ag(I) ions, and time of anther cultures on the tissue culture-induced variation and the number of green regenerants obtained.Green plant regeneration efficiency was calculated for each trial based on the estimated number of green regenerants per 100 plated anthers.The donor plant and its regenerants were assessed for the following physical features (plant growth, leaf shape, color and width, tillering mode, and spikes number) [13].
The unique properties of the isoschizomers Acc65I and KpnI, which have varying sensitivity to the site's DNA methylation pattern and its surroundings, are the foundation of the metAFLP approach.Using two metAFLP platforms (Acc65I/MseI and KpnI/MseI) for the amplification of DNAs from a donor plant and its regenerants, the method enables quantification of sequence and DNA methylation events that could be standardized, resulting in respective variations.
Samples for the mid-infrared spectroscopy were lyophilized and homogenized into powder (ball mill, MM 400, Retsch, Haan, Germany).Nicolet iN10 MX infrared imaging microscope (Thermo Fisher Scientific, Waltham, Massachusetts, USA), equipped with a deuterated triglycine sulphate (DTGS) detector and a KBr beam splitter, was used for the measurements.Using the one-bounce diamond crystal and the ATR accessory, 64 spectra per sample were obtained in the Attenuated Total Reflectance (ATR) mode with a 4 cm −1 resolution in the wavenumber range between 600 and 4000 cm −1 (Smart Orbit, Thermo Scientific, Madison, WI, USA).Before each measurement, the diamond crystal's surface was cleaned with water or propanol to get rid of any leftovers from previous samples.OMNIC software was used to record, average, and baseline-correct the spectra (v.9.0, Thermo Fischer Scientific Inc.).The ChemoSpec [41] package of the R programming language [42] used to accomplish the normalization to the unit area inside the 1800 − 900 cm −1 wavenumber region, statistics (mean, SD), and printing the spectra.
In order to build a theoretical illustration of green plant regeneration efficiency using AMOS [43] under SPSS v. 28 [44] software, metAFLP data evaluated based on the donor plant and its regenerants as well as the absorbance of the FTIR spectra bands assigned to LMP, SAM, and GSH were incorporated in a structural equation model.

Fig. 1
Fig. 1 The schematic illustration of relationships between the biochemical pathways and cycles and the hypothesized SEM model, along with FTIR spectra bands specific for the metabolomic compounds included in the model The hypothesized SEM model with variables Cu(II), CHH_DNMV, CHH_SV, the FTIR absorbances at: 990.950 cm −1 assigned to low methylated pectins (LMP), 1630…1470 cm −1 assigned to S-adenosyl-L-methionine (SAM), and 2550 − 2540 cm −1 assigned to glutathione (GSH) as a function of GPRE (in purple).Red arrows indicate the SEM paths, whereas metabolites in the SEM are indicated by blue fonts and by blue arrows, indicating their position in biochemical paths and cycles and respective FTIR bands.The action of pectins via ascorbic acid (AA) on the tricarboxylic acid cycle (TCA) / Krebs cycle and the involvement of AA in the ascorbate-glutathione cycle, as well as other paths and cycles are given in green.Cu(II) (in red) denotes the concentration of copper ions (M) in the IM.GPRE stands for green plant regeneration efficiency (number of regenerants per 100 plated anthers).CHH_SV and CHH_DNMV are the metAFLP quantitative characteristics of sequence variation and de novo DNA methylation variation affecting CHH sequence contexts (in the dark), respectively λ 1 -λ 12 path coefficients, δ 1 -δ 5 residuals (experimental errors)

Table 2
Pearson's linear correlation coefficients for analyzed variables *Correlation is significant at the 0.05 level (2-tailed) **Correlation is significant at the 0.01 level(2-tailed)

Table 3
Summary of the analyzed structural equation model

Table 5
Direct, indirect and total effectsss for the analyzed model