Hybrid aspen (Populus tremula L. × tremuloides Michx.) clone T89 was used as wild-type and all transgenic lines were made in this genetic background. Transgenic lines initially tested in the greenhouse included those expressing 35S:AnAXE1 (Pawar et al., 2017b), 35S:HjAXE and WP:HjAXE (Ratke et al., 2015; Wang et al., 2020), as well as lines with RNAi constructs targeting hybrid aspen RWA-C and RWA-D genes, denoted 35S:RWA-CD (previously called 35S::CD-RWA RNAi), and all four RWA genes, denoted WP:RWA-ABCD (previously called pGT43B::RWA-ABCD RNAi) (Pawar et al., 2017a). Two lines expressing a β-glucuronidase (GUS) gene under the control of the WP promoter (Ratke et al., 2015) were used as transgenic controls. Additional lines were generated that expressed WP:AnAXE1 using the pK-pGT34B-GW destination vector (Ratke et al., 2015) and the AnAXE1 cDNA as previously described (Pawar et al., 2017b). Each construct was represented by two to four lines selected from among approx. 20 independent lines, as described previously, based on the strength of transgene expression and superiority of greenhouse performance with regard to growth and saccharification properties (as described by Ratke et al., 2015; Pawar et al., 2017a, b; Wang et al., 2020), with the exception of WP:AnAXE1. The WP:AnAXE1 lines were selected based on the strength of transgene expression in plants cultivated in vitro, as determined by RT-PCR analysis (Supplementary Figure S1). A list of the lines and constructs is given in Table 1.

Transgenic trees were propagated in vitro at the Umeå Plant Science Centre transformation facility in Umeå, Sweden (Nilsson et al., 1992) and transplanted into soil for one month of acclimatization in the greenhouse in early spring 2014. Then, with permission from the Swedish Board of Agriculture (DNR. 4.6.18-761/14), the trees were moved to sheltered outdoor premises in Umeå for a two-week hardening period, before translocation to the field site (ca 1000 km south of Umeå) in Våxtorp, Laholm community, Sweden (56.42°N, 13.07°E). Between August 4 and 8, 2014, the trees were planted in the field, with a 3 m spacing, on abandoned farm land fenced according to the requirements for genetically modified plants. In total, 636 trees included in the current analysis, along with other trees not analyzed here, were arranged in a 14 block design, with two trees of each transgenic line randomly distributed within each block along with four wild-type (WT) trees (Supplementary Figure S2). For weed control, the field was harrowed twice a year during the first two years following planting and grass was mowed twice a year during subsequent years. All biosecurity and safety procedures for field trials with transgenic plants required by the Swedish Board of Agriculture have been adhered to.

The stability of GUS gene expression was investigated in July of year four (2017) during the period of active cambial growth. The basal part of a two-year-old branch was hand sectioned and the sections were prefixed in acetone for 30 min, washed with water, placed in the reaction solution (1 mM X-GlcA (5-bromo-4-chloro-3-indolyl β-D-glucuronide), 50 mM Na-phosphate buffer pH 7, 0.1% (v/v) Triton, 1 mM K₃[Fe(CN)₆], 1 mM K₄[Fe(CN)₆]), and incubated for 3 days in the dark at room temperature. Sections were then fixed in FAA (50% ethanol, 5% formaldehyde, 10% acetic acid, all v/v) overnight followed by clearing and dehydration in an ethanol series. Samples were rehydrated, mounted in 50% (v/v) glycerol, and imaged with a Zeiss Axioplan 2 microscope using a 40× objective. Micrographs were taken with an Axiocam HRc camera and Axiovision V 4.8.2 Software (Carl Zeiss Light Microscopy, Göttingen, Germany). Images were combined into panoramas covering the section from bark to pith using the program Adobe Photoshop CS6.

Growth parameters (plant height and root collar diameter) were assessed at the end of each growth season. Final tree height and diameter were also measured in June 2018, before harvesting. The height was assessed with a measuring stick and the stem diameter with a caliper (3 cm above ground level). Stem volume was calculated as: V = 1/3πR²H, where R – stem radius, H – stem height.

Standard assessments (Nilsson and Örlander, 1999) of damage were conducted four times a year; they included cause of damage (fungi, frost, drought, waterlogging, rodents, herbivores, insects, vegetation, unknown) and severity on a six-level scale (0 = undamaged, 1 = slight damage, 2 = uncertain or moderate damage, 3 = severe damage, 4 = life-threatening damage, and 5 = dead).

A detailed assessment of leaf damage was performed in July 2017 during the fourth growing season, according to previous methodology (Albrectsen et al., 2010; Robinson et al., 2012). Chewing damage by arthropods (chewing) was assessed as percentage of chewed leaves in the canopy. Evidence of other types of damage was scored in terms of presence (1) or absence (0). This was done for damage caused by arthropods including aphids, miners, gall-producing organisms, and pathogens including rust (Melampsora spp.) and venturia (Venturia spp.). Symptoms of chlorosis, necrosis, and hypersensitive response (HR) were recorded in the same way.

Architectural traits, chlorophyll index, and leaf CT contents were assessed for a subsample consisting of 50% of trees having superior height selected from each transgenic line and wild type. The rationale for stratifying the samples by height rather than randomly selecting 50% of trees from each line was to avoid those trees that had been damaged by planting, field work or other types of random disturbance.

All analyses were performed using the software package JMP 14.0.0 2018 (SAS Institute Inc.). The consequences of decreased acetylation for field performance were analyzed for a total of 19 growth related traits, 14 biotic stress related traits, and 23 foliar defense related chemicals (mainly phenylpropanoids). Effects of individual lines were tested by a one-way ANOVA with “line” used as fixed effect (Supplementary Tables S1, S2). Similarity of individual lines to WT was evaluated by a Dunnett’s test, and consistent line effect within a construct were assessed by a contrast analysis (all lines for a given construct versus WT). These results were used to identify cases of reproducible construct effects among different transgenic lines.

A nested-ANOVA model design was used to answer questions about the impact of construct on phenotypic trait expression. “Line” nested in “construct” and block (random, considered when possible) effects were included (Supplementary Tables S3, S4). A multiple comparison Tukey test was used for cross-comparisons among different constructs.

Impacts of the promoter and the transgene (fixed effects) were analyzed by a two-way ANOVA with an interaction (Supplementary Table S6), and comparison of deacetylation strategy was carried out by a nested ANOVA model with “construct” nested in “pre- or post-synthetic strategy” and “line” nested within “construct” (Supplementary Table S7), all used as fixed effects.

Uniform growth and a stable survival rate of close to 100% (86–100%) characterized the majority of the transgenic lines throughout the period of testing at the field site (Figure 1A and Supplementary Table S1). However, three out of 20 transgenic lines for three different constructs (namely line 22 for construct 35S:RWA-CD, line 15 for construct WP:RWA-ABCD, and line 11 for construct WP:HjAXE) exhibited distinct deviations from the general growth pattern. These lines were dwarf, reaching only 1, 26, and 2% of WT stem volume, respectively, and their apical dominance was reduced compared to that of WT and other lines with the same construct (Figures 1B,C and Supplementary Table S1). Line 11 of WP:HjAXE had significantly higher mortality than the other transgenic lines and WT (Figure 1D), and approx. 50% of trees of this line showed a striking variegated phenotype that suggested genomic instability (Figures 1D,E). None of these phenotypes were seen in greenhouse trials (Ratke et al., 2015; Pawar et al., 2017a, b; Wang et al., 2020). The three dwarf lines were also more affected by hare browsing and multiple injuries than the other lines or WT trees (Supplementary Table S1).

Since other lines with the WP:HjAXE and 35S:RWA-CD constructs did not exhibit dwarf phenotypes, and moreover since line 11 of WT:HjAXE had lower transgene expression levels than the other lines carrying this construct (Wang et al., 2020), we do not consider these anomalies to have been caused by the respective transgenes. Rather these phenotypes should be attributed to mutations induced either by random transgene insertions in the genome or by somaclonal variation. However, the WP:RWA-ABCD construct was represented by only two lines that had very different phenotypes (Figures 1B–D and Supplementary Table S1). It was therefore not possible to infer any true construct effect from the data and both these lines were subsequently omitted from analyses testing construct effects. Thus, the subsequent analyses evaluate the effects of five constructs: 35S:RWA-CD, 35S:AXE1, 35S:HjAXE, WP:AnAXE1, and WP:HjAXE, with the anomalous lines removed from analyses.

Stem growth parameters (height and diameter) for the different constructs are shown in Figure 2 and Supplementary Tables S2, S3. Trees expressing WP:AXE1 and WP:HjAXE were slightly bigger than WT in the first year, but this advantage disappeared during the subsequent years. In contrast, those with two other constructs, 35S:AnAXE1 and 35S:HjAXE, showed a small reduction in height and /or stem diameter after three and four years of growth (Figures 2A,B), resulting in a decrease in stem volume in the fourth year of 24 and 37%, respectively, as compared to WT (Figure 2C).

Shoot apical dominance and the branching pattern did not show any construct-related effects (Supplementary Tables S2, S3). Leaf dry weight and chlorophyll content were reduced in 35S:HjAXE expressing trees as compared to WT, but their leaf morphology did not change (Figure 3). Interestingly, the chlorophyll content was slightly increased in 35S:RWA-CD (Figure 3); this was not observed in any of the WP:RWA-ABCD lines (Supplementary Table S1). Unexpectedly, a small decrease in chlorophyll content was observed in WP:GUS lines.

To assess the suitability of the 35S and WP promoters and the two AXE transgenes, AnAXE1 and HjAXE for transgenic expression, we analyzed the growth traits of a subset of transgenic lines (35S:AXE1, 35S:HjAXE, WP:AXE1, WP:HjAXE) by a two-way ANOVA with promoter and transgene as fixed effects (Supplementary Table S6). This comparison revealed a positive effect of WP, compared to the 35S promoter, on stem height, diameter, volume, and chlorophyll index (Figures 2A–C, 3B and Supplementary Table S6). In contrast, there was hardly any effect of the transgene (AnAXE1 vs. HjAXE) on growth, whereas the chlorophyll index was decreased in HjAXE compared with AnAXE1 but only when combined with the 35S promoter (Figure 3B and Supplementary Table S6).

Specificity of WP activity has been previously tested in greenhouse conditions (Ratke et al., 2015). We therefore investigated whether its activity and expression pattern are maintained in field conditions, using histochemical β-glucuronidase analysis of two WP:GUS lines during the fourth growing season. The test was carried out on branches with 1- and 2-year old cambia. In both samples and both lines, β-glucuronidase expression was detected in differentiating secondary xylem cells and secondary phloem fibers and sclereids depositing secondary walls (Figure 4 and Supplementary Figure S3). This expression pattern was consistent with that previously observed in the same lines in the greenhouse (Ratke et al., 2015). We conclude that the activity and specificity of WP was not altered in the field.

We also investigated whether growth was differentially affected in plants in which xylan acetylation was modified during biosynthesis in the Golgi compared to plants in which post-synthetic xylan deacetylation was implemented (denoted post- vs. synthetic comparison). To answer this question, the lines with RWA-CD RNAi suppression driven by the 35S promoter were compared with those where fungal AXEs were driven by 35S using ANOVA (Supplementary Table S7). For these constructs, growth, assessed by stem height, diameter and volume, and leaf dry weight, was scarcely affected by the engineering strategy used; only the final height and diameter (measured in the middle of the fifth growing season) were slightly reduced in lines with the post-synthetic deacetylation strategy compared to those with synthetic reduction (Supplementary Table S7). This growth inhibition was preceded by a decrease in leaf dry weight and chlorophyll contents in the fourth year in lines with post-synthetic reduction (35S:AnAXE1 and 35S:HjAXE) (Figure 3).

Stress-related responses were recorded regularly throughout the four-year field test period, and detailed mapping of necrosis, rust and chewing symptoms was additionally conducted in 2017 simultaneously with collection of leaf material for phenolic profiling. Some of these traits were related to construct identity and, in particular, to the choice of promoter (Figure 5 and Supplementary Tables S2, S3, S5–S7). Necrosis was generally elevated for 35S:HjAXE plants, potentially at the expense of rust symptoms (Figure 5A), which were generally reduced on plants belonging to the same construct. Moreover, both kinds of fungal AXE transgenes under the 35S promoter (Figure 5B) suffered from an increase in the extent of chewing symptoms, with damage being increased by 187% (35S:AnAXE1) and 44% (35S:HjAXE) compared to wild type plants. Synthetic acetylation reduction (35S:RWA-CD), on the other hand, had no impact on chewing damage (Figure 5B).

Levels of phenolic compounds that are often associated with damage risk varied in a genotype-related way. For example, salicortin and HCH-acetyl-salicortin were elevated in 35S:RWA-CD trees (Table 2). 35S:AnAXE1 and 35S:HjAXE lines also expressed higher levels of certain SPGs, such as OH-tremuloidin, whereas acetyl-tremulacin and p-coumaric acid were greatly elevated in 35SAnAXE1. Salicylic acid, on the other hand, was reduced in 35S:HjAXE trees. Thus, use of the constitutive 35S promoter led to more alterations in concentrations of SPGs compared with transgenics created with the help of the WP promoter, which did not deviate significantly from the wildtype (WT, Table 2).

Catechin (the precursor of CTs) was elevated in 35S:AnAXE1, which also had higher levels of CTs (Figure 5C). The elevated CT level characteristic of genotype 35S:AnAXE1 was accompanied by very much higher levels of damage caused by chewing herbivores, whereas a general decrease in CT levels for 35S:HjAXE was also associated with elevated herbivory (by 44%). That the two fungal 35S transformations were associated with varying responses in terms of CTs, but at the same time had similar susceptibility to herbivores, indicated a lack of any general relationship between CT levels and the risk of chewing damage.

In addition to SPGs, we monitored the foliar levels of arachidonic acid, a polyunsaturated fatty acid (20:4 Δ5,8,11,14) regulating different physiological and stress responses that is found in some plant species including poplars (Groenewald and van der Westhuizen, 1997). It has been shown to trigger different plant stress responses and induce resistance to fungal pathogens (Savchenko et al., 2010). Interestingly, the levels of arachidonic acid were decreased compared to WT in lines with all constructs except for those containing AnAXE1 (Table 2). This suggests that the quality of lines carrying AnAXE1 constructs is potentially superior compared to that of other transgenic lines with regard to biotic stress resistance.

We have tested for the first time the field performance of trees with transgenically reduced cell wall acetyl content. We found that, in general, apart from those lines with abnormal phenotypes, the reduction of acetylation in cell walls did not lead to either major detrimental effects or significant growth stimulation in the field (Figures 2, 3). The level of reduction in acetyl content for the lines previously tested varied between 10 and 16% in WP:HjAXE (lines 11, 14B, 14C; Wang et al., 2020), and 13 and 16% in 35S:AnAXE1 (lines 4, 8, 17; Pawar et al., 2017a), and it was 20% in 35S:RWA-CD (line 10; Pawar et al., 2017b). The reductions in acetyl content in WP:AnAXE1 lines have not been analyzed, but based on previous comparisons between 35S and WP-driven transgenes (Ratke et al., 2015), and the documented observation that the specific activity of the WP is stable (Figure 4), we expect a slightly stronger effect with WP:AnAXE1 than with 35S:AnAXE1. Thus, the mild reductions in cell wall acetylation levels (by 30% or less) are well supported by plants both, in the field (Figure 2), and in the greenhouse conditions (discussed in Pawar et al., 2017b; Donev et al., 2018).

Analyses of growth and biotic stress responses in lines harboring 35S- and WP-driven fungal AXEs point to a clear advantage offered by the specific transgene expression achieved with the WP. Both height and diameter growth were reduced in 35S lines compared to WP lines, and the stem volume was reduced by as much as 21% and 41% in, respectively, 35S:AnAXE1 and 35S:HjAXE compared to the corresponding WP constructs (Figure 2). These growth penalties in 35S lines were seen only in the field; these lines did not exhibit growth defects in the greenhouse (Ratke et al., 2015; Pawar et al., 2017b; Wang et al., 2020). One factor which could have contributed to the growth penalty in the field is the alteration in interaction with herbivores as revealed by the more extensive foliar chewing damage (Figure 5B). The increase in susceptibility to herbivores is likely to be due to metabolic changes in the leaves in 35S plants caused by AXE activity, changes which are largely avoided when WP targets the transgene expression to the developing wood. The fungal AXEs targeted to cell walls are expected to hydrolyze acetyl esters liberating acetic acid, thus changing the pH of the leaf, and since acetic acid can cross membranes in uncharged form, the reaction could contribute to the biosynthesis of acetyl-CoA. This compound is used in a variety of reactions, including the TCA cycle, glyoxylate cycle, lipid biosynthesis, mevalonate pathway, and it is considered an energy-status marker for a eukaryotic cell (Cai and Tu, 2011). It is therefore perhaps not surprising that ectopic manipulation of acetyl-CoA pools can have far-reaching consequences, and our results demonstrate that restricting the transgenic modification to specific tissues, such as developing wood, can prevent or attenuate such undesirable side-effects.

Assuming that the RWA genes encode Golgi-localized acetyl-CoA transporters (Gille and Pauly, 2012; Pauly and Ramírez, 2018), cytosolic accumulation of acetyl-CoA is expected when expression of these genes is suppressed. Apoplastic AXE expression which results in high acetic acid levels in the apoplast might eventually result in a similar outcome, assuming that acetic acid diffuses via membranes and is converted to acetyl-CoA by acetyl-CoA synthases. Even though the cytosolic accumulation of acetyl-CoA might be similar with the synthetic and the post-synthetic strategy, these two strategies might lead to differences in cell wall polymer structures. For example, increased glucuronosylation of xylan is expected when the xylan acetylation machinery is suppressed since the two processes apparently compete for the same substrate (Chong et al., 2014; Xiong et al., 2015; Grantham et al., 2017). Post-synthetic deacetylation might be also more specific toward a targeted polymer – xylan – in particular, HjAXE, which was used in this study for the post-synthetic modification, has documented xylan specificity (Koutaniemi et al., 2013).

Comparisons of the phenotypic effects of synthetic versus post-synthetic strategies (both using the 35S promoter) revealed that some leaf-related traits were affected. Leaf weight, chlorophyll content, and chewing resistance appeared to be lower in the case of post-synthetic modification (Figures 3, 5). Many SPGs were also affected by the deacetylation strategy (Supplementary Table S7). These foliar changes were not matched by stem growth during the first four years in the field, but in the final (fifth) year, both stem diameter and stem height were somewhat reduced by the post-synthetic modification strategy compared to the synthetic one (Supplementary Table S7). The lines available only allowed us to draw conclusions about the pre- vs. synthetic strategy in the case of ectopic modification using the 35S promoter. It would be interesting to investigate whether the same conclusion applies to modification targeted specifically to developing wood.

Although the trees in this study were never exposed to an outbreak of severe attack by a particular herbivore or pathogen, the relatively low and variable relationships in the field between various kinds of biotic stressors and genotypes confirmed that the transgenic procedure in itself is unlikely to be associated with any systematic impact on surrounding organisms and vice versa (Strauss, 2003).

Leaf CTs of natural aspen populations are strongly tied to genotype (Lindroth et al., 2011; Robinson et al., 2012; Bandau et al., 2015; Decker et al., 2016), a feature that was also observed in our transformed genotypes. CTs are considered to be anti-oxidant phenolic polymers (Gourlay and Constable, 2019) that are expected to influence the presence and impact of plant-consuming microorganisms and herbivores (Mutikainen et al., 2000; Bailey et al., 2005; Barbehenn and Constabel, 2011), although they also express a high degree of plasticity in response to environmental factors such as nitrogen addition (Bandau et al., 2015) and they may be equally important and indicative of the extent of internal recovery and the mode of growth (Harding et al., 2013; Lindroth and Madritch, 2015; Decker et al., 2016). However, the two lines in this experiment that suffered from elevated chewing symptoms varied in tannin content, with 35S:HjAXE giving lower and 35SAnAXE1 higher foliar CT concentrations when compared to WT (Figures 5B,C). No consistent relationship between CTs and chewing damage caused could therefore be deduced from this study, supporting the hypothesis that the potential defensive role of CTs in plants is indeed complex.

Salicinoid phenolic glucosides have often been investigated as suggested markers of innate resistance to herbivore damage in woody species (Mutikainen et al., 2000; Albrectsen et al., 2004; Philippe and Bohlmann, 2007; Witzell and Martin, 2008; Fabisch et al., 2019) and in particular as constitutive markers in aspen (Albrectsen et al., 2010; Robinson et al., 2012; Bernhardsson et al., 2013; Lindroth and St. Clair, 2013; Lindroth et al., 2015). The 35S-driven transgenes appeared to be more affected with respect to their SPG profiles compared to the WP-driven transgenes, although the low number of lines per construct tested in the present experiment (between two and four) resulted in few significant changes in SPG contents among the constructs (Table 2 and Supplementary Tables S2, S4). Despite this deficiency, the greater impact of the 35S promoter compared to the WP promoter on SPG profiles was obvious. The synthesis of phenolic compounds belonging to the SPG group is still unresolved due to reticulate pathways with no apparent direct connection to the most simple salicinoid, salicin (Babst et al., 2010; Fellenberg et al., 2020), although it is increasingly accepted that the specialized metabolism of phenolic compounds is tightly linked to primary metabolism (Harding et al., 2013), and our study further suggests that cell wall acetylation may indeed alter, and determine levels of, phenolic compounds in aspen.

The selection of enzymes appropriate for transgenic modification was addressed in this study by comparing two fungal enzymes, a CE1 representative, AnAXE1 and a CE5 representative, HjAXE, expressed from either 35S or WP promoters, for their effects on several traits related to growth, biotic stress resistance and foliage characteristics. Stem growth and leaf weight were not affected by the enzyme used. In contrast, the occurrence of necrosis was associated with HjAXE, and there was a higher incidence of chewing with AnAXE1, although both transgenes induced more chewing than was seen in WT (Figure 5 and Supplementary Table S6). CTs were more characteristic of HjAXE than AnAXE1 expressing plants, and several SPGs accumulated differentially in the leaves of transgenic plants with the two transgenes (Figure 5 and Table 2). These differences were primarily seen in the lines with 35S-driven transgenes; effects were negligible in lines with the WP promoter. These data indicate that each transgene induced different susceptibilities to specific biotic stresses, associated with different patterns of accumulation of some stress-related SPGs and CTs. Previous greenhouse studies with 35S:AnAXE1 and 35S:HjAXE expressing plants did not reveal any major morphological differences between plants with the two transgenes (Ratke et al., 2015; Pawar et al., 2017a), highlighting the importance of field testing.

The physiological background behind the contrasting phenotypes observed in 35SAnAXE1 and 35SHjAXE expressing plants is not known, and it could encompass many factors. Beside the difference in enzymatic specificities and mode of action in the cell wall (discussed by Wang et al., 2020), the two proteins could be differentially perceived by the pathogen-associated molecular patterns (PAMP) recognition system (Bellincampi et al., 2014). The induction of necrosis by ectopically expressed HjAXE seen in our trial (Figure 5) is reminiscent of the effects of several fungal xylanases from family 11, including Hypocrea jecorina xylanase II, which induce ethylene and hypersensitive responses in plants, leading to necrosis (Noda et al., 2010). A conserved amino acid motif, TEIGSVTSDGS, has been identified as being involved in induction of necrosis. Amino acid alignments of sequences used in the two constructs reveal that HjAXE, but not AnAXE1, includes a similar motif, 183-VGTCTTQG-190, and it would be interesting to test this for necrosis-inducing activity.

In this trial, three out of 18 transgenic lines with reduced acetylation exhibited dwarfism (Figure 1B) and growth abnormalities (Figure 1C), and one of them showed increased mortality (Figure 1D). These detrimental effects could not be associated with the transgenes introduced, and they are likely to have been caused by somaclonal variation or positional effects. Such a high (17%) incidence of dwarfism in acetylation-modified lines is, however, remarkable, and it was not predicted on the basis of the growth observed during greenhouse trials with the same transgenic lines. We also observed no dwarfism other than dwarfism related to transgenes among another set of 48 transgenic lines modified for other qualities that were grown in nearby transgenic fields. Reports from previous American field trials support the conclusion that somaclonal variations or detrimental positional effects are rare in transgenic poplars. For example, in a long-term trial with 948 lines engineered for sterility, not a single incident of detectable somaclonal variation was reported (Klocko et al., 2018); similarly, in a survey of field trial studies in United States covering a period of over 20 years and more than 100 transgenic poplar lines, only 0.1–1% dwarfism that could putatively be linked to positional effects or somaclonal variation was detected (Strauss et al., 2016).

The exceptionally high occurrence of random dwarfism and abnormalities among our transgenic lines with reduced acetyl content suggests a potential link between the acetylation status and genomic stability. Indeed, studies in other eucaryotes including mammals and yeasts showed that cellular levels of acetyl-CoA are directly associated with histone acetylation, which in turn regulates chromatin epigenetic state (Cai and Tu, 2011; Etchegaray and Mostoslavsky, 2016). In plants, epigenetic changes in chromatin state have been linked to the activation of transposable elements under stress conditions, thus contributing to somaclonal variation (Kaeppler et al., 2000). Moreover, in mammalian cells, the cellular ability to repair double strand breaks in DNA requires histone acetylation (Sivanand et al., 2017). These data support the hypothesis that the higher levels of acetyl-CoA expected to be induced by our engineering strategies could indeed lead to increased rates of mutation, especially when combined with stress. This hypothesis could be addressed by field testing and whole-genome sequencing of different acetyl-modified plants.