Antibiotic treatment impaired growth of tobacco in vitro shoots is associated with oxidative injury and reduced microbiome diversity

Plant in vitro cultures initiated from surface-sterilized explants often harbor complex microbial community. Antibiotics are commonly used to decontaminate plant tissue culture or during genetic transformation, however, the effect of antibiotic treatment on the diversity of indigenous microbial population and consequences for performance of the tissue culture are not completely understood. Therefore, the aim of this study was to assess the effect of antibiotic treatment on the growth and stress level of tobacco (Nicotiana tabacum L.) in vitro shoots as well as the composition of plant-associated microbiome. The study revealed that shoot cultivation on medium supplemented with 250 mg L −1 timentin resulted in 29 ± 4% reduced biomass accumulation and 1.2–1.6 fold higher level of oxidative stress injury compared to control. Moreover, the growth properties of shoots were only partially restored after transfer to medium without antibiotic. Shoot microbiome analysis using multi-variable region-based 16S rRNA sequencing revealed abundant microbial community in the control tobacco shoots, including bacteria from 59 families, however, it was largely dominated by Mycobacteriaceae. The antibiotic treatment resulted in a decline of species richness (the number of families was reduced 4.5-fold) and increased domination by the Mycobacteriaceae family. The results imply that the diversity of plant-associated microbiome might represent a signicant factor contributing to the ecient propagation of in vitro tissue culture. tobacco shoot in vitro culture and composition of plant-associated microbiome. Changes of shoot biomass accumulation and oxidative stress injury of cellular membranes during the antibiotic treatment and residual effect following the treatment were assessed. The composition of shoot bacterial population before and after the antibiotic treatment was investigated using a multi-variable 16S rRNA region-based metagenomic analysis. To further rene taxonomic assignment of the dominant taxa that was revealed by the metagenomic analysis, isolate of Mycobacterium sp. was obtained using cultivation technique and its antibiotic resistance was tested.


Introduction
Endophytes are a class of endosymbiotic microorganisms that inhabit the internal plant tissues 1 . Being protected inside explant or seed tissues, endophytic bacteria evade surface sterilization procedures used for cell culture initiation, and they are common in plant tissues grown in vitro [2][3][4] . Some bacterial endophytes maintain habitual concealed lifestyle or due to limited bacterial growth-supporting media conditions remain latent over extended periods of in vitro tissue cultivation. In fact, several studies showed a bene cial effect of endophytic bacteria on the growth of in vitro cultures of tomato 5 , grapevine 6 , sweet cherry 7 , apple 8 , purple cone ower 9 or tobacco 10 . However, the formation of bacterial colonies on culture medium and bacterial overgrowth of plant tissues are rather common and manifest as infection by a variety of pathogenic species or the non-fastidious proliferation of commensal endophytic bacteria that can be triggered by the changes in environmental conditions or plant host physiology [11][12][13] .
To eliminate contamination of endophytic origin, pretreatment with high dose 14 or medium supplementation with microbial growth-inhibiting 15 antibiotics have been used and is often reported to improve the propagation or regenerative properties of the tissues 4, 16-18 . Another common application of antibiotics for in vitro tissue culture is related to the use of Agrobacterium tumefaciens bacterium-to-plant DNA transfer machinery as an instrument in the plant genetic transformation 19 . Although alternative methods for the selection of genetically modi ed plants have been developed 20 , antibiotics remain indispensable for control of Agrobacterium growth upon plant tissue culture transformation. Left untreated, Agrobacterium, the causal agent of crown gall disease, shows excessive cell proliferation, that elicits a defense response in plant cells leading to a detrimental effect on plant tissue growth and e ciency of transformation [21][22][23] . Cephalosporin and penicillin-type antibiotics, such as cefotaxime, carbenicillin or timentin, active against gram-negative bacteria are commonly used [24][25][26] . In culture medium, carbenicillin and penicillins are broken down to physiologically active levels of the auxin-related compounds promoting organogenesis of explants 27 , however, at higher antibiotic concentration, proline accumulation, oxidative injury and reduced antioxidative activity have been reported indicating elevated stress level 28 . Previously, low cytotoxic effect of antibiotic timentin on in vitro plant tissues has been described 4,24 , and it is commonly used for transformation procedures in the range of concentrations from 150 mg L −1 24 to 500 mg L − 1 29 , sometimes in combination with lower concentrations of other antibiotics 30 .
Cultivated tobacco (Nicotiana tabacum L.) has well established in vitro culture conditions 31 and it has been commonly used as model species for genetic transformation studies as well as for practical application in molecular farming using in vitro techniques or contained greenhouse practice [32][33][34] .
Recently, the composition of the endophytic microbiome of eld-grown tobacco plants or seeds has been explored using cultivation and next-generation sequencing-based methods 35,36 and several endophytic species of the Bacillus genus were isolated from tobacco leaves 10,37 . However, the microbial diversity of in vitro cultivated tobacco tissues had not been addressed so far.
Whereas antibiotics nd application for explant and tissue culture decontamination or control of bacterial overgrowth during plant transformation, the effect of antibiotic treatment on the composition of endogenous microbial population in tissue culture and the long-term consequences on plant response to stress and growth parameters have not been completely understood. Considering the established role of endophytic bacteria, as modulator of plant growth and adaptation, it could be expected that the antibiotic-induced perturbation of the endophytic community might affect the e ciency of the in vitro culture propagation. Therefore, the aim of this study was to assess the effect of antibiotic timentin on growth, stress level of tobacco shoot in vitro culture and composition of plant-associated microbiome.
Changes of shoot biomass accumulation and oxidative stress injury of cellular membranes during the antibiotic treatment and residual effect following the treatment were assessed. The composition of shoot bacterial population before and after the antibiotic treatment was investigated using a multi-variable 16S rRNA region-based metagenomic analysis. To further re ne taxonomic assignment of the dominant taxa that was revealed by the metagenomic analysis, isolate of Mycobacterium sp. was obtained using cultivation technique and its antibiotic resistance was tested.

Results
Effect of antibiotic on tobacco shoot growth and oxidative stress injury. Antibiotics could be cytotoxic to plant tissues at concentrations required for e cient control of the bacterium, usually in the range of 100 to 500 mg l −1 38 . Although low cytotoxic effect of antibiotic timentin on in vitro plant tissues had been documented 4, 24, 29 and it had been commonly used for plant cell culture at concentrations up to 500 mg L − 1 24, 29 , our experiments showed suppressed growth of tobacco in vitro shoots on medium supplemented with 250 mg L −1 timentin. After 3 weeks of cultivation, an average fresh weight (FW) of the control tobacco shoots (TC) was estimated at 164.7 ± 3.3 mg, and FW of the shoots cultivated on medium containing timentin (TA) was reduced by 29 ± 4% (p = 1.5⋅10 −7 ) compared to TC.
An analysis of the membrane lipid oxidative injury of the TC shoots revealed a signi cant variation of the MDA concentration during the three weeks of the propagation cycle (Fig. 1). The highest value was detected at day 4 after transfer to fresh medium and was estimated at 95.6 ± 2.6 nmol·g −1 FW. The increased MDA accumulation in the control shoots was likely a consequence of the combined effect of stress associated with tissue senescence and shoot injury during the transfer to the fresh medium which was followed by adaptation and active growth on the fresh medium resulting in ~1.3 fold (p = 4⋅10 −6 ) decrease in MDA concentration at the end of the rst week of cultivation (day 7), and subsequent gradual increase in MDA concentration during the course of culture senescence over the remaining two weeks of the propagation cycle. However, the TA shoots maintained on medium with antibiotic did not follow this cycle and showed 1.2-1.6 fold (p < 0,014) higher levels of MDA compared to the TC over the entire period of propagation cycle ( Fig. 1) that could be a direct consequence of antibiotic-induced cytotoxic effect.
Interestingly, after transfer to the medium without antibiotic (PA), a consistent residual negative effect of antibiotic treatment on shoot growth vigor and stress level was observed for at least several passages used in the experiments. Cumulative data from several of these experiments revealed that growth vigor of the shoots was only partially restored and accumulation of biomass remained signi cantly lower (9 ± 3%, p = 0.013) compared to TC (Supporting Material Fig. S1). Similarly, the level of MDA accumulation was reduced as compared to TA and showed a signi cant decrease during the rst week after transfer to fresh medium, nevertheless, it remained 1.1-1.3 (p < 0.033) fold higher compared to the control over the remaining two weeks of the propagation cycle.
Analysis of bacterial diversity in tobacco in vitro shoot samples. To investigate bacterial diversity in tobacco shoot culture and to assess the effect of antibiotic treatment on the bacterial diversity, DNA extracted from shoots of the TC and PA experimental groups was subjected to metagenomic analysis using the Ion Torrent sequencing platform. Six different DNA extraction and enrichment methods were used for DNA preparation to assess capability to improve the e ciency of the bacterial 16S rRNA metagenomic analysis using as a template the DNA samples derived from the plastid and mitochondrial DNA rich plant material. The PCR amplicon libraries were generated using a multi-variable region approach comprising six regions of the 16S rRNA represented by a combination of ve primer pairs (V2, V3, V4, V6-7 and V8) included in the 16S Metagenomic kit (Thermo-Fisher Scienti c) (Supporting Material Fig. S2).
The overall number of high-quality mapped sequences with 224-225 bp read length was similar for the TC and PA libraries (1,027,576 and 1,175,938, respectively), but varied up to ~3.5-fold for different DNA preparation methods and primer pairs speci c to distinct 16S rRNA gene regions (Supporting Material   Table S1 and Fig. S3). The proportion of bacterial sequences varied from 10 to 29% for the different DNA extraction methods (Supporting Material Fig. S3A). Independent of the experimental group, consistently higher content of bacterial OTUs (>25% of all reads) was detected for samples prepared with DNA extraction methods described by Doyle 39 and Ding et al. 40 . For the TC shoots, similar results were also obtained using DNA extraction method described by Li et al. 41 . The remaining two methods resulted in a consistently lower proportion of the bacterial OTUs. Bacterial DNA enrichment by selective organelle lysis using SDS and NaCl as previously described by Wang et al. 42 for preparation of microbial metagenomic libraries of tropic tree Mallotus nudi orus did not increase bacterial OTU content for the tobacco shoot samples as compared to other DNA extraction methods and the approach would require further optimization.
The proportion of sequences assigned to the bacterial and tobacco plastid or mitochondrial 16S rRNA also varied depending on PCR ampli cation region and/or speci city. Two primer pairs speci c to the V4 and V8 regions generated the largest number of reads mapped as bacterial OTUs and the largest proportion of the bacterial OTUs as compared to the overall number of reads (Supporting Material Fig.  S3B). Meanwhile, the overall ampli cation e ciency of the V6-7 and V2 primers was 2-to 4-fold lower, the resulting proportion of bacterial sequences was comparable to the V8. The V3 primers generated a large number of reads which was comparable to V8 but resulted in the lowest proportion of bacterial sequences. Previously, the V3 region had been shown to e ciently represent the diversity of microbial communities of the fecal or sewage samples [43][44][45] , however, our analysis using plant-derived samples showed and V3 primers included in the 16S Metagenomics Kit showed a strong predisposition toward plastid and/or mitochondrial sequences. It is notable that, for all primer sets, a consistently lower ratio of bacterial to plastid/mitochondrial OTUs was obtained for the PA as compared to the TC sample which could be a result of lower yields of bacterial DNA due to lower bacterial density in the antibiotic-treated shoot sample.
Principal coordinate analysis (PCoA) of the microbiome data sets using Bray-Curtis dissimilarity matrix represented variation mainly between the two tobacco shoot experimental groups on the rst coordinate and variation resulting from the application of multi-variable region 16S analysis were plotted mainly on the second coordinate (Fig. 2). Although larger variation was observed among the primer pairs for the PA sample, the overall pattern was very similar to that observed for the TC sample where V2 and V4 data sets represented the largest difference. The larger variation within the PA group was likely the result of a 5.6-fold lower number of OTUs mapped as compared to the TA sample (Supporting Material Table S1).
The region-speci c primer capability to represent bacterial taxonomic diversity was assessed using rarefaction and alpha diversity analysis. A number of reads representing bacterial OTUs of the V3 data set were insu cient for the rarefaction analysis. For the remaining four regions the saturated curves indicated that the bacterial diversity was su ciently represented and a several-fold difference in species richness between the two experimental groups, TC and TA, was revealed (Supporting Material Fig. S4).
Alpha diversity analysis revealed a similar distribution of the Richness measure and the related Chao1 and Fisher's diversity indices (Supporting Material Table S2 and Fig. S5) among the data sets generated using distinct primer pairs. A signi cant difference was detected for the Evenness measure due to the high value estimated for the V3 region which appeared mainly related to the under-representation of the Mycobacteriaceae family in the V3 data set (0.2 and 40% of bacterial OTUs for the TC and PA, respectively) as compared to other primer pairs where the family was largely dominant (79-90% and 94-98%, respectively) (Supporting Material Table S3; Fig. 4).
To evaluate 16S rRNA region-speci c primer propensity for speci c taxonomic groups, family-level data was used to avoid bias due to the limited accuracy of genus or species identi cation using short-read sequences which could lead to underestimation of the primer speci city. The association among the OTUs obtained using different primer pairs was mapped on the UpSet plot using cumulative data from both experimental groups (Fig. 3). Among the 59 family level OTUs, 22% were detected by all primer pairs and were represented by 96% or 70% of all bacterial reads when the dominant Mycobacteriaceae family was included or excluded from the analysis, respectively. No singleton families were detected using V2speci c primers and 26 (42% of all families) singleton families were detected by the remaining four primer pairs, however, these included only 0.4% of the total number of reads mapped to bacterial OTUs. Chitinophagaceae and Sinobacteraceae (880 and 560 reads, respectively) were the most abundant singleton families detected using V4-and V8-speci c primers, meanwhile, the remaining singletons included OTUs of relatively low abundance (<100 reads). The results indicate that the multi-variable 16S rRNA region-based metagenomic analysis approach is capable to provide higher taxonomic resolution of microbial diversity as compared to results obtained using individual regions, especially for the low abundance taxa.
Previously, a variation of informative power for the same 16S rRNA region-speci c primers was demonstrated in the study with human intestinal or environmental microbiota samples [45][46][47] . In our study, the read abundance for the same OTU also varied considerably among the primer pairs (Fig. 4). For example, the most abundant Mycobacteriaceae family was represented by 37%, 18% and 39% of the total number of reads using V4, V6-7 and V8 primers, meanwhile, only 6% and <1% of the total number of reads were assigned to the family for the V2 and V3 primer data sets, respectively. In another instance, the distribution of reads assigned to the Paenibacillaceae family among V3, V4 and V8 data sets was 15%, 38% and 44%, respectively, meanwhile, V2 and V6-7 represented only 3% and <1% of the total number of reads, respectively. Regardless of the apparent taxa-speci c ampli cation e ciency among the primer pairs, the hierarchical cluster analysis of the data did not reveal consistent distribution of taxonomic groups among the primer data sets (Fig. 4), possibly due to variation of primer speci city at lower taxonomic level or the bias introduced by variation of the taxa abundance. Nevertheless, the results imply that cumulative data obtained from several primer pairs with different taxa-speci c ampli cation e ciency has limited quantitative capacity to represent species abundance variation (species evenness) within a sample, however, this would not affect the quantitative estimate of inter-sample abundance variation for speci c OTU.
Antibiotic effect on bacterial diversity in tobacco shoot culture. Although both, the TA and PA, samples had a similar overall number of reads and reads mapped to bacterial OTUs, the number of unique OTUs was reduced from 153 to 27 upon the antibiotic treatment (Supporting Material Table S1). The antibiotic effect resulted in 42% of total variability of the microbiome data between the TA and PA experimental groups as represented by the rst coordinate of the PCoA (Fig. 3), and alpha diversity analysis showed a signi cant reduction in species richness and evenness upon the antibiotic treatment (Fig. 5).
Among the six phyla detected in the tobacco shoot microbiome, Actinobacteria was dominant in both experimental groups and was mainly represented by the order of Actinomycetales including 17 and 3 families for the TC and PA experimental groups, respectively (indicated in a blue color font in Fig. 4; Supporting Material Table S3). It is remarkable that Actinobacteria included the most prevalent family of Mycobacteriaceae, representing 81% and 98% of mapped bacterial reads for the TC and PA, respectively. Microbacteriaceae represent another antibiotic treatment enduring actinobacteria that became the second most abundant OTU with a relative abundance of 1.6% in the antibiotic-treated shoot sample, meanwhile, Propionibacteriaceae and Acidimicrobiaceae were detected only at the marginal level. For the latter OTU, sequence comparison showed 98% sequence similarity to Mycobacteriaceae which could also imply inaccuracy of the sequence assignment. A similar assumption could be drawn for the assignment of Thermolithobacteraceae family of the phylum Firmicutes. Meanwhile, Bacillaceae and Paenibacillaceae were the most abundant among the remaining four families of Firmicutes representing 3% of mapped bacterial reads in the control shoots. In addition, Staphylococcaceae and Streptococcaceae were detected at low abundance but by three primer pairs each. The abundance of all Firmicutes was largely reduced (<0.1% of mapped reads) upon the antibiotic treatment.
In the control shoots, 1.6% of mapped reads were assigned to 5 families of Bacteriodetes, mainly represented by Sphingobacteriaceae, Chitinophagaceae and Flavobacteriaceae, but all of them were undetectable in the antibiotic-treated sample. Among the Proteobacteria including 8% and 0.5% of mapped bacterial reads of the TC and PA, respectively, class Alpha-proteobacteria included 10 families of order Rhizobiales representing many well-known bene cial plant-associated bacteria 48 , and also Caulobacteraceae and Sphingomonadaceae families including common environmental bacteria 49,50 . In the antibiotic-treated shoots only Caulobacteraceae and Bradyrhizobiaceae were detected at 4-and 10fold reduced abundance level compared to control shoots, respectively. Interestingly, families Kopriimonadaceae, Rhodobacteraceae and Rhodospirillaceae including species of common environmental and aquatic bacteria were detected at low abundance only in the antibiotic-treated shoots.
The Desulfovibrionaceae family of class Deltaproteobacteria was detected at low but similar abundance in both, TC and PA, shoots. However, more proliferous Beta-proteobacteria and Gamma-proteobacteria (including 5 and 7 families, respectively) were also largely reduced upon the antibiotic treatment. Among the Beta-proteobacteria were notable Alcaligenaceae and Methylophilaceae which include nitrifying and methylotrophic environmental bacteria 51, 52 , Meanwhile families Enterobacteriaceae and Pseudomonadaceae of Gamma-proteobacteria represent numerous plant endophytes or pathogens 53 . In addition, two families of the PVC superphylum, Planctomycetes and Verrucomicrobia, which includes common soil and plant root-associated bacteria 54 were both detected at low abundance in the control shoots.
Isolation of Actinobacteria and antibiotic resistance assessment. To further characterize the actinobacteria dominant in the in vitro tobacco shoot culture, the extract of antibiotic-treated shoots was plated on the Loewenstein-Jensen growth medium. The isolate obtained from bright yellow colonies forming after 6-8 weeks of incubation showed 98% identity to the M. cookii strain ATCC 49103 (GenBank accession NR_114661.1) 55 based on the 1407 nt fragment of 16S rRNA. The isolate also sustained similar growth properties on the Actinobacteria Isolation Agar which was used for later cultivation and antibiotic resistance test. The isolate showed resistance to timentin at the concentration used for the tobacco shoot treatment (250 mg·L −1 ) and chloramphenicol at 30 mg L −1 (Supporting Material Fig. S6), however, the growth of the isolate was suppressed by rifampicin at 25 mg L −1 which could potentially be used for the mycobacterium elimination from the plant in vitro culture.

Discussion
An abundant microbial community of the control tobacco shoots was revealed by metagenomic analysis using a multi-variable 16S rRNA region approach and 153 OTUs representing 59 families of the bacteria domain were detected (Fig. 4). Bacterial endophytes are common in the in vitro cultures initiated from surface-sterilized plant dormant tissues, such as seeds or buds, which contain a complex community of endophytic bacteria. Previously, the study by Thomas et al. 56 revealed a vast diversity of endophytic bacteria prevailing in grapevine eld shoots (mainly Proteobacteria but also Actinobacteria, Firmicutes and several less abundant phyla) and their introduction to in vitro shoot culture was demonstrated. The occurrence of endophytic bacteria in seeds is also well documented 57,58 . Since the tobacco in vitro culture used in our study was initiated from surface-sterilized seeds, it is possible that numerous bacterial species survived the sterilization procedure in internal tissues or even on the surface of seeds. Endophytic species vertically transmitted through seed are closely associated with plant host and often play a signi cant role in the regulation of plant growth and development physiology. Previously, Enterobacteriaceae was detected as the predominant species of the endophyte community of tobacco seeds of the four distinct tobacco cultivars, and a genotype-speci c signature was observed for Alphaproteobacteria 36 . In our study, among the bacteria identi ed in shoot microbiome, numerous species of Bacillales, Rhizobiales, Burkholderiales or Enterobacteriales could include plant endophytic bacteria adapted for vertical transmission and capable to inhabit seed tissues.
Alternatively, endospore or bio lm-forming bacteria are adapted to survive in extreme environments, many show resilience to sterilizing agents and would endure seed sterilization procedures. Such route could be adequate for soil bacteria such as Mycobacterium spp. which forms resilient endospores 59 , are capable to colonize plant tissues but have not been detected in seeds so far 60, 61 . Similarly, species of Sphingomonadaceae are widely distributed in nature, having been isolated from many different land and water habitats, as well as from plant root systems and could form resilient bio lms which often are capable to survive sterilization conditions 62 .
Despite the multi-variable region metagenomic analysis used in our study, it had limited capacity to re ect intra-sample abundance differences among the taxa. It appeared that the control tobacco shoot microbiome was largely dominated by Mycobacteriaceae, and likely one of its dominant species was shown to be closely related to Mycobacterium cookii. Mycobacteria are widely distributed in water and soil 63 , however, despite the fact that the genus Mycobacterium comprises nearly 200 species, so far the research has mainly focused on obligate pathogens and information about plant-associated mycobacteria is relatively scarce. Mycobacterium spp. have been isolated from the rhizosphere of tomato 64 and rice plants 65 . They have also been detected by sequencing-based analysis as endophytes of rice roots and stems 61 , wheat roots 66 , shoots of rock plant 67 and buds of Scots pine 68 . Although previously, Mycobacterium spp. have not been described as tobacco endophyte, it might not represent an exception and might vary depending on cultivation conditions. In particular, endophytic or epiphytic colonization would be plausible for the plants cultivated on peat substrate considering that mycobacteria are abundant in peat as they are common inhabitants of sphagnum vegetation 69 , which was also one of the rst described habitats for M. cookii 70 .
It is notable that previously Mycobacterium spp. have been detected as dominant species in sweet cherry in vitro shoots 71 and tissue cultures of Scots pine 72 . Although in the latter case, some species of Mycobacterium have been reported to suppress undifferentiated Scots pine tissue growth and reduce the length of hypocotyls of seedlings, no negative effect has been observed for other species 72 . However, plant growth regulating properties have not been described for species related to M. cookii so far.
In our study, antibiotic timentin treatment had a detrimental effect on the endophytic community of tobacco in vitro shoots. The number of detected families was reduced ~4.5-fold and most of the remaining families had largely reduced counts. Interestingly, antibiotic treatment had little effect on the abundance of the dominant Mycobacteriaceae and Microbacteriaceae families (Supporting Material  Table S3) as well as two Alpha-proteobacteria families, Caulobacteraceae and Bradyrhizobiaceae, that showed a modest reduction in relative abundance. Recently, the antibiotic effect on the diversity and structure of bacterial communities was described in the soil environment 73 . The study revealed that tetracycline addition could change the microbial community composition and relative abundance of bacteria due to a combination of growth suppressing or bactericidal effect of tetracycline and the induction of bacterial antibiotic resistance. In our study, the antibiotic resistance of the M. cookii related isolate supported the notion that increased domination of the Mycobacteriaceae family in the timentin treated shoots is a consequence of antibiotic resistance. Resistance to carboxypenicillins, such as ticarcillin, had been well documented for a variety of bacteria including actinobacterial taxa 74 . In addition, the resistance to clavulanic acid inhibitor would be required to ensure effective growth observed for the Mycobacterium isolate, which is relatively common among bacteria as well 75 . Resistance to antibiotics is also the most likely explanation for the unchanged or moderately decreased abundance of other bacterial taxa detected in the antibiotic-treated shoots, however, this would require further experimental con rmation.
The antibiotic-induced perturbations in composition and/or interactions within the plant-associated microbial community of the tobacco shoots could be a signi cant factor contributing to the enduring negative effect on the growth and adaptive capacity of the plant tissue culture. Plant microbial community has a profound effect on the health of the host plant and reduced microbial diversity often leads to disease development 76, 77 . It could be proposed that the antibiotic-induced perturbations in microbial community structure could eliminate pathogen suppressing antagonistic interactions and facilitate the spread of pathogenic microorganisms or disease susceptibility of the plants could be increased due to elevated stress level. However, the antibiotic-treated tobacco shoots did not show symptoms that could be attributed to microbial pathogenesis, such as distorted morphology or necrosis of tissues (Supporting information, Figure S1). Alternatively, a loss of mutualistic interaction with bene cial endophytic bacteria could contribute to reduced adaptation to in vitro conditions and inhibited shoot growth. Indeed, growth-promoting 5, 9 stress-reducing activity 6, 8, 10 of bacteria on in vitro propagated plants have been reported previously. A large diversity of bacteria affected by the antibiotic treatment detected in our analysis would lead to an extensive list of potential candidates for the bene cial interaction, therefore, further studies, potentially including speci c bacterial isolates or microbial consortia, would be required to re ne the role of plant microbiome in modulation of in vitro culture growth and adaptation to the in vitro environment.

Conclusion
Our study revealed that the antibiotic timentin induced suppression of growth and elevated stress level of tobacco in vitro shoots is followed by enduring residual negative effect after shoot transfer to medium without antibiotic. The analysis of the control tobacco shoot microbiome detected an extensive bacterial community dominated by the Mycobacteriaceae family which potentially could originate from seeds used to initiate the in vitro culture and might include endophytic species closely associated with the plant host as well as contamination of resilient environmental bacteria. The antibiotic treatment-induced decline of species richness, increased domination of Mycobacteriaceae and several other families were likely associated with antibiotic resistance traits of the bacteria. It was proposed that antibiotic-induced perturbation of shoot microbiome composition and/or interactions might contribute to the reduced adaptive capacity and impede the growth of the tobacco shoots leading to the reduced e ciency of the in vitro culture propagation. Therefore, further insights into the speci c role of the members and interactions of the tobacco shoot microbiome would ensure a better understanding of the antibiotic effect on the in vitro tissue as holobiont and would potentially provide solutions required to improve the e cacy of the in vitro tissue culture.

Methods
Tobacco shoot culture in vitro and antibiotic treatment. Cultivated tobacco (Nicotiana tabacum cv. Samsun-NN) in vitro shoot culture was derived from seeds kindly provided by Prof. J.P.T. Valkonen (University of Helsinki). Plants were cultivated and all experiments were carried out in accordance with relevant guidelines. Tobacco shoots cultivated in vitro for over 5 years were used as a control (TC) and were maintained on solid in Murashige-Skoog (MS) medium 78 , supplemented with 0.75 mg·L −1 6benzylaminopurine, 30 g·L −1 sucrose, 0.8% agar at 25 ± 1 ºC, under 150 µmol·m −2 ·s −1 intensity illumination of 16 h photoperiod. Antibiotic-treated tobacco shoots (TA) were maintained as described for the TC, except the medium was supplemented with timentin at 250 mg·L −1 . To ascertain stable and homogeneous physiological response in the shoot culture, the duration of the treatment was 6 months. Prolonged cultivation was also used to clear residual DNA from antibiotic-inactivated bacteria to avoid false-positive results during the shoot microbiome analysis based on metagenomic DNA sequencing. After transfer to medium without antibiotic (PA), the shoots were maintained under the same conditions as TC and at least one culture passage (one-month duration) was used before collecting samples for the analysis.
Samples for shoot fresh weight (FW) and microbial composition analysis were collected three weeks after transfer to fresh medium, and oxidative stress injury was assessed at four distinct time points during the propagation cycle.
Assessment of oxidative stress injury. Oxidative injury of tobacco shoot cellular membranes was estimated based on quantitative analysis of the lipid peroxidation product malondialdehyde (MDA) using previously described method 79,80 . Homogenized frozen tobacco shoot powder was extracted with 50mM Tris-HCl pH 7.4, containing 1.5% of polyvinylpolypyrrolidone for 30 min at 4°C and centrifuged at 10,000 ⋅g for 15 min at 4°C. Equal amounts of tissue extract and 0.5% thiobarbituric acid in 20% trichloroacetic acid were mixed, heated at 95 ºC for 30 min, cooled on ice and centrifuged at 10,000 ⋅g for 5 min. The absorbance measured at 532 nm was corrected by subtracting the absorbance value at 600 nm and MDA concentration was estimated using ε = 155 mM −1 cm −1 . The absence of interference from the absorbance of anthocyanins at 532 nm was veri ed using control samples without thiobarbituric acid.
Bacterial 16S RNA metagenomic analysis. The metagenomic analysis samples from the TC and PA experimental groups were fresh frozen in liquid N 2 and stored at -70°C. To assess methods most e cient for bacterial DNA extraction from plant material, six different DNA extraction methods described by 39 Mycobacterium isolation, identi cation and antibiotic resistance test. Tobacco shoots were homogenized with a razor blade in MS medium, the homogenate was applied to Loewenstein-Jensen medium prepared from TB-Medium base (Merck, Germany) and incubated at room temperature for 6-8 weeks. The bacteria isolate was maintained on the Actinobacteria Isolation Agar (AIA). Bacterial DNA was isolated using the GeneJET Genomic DNA Puri cation kit (Thermo Fisher Scienti c, USA). The 16S rRNA gene fragment was ampli ed using the universal primers as described previously 10   Principal coordinate analysis (PCoA) of variability among the bacterial OTU data sets generated for the tobacco shoot control (TC) and post-antibiotic treatment (PA) samples using ve 16S rRNA variable region-speci c primer pairs. The analysis was carried out using the Bray-Curtis dissimilarity matrix.

Figure 3
Family level OTU association among data sets generated using ve 16S rRNA variable region-speci c primer pairs.  Representative plots of the alpha diversity indices for tobacco in vitro shoot control (TC) and postantibiotic treatment (PA) microbiome data sets. Cumulative data of all ve primer pairs was used to estimate Richness and Chao1 indices, meanwhile data of the V3 region was excluded from the calculation of Evenness and Shannon indices as an outlier.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.