Microbial methanol sink of a grass and a flower host species from a temperate 1 grassland 2

25 Background: Managed grasslands are global sources of atmospheric methanol, which is one of the most abundant volatile organic compounds (VOCs) in the atmosphere and 27 promotes oxidative capacity for tropospheric and stratospheric ozone depletion. The 28 phyllosphere is a favoured habitat of plant-colonizing methanol-utilizing bacteria. These 29 bacteria also occur in the rhizosphere, but their relevance for methanol consumption and 30 ecosystem fluxes is unclear. The estimated global methanol emission rate is considerably 31 higher than those escaped into the atmosphere. Thus, methanol utilizers in the plant 32 microbiota might be key for the mitigation of methanol emission through consumption. 33 However, information about grassland plant microbiota members, their biodiversity and 34 metabolic traits, and thus key actors in the global methanol budget is largely lacking. 35 Results: We investigated the methanol utilization and consumption potentials of two 36 common plant species ( Festuca arundinacea and Taraxacum officinale ) in a temperate and 37 fertilized grassland. The selected grassland exhibited net methanol emission. The detection 38 of 13 C derived from 13 C-methanol in 16S rRNA of the plant microbiota by stable isotope 39 probing (SIP) revealed distinct methanol utilizer communities in the phyllosphere, roots and 40 rhizosphere but not between plant host species. The phyllosphere was colonized by 41 members of Gamma - and Betaproteobacteria . In the rhizosphere, 13 C-labelled Bacteria were 42 affiliated with Deltaproteobacteria , Gemmatimonadates, and Verrucomicrobiae. Less- 43 abundant 13 C-labelled Bacteria were affiliated with well-known methylotrophs of Alpha -, 44 Gamma -, and Betaproteobacteria . Additional metagenome analyses of both plants were 45 consistent with the SIP results and revealed Bacteria with methanol dehydrogenases (e.g., MxaF1 and XoxF1-5 ) of known but also unusual genera (i.e., Methylomirabilis , , , Verminephrobacter Conclusions: The rhizosphere has been shown to be an overlooked hotspot for methanol 51 consumption in grasslands. We also identified unusual methanol utilizers in the phyllosphere 52 and rhizosphere. We did not observe a plant host-specific methanol utilizer community. Our 53 results suggest a model for methanol turnover in which both the sources (plants) and sinks 54 (microbiota) of a volatile are separated but in the same ecological unit. 55

dataset) were removed. Consensus sequences were determined for each OTU at 3% 214 genetic divergence using USEARCH and classified by BLAST alignment against the SILVA 215 SSURef 119 NR database [48]. Sequences were classified with respect to the SILVA 216 taxonomy of the best hit. Rarefaction curves and Shannon diversity indices [49] were 217 calculated as previously described [50]. In addition, the maximal number of OTUs (nmax) was 218 estimated for each sample using the Michaelis-Menten function fit. The OTUs were analysed 219 for confirmation of 13 C-labelled microbe-specific selection criteria as described previously [51, 220 52]. A few modifications were made to those criteria due to the labile nature of RNA, i.e., (i) 221 the relative abundance of a specific taxon in the 13 C treatment's heavy fraction should be 222 higher than that in same fraction of the 12 C-control treatment; (ii) the relative abundance of a 223 specific OTU in the heavy fraction should be higher than that in the light fraction of the 224 gradient of the 13C treatment by a factor K=2, due to the substrate-based stable isotope 225 approach and short incubation periods; and (iii) the relative abundance of a specific OTU in 226 the heavy fraction of the 13 C treatment should be larger than or equal to 0.05%. 227 228

Metagenomes from both plant species including bioinformatics analyses 229
Metagenome sequencing was performed for two DNA samples from pooled plant material 230 (phyllosphere, roots, rhizosphere soil) of both plant species incubated with [ 13 C]-CH3OH for 8 231 h. Sequencing was performed on the Illumina NextSeq platform, and raw read data were pre-232 processed and trimmed by a commercial service (LGC Genomics GmbH, Berlin). 233 Taxonomical analysis of the processed and trimmed reads was performed with Kaiju version 234 1.7.3 [53]. Processed reads were assembled with SPAdes 3.12.0, which includes the 235 metaSPAdes pipeline (Nurk & Bankevich et al., 2013) with default options. QUAST v4.0 was 236 used to check the assembly statistics for both metagenomes (Table 1). The rarefaction 237 curves of both metagenomes achieved sufficient coverage. Nonetheless, the F. arundinacea 238 metagenome had a slightly higher quality and coverage than that of T. officinale 239 ( Supplementary Fig. S4

246
Assembled contigs were again taxonomically and functionally classified using the MG-RAST 247 server [54]. KEGG pathways for methanol assimilation were also obtained from MG- RAST. 248 Assembled sequences were binned using MYCC [55] with 4mer and a minimum contig 249 length of 1000 bp. The coverage profiles were produced through MYCC and BAM files using 250 Bowtie 2 and MetaBAT to produce a depth file. The quality of the bins was estimated using 251 Radioactive labelling with 14 C1-methanol to determine methanol oxidation rates 267 To quantify potential methanol turnover rates in the phyllosphere, roots and rhizosphere soil 268 of the investigated plants, freshly excised plant material (leaves, roots, or rhizosphere soil) 269 was incubated with 14 C-CH3OH for 3.5 h at 20 °C (performed in a climate chamber). was incubated with 631 kBq of 14 C-CH3OH for 3.5 h in 4 replicate glass bottles (SCHOTT 275 DURAN, 100 mL), each with a sterile inlet and outlet for gases placed in a climate chamber 276 on a shaker. The inlet of the glass bottle was connected to test tubes with water to maintain 277 humidity. The outlet was connected to both a CO2 trap (NaOH solution) and a methanol trap 278 (water). The water to trap vapourized 14 C-CH3OH was maintained at 3 °C using a cryostat. 279 Thus, evaporation of condensed methanol in water was prevented. For all the traps, test 280 tubes (three of them, always connected in parallel) with 12 mL of water or NaOH were used. 281 At the rear end, after the 14 C-methanol trap, a mass flow controller with a constant gas flow 282 (15 mL min -1 ) and a pump were installed. During the experiment, test tubes with 1 M NaOH 283 solution (CO2 traps) were collected every 30 min, while methanol traps with cold water were 284 collected only once at the end of the experiment. 285 The activity of 14 C in all the traps was determined by a TriCarb 2900 TR liquid scintillation 286 counter (PerkinElmer). The scintillation mixture (15 mL) was prepared with 12 mL of 287 UltimaGold (PerkinElmer) and either 3 mL of pooled NaOH solution or water from both the 288 13 CO2 and methanol traps. Linear regression analyses were used to determine the slope of 289 methanol turnover for each incubation setup, where the mean R 2 of all 24 incubation 290 experiments was 0.986 ± 0.017. Then, the slope values and dry biomass of excised plant 291 material (leaves, roots, and rhizosphere soil) from both plant species (F. arundinacea and T. 292 officinale) were used to determine the methanol turnover rate (nmol g dry wt -1 h-1 ). The dry 293 biomasses of leaves, roots, and rhizosphere soil were 50, 32 and 28 mg, respectively. The 294 efficiency of the labelling method was tested by measuring the total content of CO2 collected 295 in NaOH using 0.5 M BaCl2. Precipitated CO2 (BaCO3) was washed on a membrane filter 296 with ultrapure water and dried at 104 °C. Then , the membrane filters with BaCO3 were 297 combusted at 1350 °C (multi EA 4000, Analytik Jena, Germany) with a continuous flow of 298 oxygen, and the resulting CO2 was trapped in 7 mL CarboSorb E (PerkinElmer). 299 Subsequently, 3 mL of the CarboSorb E sample was mixed with 12 mL of Permafluor E+ 300 (PerkinElmer), and then, the activity of 14 C was determined by using a liquid scintillation 301 counter. The total activity of all BaCO3 precipitates determined here was 3.3% more than the 302 expected value resulting from the 14 C activities of the NaOH samples. This result indicates a 303 minor methodological error. However, the accumulation of volatile methanol in the CO2 traps 304 can be excluded. In addition, the linear regression analysis of the 14 C activities of the NaOH 305 and BaCO3 samples showed an R 2 of 0.945. 306 307

Results and discussion 308
Previous investigations on methanol utilizers have mostly focused on forest soils [6,12]. 309 Fewer studies on plant-associated methanol utilizers have provided valuable insights and 310 highlighted their importance in global methanol emissions [19,22,24]. Recently, Macey and 311 coauthors revealed the importance of methanol utilizers in bulk and plant-associated soils 312 using a combined approach with molecular probes, DNA SIP and metagenomics [51]. 313 Nevertheless, by separating the rhizosphere soil from the plant, they provided the first 314 evidence that the rhizosphere has methanol consumption activity. However, the study only 315 14 analysed rhizosphere soil after destructive sampling. Our study provides detailed information 316 about the role of methanol utilizers in intact plant methanol consumption rates and active 317 methanol-incorporating bacteria, and we were still able to experimentally separate the 318 phyllosphere and rhizosphere. 319 320 Active bacterial methanol utilizers of the phyllosphere and rhizosphere of both plant species 321 It is well known that RNA has higher sensitivity and exhibits more rapid metabolic turnover 322 than DNA [60]. Thus, using RNA SIP instead of DNA SIP to identify active methanol utilizers 323 have an added advantage in our study; we avoided unnecessarily long incubation periods 324 and hence minimized potential stress for the plants caused by labelling in closed chambers. The presence of Gemmatimonadetes in both metagenomes and in the active methanol 358 community as revealed by RNA SIP is striking and suggests that these species are methanol 359 utilizers. Despite their tenacious presence and high abundance (ca. 2%) in many rhizosphere 360 and soil studies, they are frequently ignored due to a lack of cultivable isolates [70]. 361 Representatives from this phylum are often involved in nitrogen and sulphur cycles, but 362 recent studies on their genomes revealed the presence of MDH genes, and thus, can be 363 considered as methylotrophs [71]. We also detected members of the phylum Acidobacteria 364 (Holophagae). This phylum is also lacks cultivable isolates and was found to be abundant in 365 a study on methanol-utilizing bacteria in a forest soil [72]. Recently, a member of 366 Acidobacteria (Solibacter) was described as a methylotroph. It possess the xoxF3 gene. 367 Thus, our study proved that these phyla that have been long overlooked in regard to 368 methanol utilization are relevant in common grassland plant hosts for methanol turnover. Verrucomicrobia (3%) and Firmicutes (5%) (Fig. 2, Supplementary Fig. S9; Additional file 9). 377 The predominance of these members was expected since they have often been identified in 378 various studies on methanol-degrading microbes. However, based on average frequencies in 379 both metagenomes, a few phyla were abundant in our study and have gone unnoticed thus 380 far, such as Deltaproteobacteria (6%), Plantomycetes (3.5%) Acidobacteria (3%), and 381 Gemmatimonadetes (1%). The highly abundant genera and families in both metagenomes 382 were consistent with previous studies on methanol utilizers in plants and soils [14-16, 22, 51]  The complexity of metagenomic datasets and their processing can lead to a high level of 399 genome fragmentation and heterogeneity, which might shift the microbe distribution patterns 400 and can imbricate the microbiota compositions [78,79]. These technical challenges can be 401 overcome by binning, as these approaches often use abundance information from scaffolds 402 or contigs. Binning of the contigs from MAGs of both plant species revealed 29 and 14 403 annotated bins (Fig. 4). Annotated bins identified as Bacteria were shortlisted. All genome 404 bins were screened for PQQ-dependent MDH gene markers, such as mxaF or xoxF (1-5) 405 (Table 2). Both plant species were dominated by typical representatives from Alpha-, 406 and Gammaproteobacteria and Actinobacteria and a few unexpected members affiliated with 407 Deltaproteobacteria, Acidobacteria, Gemmatimonadetes and Bacilli. The genera of methanol 408 utilizers detected in metagenomes from both MAGs were Methylobacillus, Methylosinus, 409 Methylomirabilis, Methylooceanibacter, Gemmatimonas and Verminephrobacter (Fig. 4). A 410 few detected members of Acidobacteria, Gemmatimonadetes and Bacilli (e.g., 411 Gemmatimonas and Verminephrobacter) have never been detected previously. However, 412 these taxa have been observed in many soil-and plant-associated habitat-based studies on 413 methylotrophs. Therefore, we aimed to provide metabolic insights into these methanol 414 utilizers. Only a few recent 16S rRNA-based, metagenome-and proteogenome-targeted 415 studies have suggested the presence of methylotrophy in low-abundant phyla (e.g., 416 Acidobacteria, Gemmatimonadetes and Firmicutes), thus suggesting their role in terrestrial 417 methanol consumption [26,51,80]. Interestingly, in a recent proteome study, a PQQ-418 dependent MDH from Gemmatimonadetes was detected as the most abundant protein [26]. leading to other crucial methanol utilizers being overlooked. Due to their low relative 453 abundances, the above described 'unusual' methanol utilizers have never been detected and 454 identified. We applied, for the first time, a 13 C-labelling approach in a plant microbial 455 interaction study to separately label the phyllosphere and rhizosphere compartments while 456 leaving the plants intact. Hence, we could exclude the exchange of the labelled compounds 457 between the plant compartments. We also ensured that sufficient 13 C labelling occurred by 458 investigating 13 C incorporation within 8 h and 24 h of incubation with 13 C-CH3OH by 459 subjecting the plant material (leaves, roots and rhizosphere soil) to further analyses or by 460 examining their associated microbial communities using EA/IRMS ( Supplementary Fig. S12, 461 Additional file 12).  13 C values of unlabelled leaves and roots of F. arundinacea were -28.8 to 462 We used radioactive isotope turnover measurement with [ 14 C]-methanol as a tracer to reveal 475 potential methanol oxidation rates in the investigated grassland species in all plant 476 compartments. We used newly developed water traps to determine radioactivity loss through 477 evaporated 14 C-methanol. The water was cooled at 3 °C for maximal trapping of 14 C-478 methanol. The efficacy was 97.5% with only 0.23% methanol in CO2 traps. Thus, we were 479 able to quantify the potentially higher recovery rates for CO2 formation, which we used to 480 calculate potential methanol oxidation rates. 481 20 The methanol oxidation rates were dependent on the plant species. T. officinale samples had 482 higher rates than F. arundinacea samples (Fig. 6). The phyllosphere of T. officinale exhibited 483 the highest methanol oxidation rates (149 ± 15 nmol g dry wt -1 h -1 ). Roots of T. officinale had 484 higher methanol oxidation rates (131 ± 26 nmol g dry wt -1 h -1 ) than the rhizosphere soil (87 ± 485 12 nmol g dry wt -1 h -1 ), while F. arundinacea revealed the opposite trend. The rhizospheres 486 (roots and rhizosphere soil) of both plant species showed similarly high methanol oxidation 487 rates as the respective phyllosphere compartments. ANOVA (two-way and one-way), T-test 488 and Welch's test revealed a significant difference between both the plant species and plant 489 materials (α = 5%). Thus, our study proved that the rhizosphere of the two common 490 grassland plant host species is a highly active and therefore relevant methanol sink in such 491

ecosystems. 492
Spot check quantification of methanol emissions using closed flux chambers and SIFT MS in 493 a managed grassland provided insights into sources and sinks within the plant holobiont 494 ( Supplementary Fig. S13, Additional files 13). Methanol mixing ratios in the air from three 495 experimental plots with two different plant species (approx. 26.5 ± 1.2 ppb) showed higher 496 methanol emissions than the plot without plant biomass (17.1 ± 0.7 ppb) (Fig. 7). Therefore, 497 we confirmed that both above-and belowground plant parts are net methanol emitters. 498 499

Conclusions 500
Our study revealed the rhizosphere of temperate grassland plant species as an overlooked 501 local methanol sink and revealed new bacterial taxa that together with known ones represent 502 the plant host-associated methanol sink. This finding led us to re-evaluate the canonical 503 concept of members of the family Methylobacteriaceae being the key sink of methanol in 504 plant species. The rhizosphere of both plant species was identified as a major sink for 505 methanol in terrestrial ecosystems. To our knowledge, there has been no previous study that 506 quantified this sink activity in the plant rhizosphere by direct measurement. Our study 507 confirms a long-held assumption that the rhizosphere is one of the hotspots for methanol 508 consumption in grasslands. Eventually, this finding implies that the observed net surface 509 fluxes from grasslands and their responses to land use and climate change can be 510 understood only if the belowground microbiota and its activity are considered. Our study 511 suggests a conceptual model of plant-associated microbes as modulators of methanol fluxes 512 at the ecosystem scale. This will help resolve uncertainties in the current global methanol 513 budget models. 514

Declarations 515
Not applicable 516

Ethics approval and consent to participate 517
Not applicable 518

Consent for publication 519
Not applicable 520

Availability of data and material 521
Read data of the two metagenomes have been submitted to the National Center for 522 Biotechnology Information (NCBI) database under BioProject number PRJNA715626. Raw 523 read data of all 60 samples were deposited in the NCBI sequence short-read archive under 524 the same BioProject (PRJNA715626) with accession numbers SRA14001742 to 525

Competing interests 527
The authors declare that they have no competing interests. 528

Authors' contributions 532
SKo conceptualized the study. SKa designed and performed a series of experiments and 533 analysed the data. RR conducted and analysed the radioactive labelling experiments. CBT 534 and AMR interpreted the EA/IRMS and SIFT-MS data, respectively. MH provided support in 535 spot check quantification. SKa and SKo together wrote the manuscript. 536

Acknowledgements 537
We thank all members of the group Microbial Biogeochemistry for their support, especially 538