Modes of carbon fixation in an arsenic and CO2-rich shallow hydrothermal ecosystem

The seafloor sediments of Spathi Bay, Milos Island, Greece, are part of the largest arsenic-CO2-rich shallow submarine hydrothermal ecosystem on Earth. Here, white and brown deposits cap chemically distinct sediments with varying hydrothermal influence. All sediments contain abundant genes for autotrophic carbon fixation used in the Calvin-Benson-Bassham (CBB) and reverse tricaboxylic acid (rTCA) cycles. Both forms of RuBisCO, together with ATP citrate lyase genes in the rTCA cycle, increase with distance from the active hydrothermal centres and decrease with sediment depth. Clustering of RuBisCO Form II with a highly prevalent Zetaproteobacteria 16S rRNA gene density infers that iron-oxidizing bacteria contribute significantly to the sediment CBB cycle gene content. Three clusters form from different microbial guilds, each one encompassing one gene involved in CO2 fixation, aside from sulfate reduction. Our study suggests that the microbially mediated CBB cycle drives carbon fixation in the Spathi Bay sediments that are characterized by diffuse hydrothermal activity, high CO2, As emissions and chemically reduced fluids. This study highlights the breadth of conditions influencing the biogeochemistry in shallow CO2-rich hydrothermal systems and the importance of coupling highly specific process indicators to elucidate the complexity of carbon cycling in these ecosystems.


1-Sediment mineralogy and porewater chemistry
The XRD analysis of the white-capped sediments revealed quartz, birnessite (Mn oxides), (Na,Ca)Mn7O 14 )2.8H2O), at all depths. Alunite (KAl3(SO4)2(OH)6) at 6-8cm (Table S1) was detected in the top and middle depths. In the brown-capped sediments, quartz and birnessite were the main mineral phases at all depths detected by XRD. In the deepest sections, the mica-phases were prevalent (Table S1). In the reference sediments, as is the case for the brown-capped sediments, quartz and birnessite were detected at all depths, in addition to the mica-phase.
The Raman data further reveals graphite as a key component of all the analysed samples, while feldspar, opal, illite and chlorite were detected in most of the samples (Table S2). Hematite was found in all reference and brown-capped sediment samples except in the deepest sandy sample and only in the top layer of the sediment capped by the white deposits (Table S2). Pyrite and marcasite were detected in the deepest white-capped sediments and sulphur in the surface layer, but not elsewhere or in the other samples (Table S2).
Between 3-5 ml of porewater was recovered from most samples and filtered under anoxic conditions in an anaerobic glove bag (Table S3) for chemical analysis. Porewater samples with a volume <1 ml, especially for the white-capped and reference sediments (Table S3), was associated with sediments exhibiting higher porosity. With the exception of the top and the deeper sections of the push core samples from the white deposits, no porewater was recovered for the middle depths after centrifugation (Table S3).
Concentrations of Ca, K and Mn were higher in porewater collected from the white-and brown-capped deposits and lower in reference sediments (Table S3). Total dissolved S, determined as total dissolved sulphur in acidified filtered samples was used as a proxy for sulphate concentrations 1 . S concentrations showed a larger variability in the brown-capped deposit (7.8×10 2 -3×10 6 ppm) compared to the white-capped sediment (3.7×10 3 -8.4×10 3 ppm) and sandy reference also had more variable concentrations (0-8.8×10 3 ppm) (Table S3). Low concentrations of porewater Fe were measured; 94, 209 and 946 ppb for the white-capped, brown-capped and reference sediments, respectively.

2-Statistical analysis
The MANOVA test indicates that the 3 habitats: white-, brown-capped and reference sediments, generally plays an important role in genetic abundance (P<0.005). However, there is no obvious difference between archaeal 16S rRNA gene and aclB gene abundance in the browncapped and reference sediments (Table S6). The same was observed for the dsrA gene distribution between the white and the brown-capped sediments (Table S6). Considering the specific gene abundance and the individual push cores, the MANOVA test pointed out that replicate cores collected in the white-capped sediments show no statistically supported differences. However, the MANOVA test showed that for several of the quantified genes: Zetaproteobacteria 16S rRNA, dsrA, aclB, cbbL and cbbM, the habitat and the replicate cores are a significant factor controlling gene abundance (Table S7). Also, the dsrA gene abundance does not vary with replicate samples between the three habitats. Related to gene abundances and depth, a global transition can be observed in between 4-8 cm, with the exception of the RuBisCO encoded genes (Table S8).
Generally, before and after this transition, most of the genes show no significant depth-dependent abundance.
Significant correlations are found between Zetaproteobacteria abundance and the cbbM genes in all cores (R 2 > 0.90), except in the case of push core 2 collected in the brown-capped sediments. A strong correlation (R 2 > 0.93) between the archaeal 16S rRNA gene abundance and the aclB gene is also established in the brown-capped sediments. The pmoA and mmoX gene abundances, key indicators of the aerobic methanotrophy, as well as the anaerobic methane cycle mcrA genes, were quantified in all habitats and at most depths (Fig. 4). The sulphate-reducing bacteria dsrA gene was also detected in all habitats and depths (Fig. 4). Between the dsrA and mcrA genes, a significant correlation was found for the white-capped and reference sediments, with R 2 =0.98 and 0.69, respectively. In contrast, no correlation (R 2 = 0) is observed between the dsrA and mcrA genes in the brown-capped sediments. With the exception of push core 2 collected in the white-capped sediments, all push cores showed a significant correlation between the mmoX, pmoA and mcrA genes, R 2 > 0.73 and R 2 > 0.90, respectively. With the exception of the cbbM and the Zetaproteobacteria 16S rRNA gene abundances, the highest genetic abundance is in the reference sediments and lowest in the white-capped sediments.

1-Sediment mineralogy and porewater chemistry
The Raman data acquisition was obtained with an 1800 lines/mm grating and the excitation was done with an Ar-ion laser (514 nm) source. A confocal Olympus BX41 microscope was combined to the instrument. The laser beam, with a power at the sample surface at 8 mW, was focused through an 80x objective with a long-working distance of 8mm. The analysed spot size was around 1 µm. The spectral resolution was ~0.4 per cm/pixel and the accuracy of the instrument was controlled by repeated use of a silicon wafer calibration standard. The data collection and spectral baseline correction were done with the LabSpec 5 software.
At the Stockholm University, the porewater from the geomicrobiology push cores was extracted anaerobically. The sediments samples were centrifuged 10 min at 6 000rpm and then the porewater was recovered, filtered through 0.2 µm (Filtropur S, Sarstedt) membrane and stored at 4°C. For several sediment samples, porewater quantities were low. Samples from which <1 ml porewater was collected were analyzed without filtration. The seawater overlying the top of the geomicrobiology push core was also analyzed. Prior to elemental analysis, samples were acidified to 0.28 mol L -1 with ultra-pure grade HNO3 and further diluted 100-fold with 0.28 mol L -1 HNO3.

2-Sediment carbon isotope analysis for the assessment of δ 13 C DIC and δ 13 C POC
Samples were collected for δ 13

Supplementary Information
The inherent bias of the qPCR: The sensitivity and specificity of qPCR makes it a valuable screening and quantitative tool, but also introduces biases through inefficient nucleic acid extraction, poor purity, and yield [3][4][5][6][7] .
Primers set specificity and qPCR parameters are also critical 8,9 . None of the commonly used nucleic acid extraction protocols are universally perfect. Consequently, different techniques may yield different results 3,5 . We used the PowerSoil DNA extraction kit (MoBio), often employed in DNA extraction from marine hydrothermal systems 7,10 . The universal primers used, were designed from gene bank database sequences, consequently we cannot exclude the possibility of lower coverage of 16S rRNA gene abundances, if our samples contained native lineages not having representative sequences in the gene bank and thereby not recognized by the universal primers 11 .
For example, universal 16S rRNA gene primers failed to detect anammox lineages in the Plamctomycetes identified with specific primers in deep-sea hydrothermal mats 12 . Similar observations were also found in studies using universal 18S rRNA gene and specific Botryococcus sp. qPCR primers, in which the specific primers revealed higher abundance of the specific taxa than the total Eukarya 13 . The case of the Zetaproteobacteria 16S rRNA specific and total bacteria 16S rRNA gene quantification is most obvious in our data (Fig. 5), which may also be further coupled to gene copy number variability. Moreover, because we compare qPCR data treated with the same methods, we minimize sample treatment biases, enabling comparative analysis across all habitats.

The six main pathways of the inorganic carbon fixation:
The Calvin cycle-the most studied carbon fixation pathway-is found in many photosynthetic and chemolithotrophic prokaryotes, algae and plants 14,15 . Within the Prokaryotes, the Calvin cycle is known to be present in Alpha-, Beta-, Gamma-, and Zeta-proteobacteria, the Cyanobacteria and as well as in some Firmicutes, Chloroflexi and archaea 14,[16][17][18] . RuBisCO, the main enzyme used for ribulose 1,5-bisphosphate (RuBP) carboxylation in the first critical step of inorganic carbon fixation to organic carbon, is a marker of the Calvin cycle 14,16 . RuBisCO is currently distinguished into forms I, II, III and IV 16,19 . Only forms I-III have RuBP-dependent CO2 fixing activity, while form IV encodes RuBisCO-like-proteins unable to fix CO2, but apparently involved in sulphur metabolisms 14,[19][20][21] . In deep and shallow seafloor hydrothermal systems, the cbbL and cbbM genes encoding the large unit of RuBisCO forms I and II, respectively, have been produce small δ 13 Corg values of up to -3.8‰ 14,15 . Finally, the reductive acetyl-CoA pathway (Wood-Ljungdahl pathway), widespread in acetogenic bacteria and methanogenic archaea, characteristically produces δ 13 Corg values >-30‰ 14,15 . Biological methane cycling by microorganisms in this pathway may contribute to carbon cycling and to CO2 production through anaerobic and aerobic methanotrophy in various submarine hydrothermal ecosystems.
Both methanogenesis and anaerobic methanotrophy are carried out by members of the domain archaea 29 . However, there is a paucity of direct δ 13 Corg signature linked to the anaerobic methanotrophy because even some ANME are also able to live alone 30,31 , while some live often in consortia with sulfate-reducers, iron-reducers or nitrate-reducers 32 , and so far, no pure strain has been isolated. Aerobic methanotrophy is widespread in some Gammaproteobacteria and Alphaproteobacteria 33,34 and the first cited class is believed to predominate hydrothermal ecosystems [35][36][37] . While anaerobic methane oxidation is known to be widespread in sedimentary hydrothermal systems 10,38 , it is rarely detected in hydrothermal chimneys 39 . The δ 13 Corg signature of the various methanotrophic processes varies dependent on the source of methane (thermogenic, abiogenic or biogenic). Furthermore, the ribulose monophosphate aerobic methane oxidation pathway in the Gammaproteobacteria and mostly the serine pathway for the Alphaproteobacteria members 33,34 , fractionate carbon differently 40,41 .

Supplementary tables:
Supplementary Table 1. Minerals identified by (X-ray diffraction) XRD in various sediment types and the depth. Yellow represents minerals found in all the habitats and depth. Purple represents mineral detected only in the white-capped sediments. Green represents minerals identified in the in the brown-capped sediments and the sandy reference sediment.  Green, represent minerals found detected in the brown-capped sediments and the sandy reference sediment.