Samples were collected in lakes and lagoons on three different islands in the Azores Archipelago in June 2018 (Fig. 1 and Table 1): São Miguel (lakes Azul, Verde, and Empadadas), São Jorge (Lake São Jorge and lagoons Cubres East and Cubres West), and Flores (lakes Funda, Lomba, and Negra). June is considered part of the dry season in the Azores, which ranges from April to August. From September to March, the Azores Archipelago receives more rain and is exposed to high winds (Santos et al., 2004; Hernández et al., 2016). During the summer months, deeper lakes such as Azul, Verde, Funda, and Negra stratify and the bottom water becomes suboxic and, in some years, even fully anoxic (Gonçalves, 2008; Gonçalves et al., 2018). In contrast, shallow lakes are typically considered polymictic but might undergo short periods of stagnation during the summer months (Gonçalves, 2008). The lakes sampled for this study are either mesotrophic or eutrophic and range in water column depth from 2 to 115 m. During the summer months, lakes Verde and Funda experience large cyanobacterial blooms (Cordeiro et al., 2020). Both lakes Verde and Funda were experiencing a cyanobacterial bloom at the time of sampling.

Suspended particulate matter was collected from the surface water, near the oxycline and/or chemocline at the time of sampling and from the bottom water of the lake without disturbing the sediment–water interface. Water (5 to 10 L) was immediately filtered through Whatman GF/F 142 mm 0.7 µm pore size filters and frozen in liquid nitrogen. Surface sediment samples (0–1 cm) were also obtained at the same time using a gravity corer (UWITEC 90 mm, Austria). In Azul and Verde, surface sediments were collected near the lakeshore and in the deepest part of the lake. Replicate samples were collected for sediment cores from lakes Azul (n=2), Empadadas (n=4), and Negra (n=2) and analyzed for reproducibility. All samples were immediately frozen in a liquid nitrogen dewar and transported to the NIOZ Royal Netherlands Institute for Sea Research, where SPM and sediment samples were stored at -80 and -40 ∘C, respectively, until analysis. At the time of sample collection, water column profiles were obtained using a Hydrolab (Horiba U-50) for dissolved oxygen content (percent saturation and mg L-1), temperature, salinity, conductivity, pH, and turbidity (NTU, nephelometric turbidity unit). Additional long-term data were obtained from the Universidade dos Açores long-term monitoring program (data from 2003–2020).

An enrichment co-culture consisting of a methanotroph (43 % Methylobacter sp.) and a methylotroph (21 % Methylotenera sp.) was previously isolated from a seasonally stratified and hypereutrophic lake (Lacamas Lake, WA, USA; van Grinsven et al., 2020). This culture was grown in triplicate under oxic conditions in the dark at 15 ∘C on a nitrate mineral salt (NMS) medium (Whittenbury et al., 1970). For lipid extractions, the cultures were grown on 250 mL NMS media in 580 mL acid-washed and autoclaved glass pressure bottles with butyl rubber stoppers with an addition of methane (16 mL, 99.99 % pure) to the headspace. After the methane gas was consumed, the cultures were filtered onto muffled Whatman GF/F 47 mm 0.3 µm pore size filters and frozen at -80 ∘C. The methylotroph identified in the co-culture is closely related to Methylotenera versatilis and Methylotenera mobilis (van Grinsven et al., 2020). Therefore, we obtained freeze-dried biomass of Methylotenera mobilis (DSM 17540) from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ) culture collection to investigate the presence of BHPs in this methylotroph.

All samples were freeze-dried and extracted using a modified Bligh–Dyer method (Bligh and Dyer, 1959; Bale et al., 2021). Samples were submerged in methanol (MeOH), dicholoromethane (DCM), and phosphate buffer (2:1:0.8, v:v:v) and ultrasonically extracted twice. The solvent was collected in a separate flask after each extraction. DCM and phosphate buffer were added to the resulting solvent to obtain a new volume ratio of 1:1:0.9 (v:v:v). The DCM layer was collected, and the aqueous layer was washed two more times with DCM. The extraction was repeated using MeOH : DCM : aqueous trichloroacetic acid solution (2:1:0.8, v:v:v) following the same procedure described above. The combined DCM layers were dried under N2 gas and stored at -20 ∘C until analysis. Prior to analysis, deuterated diacylglyceryltrimethylhomoserine (DGTS D-9; Avanti^® Polar Lipids, USA) was added to the extracts as an internal standard, and the samples were redissolved in MeOH:DCM (9:1, v:v) before filtering the samples through a 0.45 µm regenerated cellulose syringe filter (4 mm diameter; Grace Alltech, Deerfield, IL).

Samples were analyzed following the methods described in Hopmans et al. (2021). Briefly, samples were analyzed on an Agilent 1290 Infinity I UHPLC (ultra-high-performance liquid chromatography) coupled to a Quadrupole-Orbitrap HRMS (high-resolution mass spectrometry) (Q Exactive, Thermo Fisher Scientific, Waltham, MA) equipped with an Ion Max source and heated electrospray ionization (HESI) probe (Thermo Fisher Scientific, Waltham, MA). Separation was achieved on an ACQUITY BEH C18 column (2.1×150 mm, 1.7 µm particle; Waters) and precolumn. The solvent system consisted of (A) MeOH : H2O (85:15) and (B) MeOH : isopropanol (1:1) with both containing 0.12 % (v/v) formic acid and 0.04 % (v/v) aqueous ammonia. Lipids were detected using positive ion monitoring of m/z 350–2000 (resolution of 70 000 ppm at m/z 200) with an inclusion list of calculated exact masses of BHPs: 171 for sediment samples and 357 for water column samples. Note that the difference in the number of exact masses in the inclusion lists reflects the addition of further novel BHPs after their identification. A Lake Verde sediment sample (Verde SS1) that included all the novel BHPs was rerun with an updated inclusion list (421 exact masses) for confirmation of novel BHPs. For untargeted BHP detection and identification, we used data-dependent MS2 (tandem mass spectrometry) (isolation window of 1 m/z; resolution of 17 500 ppm at m/z 200) of the 10 most abundant ions for a total cycle of ca. 1.2 s and dynamic exclusion (6 s) with a 3 ppm mass tolerance. To obtain optimal fragmentation of BHPs, we used a stepped normalized collision energy of 22.5 and 40. Mass calibration was performed every 48 h using a Thermo Scientific Pierce LTQ Velos ESI Positive Ion Calibration Solution.

BHPs were identified based on their retention time, exact mass, and fragmentation spectra. Integrations were performed on (summed) mass chromatograms (within 3 ppm of mass accuracy) of relevant molecular ions ([M + H]+, [M + NH4]+, and [M + Na]+). We used an internal standard for normalization between sample runs by calculating the average peak area of the internal standard and correcting the peak areas to the average internal standard. However, we do not have an internal standard for BHP quantification; therefore all BHPs are reported as response units (RU). BHPs identified in water column and surface sediment samples are normalized to liters of water filtered (L) and grams of freeze-dried sediment (g), respectively.

Total organic carbon (TOC) and total nitrogen (TN) content was measured for the surface sediments (n=18) as described in Mitrović et al. (2023). Briefly, the samples were decalcified, powdered, and freeze-dried prior to analysis. All samples were analyzed using an Elementar vario ISOTOPE cube coupled to an isotope ratio mass spectrometer (IRMS) Elementar ISOPRIME visION. Prior to TOC analysis, carbonates were removed by pretreating the samples with excess 2 M hydrochloric acid (HCl) on a shaker overnight. The resulting samples were neutralized with bidistilled water and freeze-dried. All TOC samples were measured in duplicate with a reproducibility of <0.5 %.

δ13C and δ15N values of bulk organic matter were measured on decalcified samples of surface sediments (n=18) using an Elementar analyzer (Elementar vario ISOTOPE cube) coupled to an IRMS (Elementar ISOPRIME visION). Stable isotope ratios of 13C and 15N are expressed using the δ notation in units per mil, reported relative to Vienna Pee Dee Belemnite (VPDB) and atmospheric N2, respectively. The results were normalized to certified standards, and standard deviations were <0.05 ‰.

A soil index (Rsoil) was developed to trace inputs of terrestrially derived BHPs into the marine environment (Zhu et al., 2011). This index relies on the observations that BHT is present in higher abundance relative to nucleosides in the marine environment than in the soils (Pearson et al., 2009; Rethemeyer et al., 2010; Zhu et al., 2011). The index is calculated as follows: 1Rsoil=(soil-marker BHPs)(soil-markerBHPs+BHT). Previously, “soil-marker BHPs” included only adenosylhopane, N1-methylinosylhopane (i.e., type-2 adenosylhopane after Hopmans et al., 2021), adenosylhopaneHG-Me (i.e., type-3 adenosylhopane after Hopmans et al., 2021), and their ring-methylated counterparts. BHT in the Rsoil equation only refers to the isomer with 17, 21(H), 22R, 32R, 33R, 34S stereochemistry. Previous high-latitude studies have revised this index, Rsoil′, to exclude methylated nucleosides (Doðrul Selver et al., 2012): 2R′soil=(adenosylhopane+N1-methylinosylhopane+adenosylhopaneHG-Me)(adenosylhopane+N1-methylinosylhopane+adenosylhopaneHG-Me+BHT). In this study, we tested both of these indices at our sample sites.

Non-metric multidimensional scaling (NMDS) analysis was conducted using a Bray–Curtis dissimilarity matrix to visualize the similarities between BHP compositions at our sample sites. All analyses were performed and visualized in R (version R-4.2.1; R Core Team, 2022) using the vegan package (version 2.5–7; Oksanen et al., 2020) and ggplot2 (version 3.3.5; Wickham, 2016). NMDS plots were generated for both the water column and sediment samples. To test whether there was a significant difference between our sample sites and sample site types (for sediment samples: deep lakes, shallow lakes, and coastal environments; for water column: surface water, chemocline, and bottom water), we performed analysis of similarities (ANOSIM) tests in R using the vegan package for the sediment samples and water column samples. A Bray–Curtis dissimilarity matrix was used for these tests.

At our study sites, we identified 83 BHPs in total (Table A1). For the identification of novel BHPs, we discuss the carbon positions, rings, and modifications to the side chain of the core BHP structure as shown in Fig. B1. The novel BHPs are briefly discussed here and described in more detail in Appendix B. In the Azorean lakes, BHPs were more diverse in surface sediments (average of 51 BHPs per sediment sample) than in the water column SPM (average of 11 BHPs per SPM; Fig. 2).

Nucleoside BHPs were particularly diverse in the water column SPM and surface sediments in lakes Empadadas, São Jorge, and Lomba. We identified 14 nucleoside BHPs, including 3 inosylhopanes, described by Hopmans et al. (2021; Appendix B; Figs. B2, B3). Further, we identified five additional adenosyl-BHP isomers (Fig. B2), including early-eluting isomers (peaks b*, f*, g*, and k*). We also detected four additional inosylhopane isomers in surface sediments from Lake Empadadas with modifications in the position of the methyl group on the ring structure and/or head group (Fig. B3, peaks b*, c*, f*, and g*).

BHT (17, 21(H), 22R, 32R, 33R, 34S stereochemistry) was predominant in most samples, along with unsaturated and methylated versions of BHT. In the surface sediment from the coastal lagoons, Cubres East and Cubres West, we identified two compounds comparable to BHT; however, in the MS2 spectrum we observe the loss a methoxy moiety (32 Da (dalton), CH3OH), as well as three hydroxyl moieties (Appendix B2, Fig. B4). We tentatively identify these compounds as methoxylated BHTs. BHT cyclitol ether (CE) was also abundant in both water column SPM and surface sediments (Fig. 2). In addition, we detected MeBHT-CE in surface sediments from Lake Verde, as well as an isomer where, based on the MS2 spectrum, the methylation occurs on the cyclitol ether head group (Appendix B3, Fig. B5). We putatively identify this isomer as BHT-MeCE. BHpentol, BHpentol-CE, BHhexol, and BHhexol-CE were also identified at our sites but only in the lake surface sediments.

Aminotriol, as well as isomers of aminotriol and unsaturated aminotriol, was abundant in the water column SPM and surface sediment of the lakes (Fig. 2). Methylated and acylated versions of aminotriol were also identified but only in the surface sediment of the lakes. In contrast to a previous study in a hyper-euxinic and meromictic lake, we do not observe any diunsaturated aminotriols in the SPM or surface sediments at our study sites (O'Beirne et al., 2022).

Aminotetrol and aminopentol, compounds commonly associated with methanotrophic activity (Neunlist and Rohmer, 1985b; Zundel and Rohmer, 1985; Cvejic et al., 2000), were found in almost all of the water column SPM and surface sediments. However, acylated versions of aminopentol (C14:0, C15:0, and C16:0) were only observed in the surface sediments of Azul and Verde. Two isomers of methylcarbamate-aminotriol (MC-aminotriol) and MC-aminopentol, compounds also associated with methanotrophs, were identified in water column SPM and surface sediment after Rush et al. (2016). The recently described propenolamine-BHT, ethenolamine-BHT, ethenolamine-BHpentol, and ethenolamine-BHhexol (Fig. 3) were identified in our samples after Hopmans et al. (2021). In addition, we putatively identified unsaturated versions of ethenolamine-BHT (Appendix B4, Fig. B6) and ethenolamine-BHhexol (Fig. 3e) with the double bond likely occurring at the Δ6 and Δ11 positions, respectively, based on the retention times. Acylated versions of ethenolamine-BHhexol (C15:0, C16:0, and C17:0) were also observed in lake surface sediments.

In sediment from Lake Verde (Fig. 3, peaks j and m), we identified several unknown composite BHPs previously described in Hopmans et al. (2021) with an assigned elemental composition (AEC) of C36H64O4N+ (m/z 574.483) and C36H64O5N+ (m/z 590.483), but no structure was assigned. In addition, we identified the same compound but with an additional hydroxy moiety (Fig. 3, peak o). We propose that these novel composite BHPs are a series of N-formylated aminoBHPs: N-formylated aminotriol (peak j), N-formylated aminotetrol (peak m), and N-formylated aminopentol (peak o). The proposed structure for N-formylated aminopentol is shown in Fig. 3f with the diagnostic fragmentation indicated. In addition, we observe unsaturated versions of N-formylated aminotriol (Fig. 3b and Appendix B5, Fig. B7), N-formylated aminotetrol (based on the MS1 and retention time), and N-formylated aminopentol with the double bond likely occurring at the Δ6 position for N-formylated aminotriol and Δ11 position for N-formylated aminotetrol and N-formylated aminopentol. In the mass chromatogram for m/z 572.467 (Fig. 3b) from the Lake Verde sediment, we also observe a later-eluting peak (C36H62O4N+; Fig. 3b and d, peak l). Based on the MS2 spectrum, we propose that this is the same compound identified by Elling et al. (2022) in nitrite-oxidizing bacteria. This compound is likely a cyclized form of N-formylated aminotriol, and we putatively identify it as oxazinone-aminotriol (Appendix B6).

In addition to the discussed Δ6-ethenolamine-BHT, we observed a later-eluting peak at 21.88 min (Fig. 3a, peak c) in the mass chromatogram of m/z 586.483 (C37H64O4N+; Δ ppm 0.72). We propose that one of the functionalities is part of a cyclized structure and this BHP is a formylated aminotriol, where the formic acid is part of a cyclic structure and the terminal amino group is methylated. The proposed structure with key fragmentation is shown in Fig. 3c. We tentatively identify this compound as a dioxanone-methylaminotriol (Appendix B7).

Based on the NMDS visualization, the more complex BHP distributions in the surface sediments allow for these sampling sites to be placed into three lake types (Fig. 4): deep lakes (i.e., Azul, Verde, Funda, and Negra; circles), shallow lakes (i.e., Empadadas, São Jorge, Lomba; squares), and coastal lagoons (i.e., Cubres East and Cubres West; stars). Using ANOSIM we find a significant difference both between the sample sites (R = 0.89, p<0.01) and the sample site types (R = 0.72, p<0.001). The variability between our sites, particularly in the surface sediments, is reflected in the overall differences in BHP distributions. In Azul and Verde surface sediments collected near the shore and in the deepest part of the lake show comparable BHP distributions. We also observe comparable BHP distributions in sample replicates for lakes Azul, Empadadas, and Negra; thus we will only discuss the average BHP values for the replicate sites. In the water column NMDS plot (Fig. 5) we observe the bottom water samples cluster closer together, whereas the samples from the surface water and chemocline overlap. Using the ANOSIM test, we find no significant difference in BHP distributions between the surface water, chemocline, and bottom water of these lakes (R = 0.04, p=0.32) or between the water columns of different sites (R = 0.06, p=0.35).

The deep lakes are monomictic and undergo stratification from June–October, with hypoxic and even anoxic conditions occurring during the summer months in the hypolimnion (Gonçalves, 2008; Raposeiro et al., 2018). These lakes contain a high abundance of BHT, BHT-CE, and aminotriol, particularly in the surface waters and sediment samples (Fig. 2 and Tables A3–A4). Previous studies that identified BHPs in lake sediments also described a high abundance of BHT, BHT-CE, and aminotriol (Farrimond et al., 2000; Talbot and Farrimond, 2007). Similarly, BHP distributions in the water column of a hypereutrophic, meromictic lake, Mahoney Lake (Canada), are dominated by BHT, unsaturated BHT, aminotriol, and BHT-CE in the surface waters and BHT, diunsaturated aminotriol, and aminotriol in the surface sediments (O'Beirne et al., 2022). In the Azores, the deep lakes (circles) are further distinguished by the presence of amino-containing BHPs in the surface sediments (Fig. 2a) that cluster together in the NMDS plot (Fig. 4), including aminotetrol, aminopentol, MC-aminopentol (green), N-formylated aminotetrol, N-formylated aminopentol (gray), ethenolamine-BHpentol, and ethenolamine-BHhexol (gold). This is also reflected in the BHP distributions of the bottom water SPM samples collected in June 2018 (Fig. 6), in which we observe amino-containing BHPs (i.e., aminopentol, ethenolamine-BHhexol, aminopentol, N-formylated aminopentol) in samples collected below the oxycline. A higher abundance of penta- and hexa-functionalized BHPs, including aminopentol, was previously described in eutrophic lakes relative to lakes with low primary productivity (Farrimond et al., 2000; Talbot and Farrimond, 2007).

Although the shallower lakes might also undergo a period of stagnation, more regular mixing likely leads to oxygenation of the surface sediments. The shallow lakes are characterized by a higher abundance and diversity of nucleoside BHPs in both the surface sediment and water column SPM (Fig. 7). In addition, Me-inosylhopane, Me-N1-methylinosylhopane, and the early-eluting adenosylhopane type BHPs (brown) all cluster together near the shallow lakes, Empadadas, São Jorge, and Lomba (squares), in the NMDS plot (Fig. 4). Nucleoside BHPs (i.e., adenosylhopane, adenosylhopane type, and 2Me-adenosylhopane type) were previously reported in the sediment samples of two lakes, Loch Ness (Scotland) and Lake Nkunga (Kenya), that were associated with high levels of terrestrial input (Talbot and Farrimond, 2007). Loch Ness and Lake Nkunga also contained relatively high levels of BHT, aminotriol, BHT-CE, and even aminopentol (Talbot and Farrimond, 2007). Similar BHP distributions are observed in the sediments of Empadadas, São Jorge, and Lomba.

Finally, Cubres East and Cubres West are shallow, coastal lagoons that are influenced by both marine and freshwater inputs. Cubres East (salinity of 9.7 ppt) and Cubres West (salinity of 23.8 ppt) are defined by a high relative abundance of ethenolamine-BHT and propenolamine-BHT and by the presence of two methoxylated BHPs. In contrast to our freshwater sites, BHT-CE is mainly absent from the coastal lagoons (except for in the surface sediments of Cubres West). Similarly, in land-to-sea transects from the Yangtze (China) and Yenisei (Siberia) rivers, BHT-CE was more abundant in rivers, estuaries, and the coast relative to the open water (Zhu et al., 2011; De Jonge et al., 2016). Marine samples and fjord samples appear to predominantly consist of BHT and aminotriol (Farrimond et al., 2000; Zhu et al., 2011; Kusch et al., 2021a). In the estuary and coastal samples from the Yangtze River drainage basin, the BHP distributions consist of BHT, 2MeBHT, aminotriol, aminotetrol, aminopentol, nucleoside BHPs, and BHT-CE (Zhu et al., 2011). In general, we observe a similar distribution of BHPs in Cubres East and Cubres West.

Aminotriol and BHT were the most abundant BHPs at all of our sample sites. The presence of more diverse and abundant aminoBHPs in the deeper and seasonally stratified lakes allow us to distinguish them from the shallower lakes that contain a higher abundance of nucleoside BHPs. Finally, the lower diversity of BHPs and unique presence of methoxylated BHPs in Cubres East and Cubres West separates the marine influenced sites from the rest of the lakes in this study.

Previously, certain BHPs identified in MOB cultures were also detected in lake settings (Talbot et al., 2001; Talbot and Farrimond, 2007; Rush et al., 2016; Neunlist and Rohmer, 1985b, c; Cvejic et al., 2000; O'Beirne et al., 2022). However, neither methanotrophs nor lacustrine settings have been analyzed since recent methodological advancements have expanded BHP identification (Hopmans et al., 2021).

Here, we investigated the BHP composition of a lacustrine methanotroph–methylotroph (Methylobacter–Methylotenera) enrichment co-culture isolated from Lacamas Lake, WA, USA (van Grinsven et al., 2020). BHPs (i.e., aminotriol, 3Me-aminotriol, MC-aminotriol, aminotetrol, 3Me-aminotetrol, MC-aminotetrol, aminopentol, unsaturated aminopentol, 3Me-aminopentol, MC-aminopentol) have been previously detected in cultures of gammaproteobacterial type I MOB Methylobacter (Osborne, 2015; Rush et al., 2016). Our aim was to potentially identify novel BHPs from a lacustrine enrichment dominated by Methylobacter. This enrichment is also abundant in the methylotroph Methylotenera (van Grinsven et al., 2020). In order to rule out the synthesis of BHPs by this methylotroph, we investigated its genomic capacity to synthesize BHPs by searching for the key gene responsible for the formation of hopane and therefore BHP (i.e., squalene-hopene cyclase). A protein blast search (NCBI, National Center for Biotechnology Information) was performed with the sequence of the squalene-hopene cyclase gene of Bradyrhizobium japonicum as a query (accession no. WP_038942977.1), which did not lead to any hits in the genomes of Methylotenera species. In addition, we analyzed the BHP composition of Methylotenera mobilis, a strain closely related to the methylotroph also present in the co-culture, which was available in the DSMZ culture collection. Analysis of the M. mobilis biomass did not lead to the detection of BHPs. We acknowledge that other bacteria were also present in the enrichment culture (see van Grinsven et al., 2020); however, their BHP contribution is likely minor compared to the more abundant Methylobacter sp. present.

In the Methylobacter–Methylotenera co-culture we identified 22 BHPs (Table A4), with the most abundant being aminotriol (43 %) and aminopentol (43 %) (Fig. 8). Aminotetrol (6 %) was also observed, along with the unsaturated and acylated forms of aminotriol and aminopentol. Although present in low abundance, we identified MC-aminotriol, MC-aminopentol, ethenolamine-BHT, ethenolamine-BHhexol, N-formylated aminotriol, and N-formylated aminopentol (Fig. 8, Table A4). Low concentrations of nucleoside BHPs, i.e., adenosylhopane and 2Me-adenosylhopaneHG-diMe, were also present. Finally, we detected two novel BHPs (m/z 594; C35H64O6N+; Fig. 9), where the MS2 spectra show the consecutive loss of five hydroxyl moieties and an additional loss of 29 Da (CH3N). Based on the MS2 spectra and the elemental composition, we tentatively identify these compounds as aminohexol BHPs (Appendix B8). Notably, 3β-methylated BHPs are absent in the co-culture. Based on past culture studies, aminopentol and aminotetrol are associated with type I and type II MOB, respectively (Neunlist and Rohmer, 1985b, c; Cvejic et al., 2000; Talbot et al., 2001; van Winden et al., 2012). The high relative abundance of aminopentol relative to aminotetrol observed in the BHP distribution of the Methylobacter–Methylotenera co-culture analyzed in this study is similar to that of other type I MOB (Neunlist and Rohmer, 1985c; Talbot et al., 2001; van Winden et al., 2012; Rush et al., 2016; Kusch and Rush, 2022). The novel BHPs detected in the co-culture, however, are reported for the first time in this study. We conclude that the suite of BHPs identified in the co-culture, including the novel ethenolamine-BHPs, N-formylated aminoBHPs, and aminohexols, are attributed to the MOB Methylobacter present and thus can be potentially considered markers of aerobic methanotrophs in lake settings.

Aminotetrol is present in all of the Azorean samples, except for in the surface water of Negra (1 m depth), at 5 m depth in Azul, and in the water column and sediment of Cubres East (Fig. 2). Similarly, aminopentol is observed in the water column and sediment of Azul and Verde (except for at a depth of 5 m in the water column of Azul). In Funda and Negra, aminopentol only occurs in the deepest part of the water column and in the sediment samples. For Empadadas and Cubres East, aminopentol occurs both in the water column and the sediment, but it is only observed in the sediment sample from Cubres West. Similarly, aminopentol is found in both sediment samples from Lomba and São Jorge. Both aminopentol and aminotetrol are also produced in small amounts by sulfur-reducing bacteria (SRB; Blumenberg et al., 2006). As SRB also synthesize relatively higher amounts of aminotriol, ratios of aminopentol and aminotetrol to aminotriol have been used to determine the origin of the penta- and hexa-functionalized aminoBHPs (Blumenberg et al., 2006, 2009, 2012). The absence of anoxic conditions in the lake water column would rule out a major contribution of SRB-produced BHPs during the time of sampling. This is confirmed by the high aminopentol : aminotriol (water column of 0 (aminopentol absent) to 0.37 and sediment of 0.01 to 1.67) and aminotetrol : aminotriol (water column of 0 (aminotetrol absent) to 1.30 and sediment of 0 (aminotetrol absent) to 0.38) ratios in both water column and sediment samples. Recently two species of nitrite-oxidizing bacteria were shown to produce aminopentol in minor amounts under certain growth conditions (Elling et al., 2022), potentially contributing to the aminopentol observed in the surface waters of Azul, Empadadas, and Cubres East. However, in the oxygen-depleted bottom waters of Azul, Funda, and Negra, MOB are likely the primary producers of aminotetrol and aminopentol. A relatively higher abundance of aminopentol than aminotetrol in the water column and sediment samples would suggest a higher abundance of type I than type II MOB, in accordance with previous studies in both dimictic and meromictic lakes (Hanson and Hanson, 1996; Oswald et al., 2015, 2016; Guggenheim et al., 2020).

We observe unsaturated aminopentol (van Winden et al., 2012; Wagner et al., 2014) in the bottom water samples of lakes Funda and Negra (Fig. 6), which could be an adaptation by MOB to colder water temperatures (Bale et al., 2019). We also observe unsaturated aminotriol in the bottom water of Azul (15 m, 14 ∘C), Funda (29 m, 12 ∘C), and Negra (100 m, temperature data unavailable). C16:0N-acyl-aminopentol was reported in trace amounts near a terrestrial methane seep (Hopmans et al., 2021). C16:0N-acyl-aminopentol, as well as C14:0 and C15:0N-acyl-aminopentol, are present in sediment samples from Azul and Verde, whereas in Lomba and Empadadas we only observe C16:0N-acyl-aminopentol in the sediment (Fig. 2). In the NMDS plot of the surface sediments, the acylated aminopentols group closely with BHPs related to MOB, aminotetrol and aminopentol (Fig. 4). In addition, C14:0 and C16:0N-acyl-aminopentol are both detected in the Methylobacter–Methylotenera co-culture, confirming that MOB could be a potential source for these acyl-aminoBHPs in lakes Azul, Funda, Lomba, and Empadadas.

MC-aminotriol and MC-aminopentol were produced by the Methylobacter–Methylotenera co-culture (Fig. 8), albeit in low abundance. However, this is the first report of ethenolamine-BHT, ethenolamine-BHhexol, N-formylated aminotriol, and N-formylated aminopentol in culture. So far, ethenolamine-BHPs and N-formylated aminoBHPs were only identified near a terrestrial methane seep (Hopmans et al., 2021). Further environmental and culture studies are needed to identify whether these are robust biomarkers for methanotrophs. In all of our samples, except for the water column of Negra at 1 and 7 m depths (Fig. 2), we detected ethenolamine-BHPs, propenolamine, and/or N-formylated aminoBHPs (Hopmans et al., 2021). The widespread distribution of these compounds suggests they are derived from multiple producers. Although we cannot identify a direct source for these compounds, the NMDS analysis shows that ethenolamine-BHpentol, ethenolamine-BHhexol, acylated ethenolamine-BHhexols (C15:0, C16:0, C17:0), N-formylated aminotetrol, and N-formylated aminopentol cluster with MC-aminopentol, aminopentol, aminotetrol, and the acylated aminopentols near the surface sediments from Azul, Verde, Funda, and Negra (Fig. 4). In addition, ethenolamine-BHhexol and N-formylated aminopentol are detected in the bottom waters of Lake Funda (Fig. 6). Azul, Verde, Funda, and Negra all stratify during the summer months, and the bottom water in these lakes ranges from suboxic to anoxic conditions (Gonçalves et al., 2018). This would suggest that ethenolamine-BHpentol, ethenolamine-BHhexol, acylated ethenolamine-BHhexols, N-formylated aminotetrol, and N-formylated aminopentol are being produced under suboxic to anoxic conditions and/or could be associated with MOB.

BHPs with a 3β-methylation, particularly 3MeBHT and 3Me-aminotriol, are often linked to methane oxidation (Neunlist and Rohmer, 1985b; Zundel and Rohmer, 1985; Cvejic et al., 2000). However, 3β-methyl-BHPs are not exclusively produced by MOB (Zundel and Rohmer, 1985; Welander and Summons, 2012). BHPs with a 3β-methylation are also produced by other bacteria in the phyla Proteobacteria (classes Alpha-, Beta-, and Gammaproteobacteria), Cyanobacteria, Acidobacteriota, and Nitrospirota (Zundel and Rohmer, 1985; Sinninghe Damsté et al., 2017; Elling et al., 2022). In addition, 3β-methylation is crucial for bacterial cell survival in the late stationary phase and thus could have a broad taxonomic distribution (Welander and Summons, 2012). We observe 3MeBHT in all sediment samples except for Cubres East and Cubres West (Fig. 2); however, it was not detected in our water column samples. In contrast, 3Me-aminotriol only occurs in sediment samples from Lomba, Cubres West, São Jorge, Empadadas, Verde, and Azul. The absence of 3Me-BHPs in SPM samples suggests that at the time of sampling the bacteria responsible for 3β-methylation were not abundant in the water column. Future work will need to evaluate annual changes in both BHP distributions and bacterial communities in the water column to confirm the primary producers of 3Me-BHPs in lakes.

AminoBHPs with a methylcarbamate terminal group (MC-aminoBHPs) were proposed as biomarkers for type I MOB in marine settings based on culture samples (Rush et al., 2016), but, so far, the occurrence of MC-aminoBHPs in lakes has not been evaluated. MC-aminotriol has been observed in nitrite-oxidizing bacteria (Elling et al., 2022). The highest rates of methane oxidation in seasonally stratified lakes typically occurs at or directly below the chemocline (Rudd et al., 1976; Hanson, 1980; Harrits and Hanson, 1980; Oswald et al., 2015). In the Azorean lakes, MC-aminotriol was found in the oxycline and bottom water of both lakes Azul and Funda (Fig. 6). In Azul and Funda we observe MC-aminotriol near the oxycline, which could also be attributed to nitrite-oxidizing bacteria (Elling et al., 2022). We also observe MC-aminotriol in sediment samples from Verde, Negra, São Jorge, Lomba, Cubres East, and Empadadas. The co-occurrence of MC-aminopentol with 3MeBHT and/or 3Me-aminotriol in sediment samples from Verde, Negra, Lomba, and Empadadas could indicate a common MOB source. Both Lomba and Empadadas are relatively shallow lakes with a fully oxygenated water column, which means the primary site of methane oxidation occurs in the surface sediment and not in the water column (Hanson and Hanson, 1996). The different MC-aminoBHP distributions in the sediment samples could indicate that we were either not sampling during the peak period of methane oxidation in the lake, there was a shift in the MOB community, or methane oxidation was primarily occurring at the sediment surface.

Based on the distributions of aminoBHPs and MC-aminoBHPs, MOB are likely present, albeit in low abundance, at all Azorean sites. In the deeper lakes (i.e., Azul, Funda, and Negra), methane oxidation is likely occurring either at or below the chemocline, which would account for the diverse aminoBHPs observed in the water column (Fig. 6). Low δ13C values in bulk sediment samples from Lake Funda and primarily reducing conditions associated with the bottom water of the lake would suggest that methanogenesis is occurring in the bottom water and/or surface sediment of Lake Funda during the summer months (Richter et al., 2022). Similarly, proxy records suggest that eutrophication in Lake Azul contributed to hypoxic conditions in the bottom water (Raposeiro et al., 2017). In both lakes Funda and Azul, an increase in organic matter deposition combined with low-oxygen conditions in the bottom water during the summer months would support methane production and, therefore, a community of methane-oxidizing bacteria in the water column of these lakes. A similar distribution of BHPs associated with methane-oxidizing bacteria (i.e., aminopentol, unsaturated aminopentol, and aminotetrol) is found in high abundance in the methane-rich environments of floodplains in the Amazon basin and wetlands in the Congo River basin (Wagner et al., 2014; Spencer-Jones et al., 2015). In contrast, in the shallow lakes and the coastal lagoons methane oxidation might occur both in the water column and in the oxygenated sediment. Further molecular and culture studies, however, are needed to validate these conclusions.

A recent study suggests that some nucleoside BHPs may be in situ produced in the chemocline or under anoxic/euxinic conditions in marine water columns (Kusch et al., 2021b). To the best of our knowledge, no studies have investigated nucleoside BHP production in lacustrine environments. Nucleoside BHPs were detected in all water column and sediment samples except the water column of the lagoon Cubres West (salinity of 23.8 ppt; Fig. 7). The distribution of nucleoside BHPs in the sediment of Lake Empadadas is reflected in the BHP distribution of its surface water, suggesting that in shallow, well-mixed lakes nucleoside BHPs like adenosylhopane and N1-methylinosylhopane are either derived from production in the water column or surface water run-off (Cooke et al., 2008; Xu et al., 2009; Rethemeyer et al., 2010). In contrast, in Lake Funda we observe an increase in adenosylhopane, adensylhopaneHG-diMe, 2Me-adenosylhopaneHG-diMe, Me-adenosylhopaneHG-diMe, N1-methylinosylhopane, 2Me-N1-methylinosylhopane, and Me-N1-methylinosylhopane at the chemocline (6 m depth). Of these nucleoside BHPs, only adenosylhopane, adenosylhopaneHG-diMe, and 2Me-adenosylhopaneHG-diMe are present in the surface water (1 m) of Lake Funda. A pycnocline associated with the chemocline in the Azorean lakes could result in the accumulation of adenosylhopane, adenosylhopaneHG-diMe, and 2Me-adenosylhopaneHG-diMe; however, the lack of Me-adenosylhopaneHG-diMe, N1-methylinosylhopane, 2Me-N1-methylinosylhopane, and Me-N1-methylinosylhopane in the surface water of Lake Funda could also indicate in situ production of these compounds in the water column. Kusch et al. (2021b) found that adenosylhopane, N1-methylinosylhopane, and 2Me-N1-methylinosylhopane were produced within the chemocline and under anoxic or euxinic conditions in marine settings. In addition, adenosylhopane is separate from the rest of the nucleosides in the NMDS plot of only water column samples (Fig. 5), falling in between samples collected near the chemocline and the surface water. N1-methylinosylhopane is also observed in the bottom water of Lake Azul, in addition to adenosylhopaneHG-Me and adenosylhopaneHG-diMe, suggesting that these nucleosides are likely produced in situ. The primary producers of nucleoside BHPs, however, remain unconstrained and are likely associated with a wide range of bacteria. Adenosylhopane, in particular, is produced as an intermediate product in BHP synthesis (Bradley et al., 2010) and has the potential to be sourced from diverse bacteria, including purple non-sulfur bacteria (Neunlist and Rohmer, 1985a; Neunlist et al., 1988; Talbot et al., 2007), nitrifying bacteria (Seemann et al., 1999; Talbot et al., 2007; Elling et al., 2022), and nitrogen-fixing bacteria (Bravo et al., 2001; Talbot et al., 2007).

The ratio of nucleoside BHPs to BHT (Rsoil index, Eq. 1; Zhu et al., 2011) and the revised soil index for high latitudes (Rsoil′, Eq. 2; Doðrul Selver et al., 2012) have previously been used as an indication of soil input to riverine and marine environments (Cooke et al., 2009; Sáenz et al., 2011; Taylor and Harvey, 2011; Zhu et al., 2011; Doðrul Selver et al., 2012, 2015; De Jonge et al., 2016; Kusch et al., 2019). However, the recent indications that nucleoside BHPs may be in situ produced in marine (Kusch et al., 2021b) and lacustrine water columns (this study) could affect the application of the Rsoil proxy. Rsoil values of marine and lacustrine sediments from the tropics to Arctic range from 0.4–0.9 (Cooke et al., 2008; Pearson et al., 2009; Rethemeyer et al., 2010; Kusch et al., 2021a), and values between 0.5–0.8 have been employed as a tentative terrestrial end-member (Zhu et al., 2011). Here we revise the Rsoil index to exclude nucleoside BHPs that might be produced in the lake water column (i.e., adenosylhopane, early-eluting adenosylhopaneHG-Me, Me-adenosylhopaneHG-diMe, N1-methylinosylhopane, 2Me-N1-methylinosylhopane, Me-N1-methylinosylhopane). We call this new index Rsoil-lake. This ratio is significantly correlated with Rsoil (R = 0.94, p<0.0001) and Rsoil′ (R = 0.60, p<0.0001). Further, this ratio is positively correlated with TOC / TN (R = 0.81, p<0.005; Fig. C1), where TOC / TN values between 4 and 10 reflect organic matter derived from lake algae and values higher than 20 indicate an increased source of vascular land plants (Meyers, 2003). The low sediment TOC / TN values (7 to 11) and isotopically depleted δ13C values (-31 ‰ to -24 ‰; Fig. C2a) indicate predominantly in situ derived organic matter, which is in accordance with the high levels of primary productivity observed in all of the lakes (Cordeiro et al., 2020; Meyers, 2003). Similarly, low δ15N values (-1 ‰ to 2 ‰; Fig. C2b) are commonly associated with high levels of nitrogen fixation in accordance with the cyanobacterial blooms observed in the lakes during the summer months (Cordeiro et al., 2020; Meyers, 2003). The slightly more enriched δ13C values (-19 ‰ to -15 ‰) in Cubres East and Cubres West are typical of marine algae, also suggesting that the organic matter is mainly derived from in situ production (Meyers, 1994).

At all of the sites, we observe lower Rsoil-lake values in the water column than the sediment. As our water column SPM samples were collected in the dry season in the Azores (April–August; Hernández et al., 2016), we would anticipate low terrestrial inputs. In contrast, the BHPs detected in the sediments represent an integrated annual signal and likely include a larger input of terrestrial material during the wet season (September–March; Hernández et al., 2016). The Rsoil-lake values found in the sediments of the deeper lakes (i.e., Funda, Negra, and Azul) range from 0.17 to 0.32 and similarly have TOC / TN values between 8.1 and 8.3. The sediments of the shallow lakes all contain relatively higher Rsoil-lake values (i.e., Empadadas of 0.57, São Jorge of 0.69, and Lomba of 0.56) and higher TOC / TN values (ranging from 10 to 11.4), which could indicate slightly higher inputs of terrestrial organic matter, particularly during the wet season. In the coastal lagoons Cubres East and Cubres West, we observe low Rsoil-lake values both in the water column (Cubres East of 0.04 and Cubres West of 0) and sediment samples (Cubres East of 0.21 and Cubres West of 0.16). This could reflect fewer terrestrial inputs relative to the other lakes in this study. Low TOC / TN values (Cubres East of 7.08 and Cubres West of 7.22) would also indicate a primarily in situ source of organic matter (Meyers, 2003); however, further work is needed to verify this.

Our results highlight that caution should be used to ensure that the nucleoside BHPs included in the Rsoil proxy are not produced in situ. Our current dataset, the Rsoil-lake index, which excludes nucleoside BHPs that likely originate from the water column, appears to work for all of our Azorean sites; however additional work is needed to distinguish the primary source of nucleoside BHPs and to expand the application of Rsoil-lake. More detailed water column and catchment studies are needed to confirm whether these nucleoside BHPs are actually being produced within the lakes. Furthermore, similar comparison studies should be performed in marine settings.