Confounding effects of oxygen and temperature on the TEX₈₆ signature of marine Thaumarchaeota

Significance

Ammonia-oxidizing archaea (AOA) are among the most abundant microorganisms in the ocean. Apart from having a major influence on the nitrogen cycle, the glycerol dibiphytanyl glycerol tetraether (GDGT) membrane lipids of AOA are widely used to reconstruct past sea surface temperatures. We here provide compelling evidence that the composition of membrane lipids of marine AOA show strain-specific dependence on temperature. We also show that oxygen (O₂) concentration greatly influences membrane lipid composition, leading to significant increases in TEX₈₆-derived temperatures with increasing O₂ limitation. This finding challenges the convention that GDGT composition correlates solely with ocean temperature and necessitates a reassessment of archaeal lipid-based paleoclimate proxies, particularly when applied to environments where O₂ is depleted.

Abstract

Marine ammonia-oxidizing archaea (AOA) are among the most abundant of marine microorganisms, spanning nearly the entire water column of diverse oceanic provinces. Historical patterns of abundance are preserved in sediments in the form of their distinctive glycerol dibiphytanyl glycerol tetraether (GDGT) membrane lipids. The correlation between the composition of GDGTs in surface sediment and the overlying annual average sea surface temperature forms the basis for a paleotemperature proxy (TEX₈₆) that is used to reconstruct surface ocean temperature as far back as the Middle Jurassic. However, mounting evidence suggests that factors other than temperature could also play an important role in determining GDGT distributions. We here use a study set of four marine AOA isolates to demonstrate that these closely related strains generate different TEX₈₆–temperature relationships and that oxygen (O₂) concentration is at least as important as temperature in controlling TEX₈₆ values in culture. All of the four strains characterized showed a unique membrane compositional response to temperature, with TEX₈₆-inferred temperatures varying as much as 12 °C from the incubation temperatures. In addition, both linear and nonlinear TEX₈₆–temperature relationships were characteristic of individual strains. Increasing relative abundance of GDGT-2 and GDGT-3 with increasing O₂ limitation, at the expense of GDGT-1, led to significant elevations in TEX₈₆-derived temperature. Although the adaptive significance of GDGT compositional changes in response to both temperature and O₂ is unclear, this observation necessitates a reassessment of archaeal lipid-based paleotemperature proxies, particularly in records that span low-oxygen events or underlie oxygen minimum zones.

Marine ammonia-oxidizing archaea (AOA) (now assigned to the phylum Thaumarchaeota) are among the most ubiquitous and abundant organisms in the ocean, constituting up to 40% of microbial plankton in the meso- and bathypelagic zones (1–4). They are generally recognized as the main drivers of oceanic nitrification (5–7), are closely coupled with anammox organisms in oxygen minimum zones (OMZs) (8–10), and have been implicated as a source of the greater part of oceanic emissions of the ozone-depleting greenhouse gas nitrous oxide (11). Their wide habitat range suggests both high ecotypic diversity and adaptive capacity (12, 13).

The adaptive basis for their dominant role in the nitrogen cycle has in part been attributed to highly efficient systems of ammonia oxidation and carbon fixation, and a primarily copper-based respiratory system that reduces reliance on iron availability in the often iron-depleted marine environment (13–16). In addition, compositional regulation of their distinctive glycerol dibiphytanyl glycerol tetraether (GDGT) lipid membrane (SI Appendix, Fig. S1) is implicated in adaptation and acclimation to energy-limited environments (17). Relative to the bacterial membrane bilayer, the membrane-spanning lipids of archaea are less permeable to ions and protons (18, 19). Lower permeability is suggested to reduce maintenance energy costs, an important adaptive feature of extreme oligophiles such as the AOA. Thus, growth temperature-dependent modulation of membrane composition is likely associated with maintenance of appropriate permeability, as well as other membrane functions (18).

The influence of temperature on membrane composition has been the major focus of studies of the environmental distribution of archaeal membrane lipids in the present and for interpreting lipids preserved in the sedimentary record (20). In particular, the correlation between sea surface temperature (SST) and the cyclopentane ring distribution of GDGTs in a sample set of globally distributed core top sediments is the basis for a widely applied paleotemperature proxy, TEX₈₆ (TetraEther indeX of lipids with 86 carbon atoms). The TEX₈₆ proxy has been used to reconstruct surface ocean temperature as far back as the Middle Jurassic (21, 22). However, the extent to which temperature is the causative agent behind the correlation has not fully been examined. Moreover, it is not evident how a group of organisms that live at depths below the upper-photic zone and are the dominant prokaryote of the abyss can record SST via a physiological response.

Interpretation of TEX₈₆ records that deviate from expectation are usually interpreted from the point of view that temperature is still the main underlying influence on TEX₈₆. That is, disagreement with other proxy records, unreasonably large swings in TEX₈₆ over short time periods, or warm biases in semienclosed basins (Mediterranean and Red Sea) have been attributed to changes in circulation, seasonal timing of production, selective export to sediments, or to autochthonous archaeal populations having slightly different temperature responses (20, 23–25). Large discrepancies between in situ and TEX₈₆-derived temperatures from suspended particulate matter (SPM) samples, particularly from low-oxygen (O₂) environments, have been observed in many regions (25–28), but O₂ has not explicitly been suggested as a cause of these discrepancies. Instead, it is thought that some local environments select for ecotypes having slightly different TEX₈₆–temperature relationships.

TEX₈₆ reconstructions of the distant past, when oceanic conditions could be quite different from today, present the greatest challenge to interpretation. For example, inferred Cretaceous SSTs are higher than physically plausible for the ocean (29). Thus, considerable efforts have been made to develop and apply new TEX₈₆ equations suitable for high-temperature environments (${\text{TEX}}_{86}^{\text{H}}$), low-temperature environments (${\text{TEX}}_{86}^{\text{L}}$), local systems (TEX₈₆′), the marine water column, and mesocosms (24, 30–33).

The availability of a number of pure cultures of marine AOA (12) now serves to test strain-specific response to temperature, as well as the influence of ecophysiological factors other than temperature, on lipid composition. Here we grew pure cultures of four marine AOA strains to evaluate the influence of growth conditions on core lipid composition. Despite relatively close phylogenetic relationships, these isolates are physiologically distinct ecotypes recovered from dimly lit deep waters, the nitrite maximum near the euphotic zone, and near-surface sediment (12). The influence of temperature and dissolved O₂ (DO) on GDGT composition and TEX₈₆ values was examined by independently varying these two important environmental variables. These experiments revealed significant strain-specific variation in temperature-associated GDGT distribution. More striking was the observation that in addition to temperature, the DO concentration had a profound impact on membrane composition, skewing TEX₈₆-derived temperatures by more than 20 °C at a single growth temperature. These findings necessitate reinterpretation and possible reformulation of this important paleoclimate proxy.

Results

Early stationary phase cultures were used to investigate total GDGT content and the distribution of GDGTs in different AOA strains [strains are described in detail in Qin et al. (12)]. Cell pellet hydrolysis ensured the complete recovery of all cellular lipids and removed polar head groups from intact GDGTs so they could be analyzed as core GDGTs (34). Different strains of marine AOA contained variable amounts of total GDGTs per cell (calculated based on cell counts and lipid abundance) and each strain contained a similar amount of total lipids per cell regardless of the growth temperature. Strain HCE1 contained 1.47 ± 0.096 fg GDGTs per cell (at 15–20 °C, n = 6), which was less than half of the total GDGTs recovered from PS0 cells (3.01 ± 0.065 fg per cell at 25 °C, n = 3) (SI Appendix, Table S1). Cells of strains HCA1 and SCM1 contained similar lipid content (2.04 ± 0.21 fg per cell at 15–30 °C for HCA1, n = 12 and 2.08 ± 0.12 fg per cell at 20–30 °C for SCM1, n = 9; SI Appendix, Table S1). These values were within the range reported for the cellular lipid contents of SCM1 and natural marine communities dominated by AOA (0.1–8.5 fg per cell) (17, 35). GDGTs of SCM1 accounted for ∼10% of dry cell weight (20 fg per cell) (13). This value is at the higher end of previous estimates for methanogenic and thermoacidophilic archaea that contain 2.5–10% total lipids on a dry weight basis (36–38).

All isolates showed substantially different TEX₈₆ values when grown at 25 °C (0.56 ± 0.02 for HCA1, 0.51 ± 0.01 for SCM1, 0.42 ± 0.01 for HCE1, and 0.38 ± 0.01 for PS0) (SI Appendix, Table S1). TEX₈₆ values calculated using the calibration curve of Kim et al. (39) correspond to TEX₈₆-derived temperatures ranging from 10.5 to 20.7 °C, all below their actual growth temperature of 25 °C (Fig. 1). To further test the strain-specific relationship between TEX₈₆ values and growth temperatures, SCM1 (optimum growth temperature at 32 °C), HCA1, and HCE1 (optimum growth temperatures at 25 °C) were incubated at temperatures ranging from 15 to 35 °C, 10–30 °C and 10–25 °C, respectively (see SI Appendix, Fig. S2 and Table S1 for experimental details and results at different growth temperatures). The relationship between TEX₈₆ values and growth temperatures of SCM1 was a linear function, showing a general pattern of increased TEX₈₆ values with increasing temperature (SI Appendix, Fig. S3A; see SI Appendix, Fig. S4A and Table S2 for additional details of relative abundance of GDGTs):

$$TE{X}_{86}=0.015T+0.21\left(r^2=0.85\right).$$ [1]

A linear relationship was also found between the TEX₈₆-derived temperature reconstructed using the Kim et al. (39) calibration equation and actual growth temperature of SCM1 (Fig. 2A). These calculated TEX₈₆ temperatures were 0.13–7.28 °C lower than incubation temperatures, explaining the difference between the global calibration line ($T=-10.78+56.2\times TE{X}_{86}$) and the culture correlation line ($T=-14+66.7\times TE{X}_{86}$). In contrast, there was no linear dependence of growth temperature on the TEX₈₆ value of strains HCA1 and HCE1. Instead, the data of HCA1 (Eq. 2) and HCE1 (Eq. 3) fitted best to polynomial curves, reaching the maximum TEX₈₆ values at 20 °C (SI Appendix, Fig. S3 B and C; see SI Appendix, Fig. S4 B and C and Table S2 for additional details of relative abundance of GDGTs):

$$TE{X}_{86}=-0.0006T^2+0.023T+0.33\left(r^2=0.94\right)$$ [2]

$$TE{X}_{86}=-0.0017T^2+0.054T+0.11\left(r^2=0.89\right).$$ [3]

Similarly, second-order polynomials resulted in good fits to TEX₈₆-derived temperatures of HCA1 and HCE1 vs. incubation temperature (Fig. 2 B and C). The temperatures derived from TEX₈₆ values of HCA1 and HCE1 reflected the in situ temperatures, within calibration error, at 20 °C (TEX₈₆ temperatures of HCA1 and HCE1 were 21.1 °C and 19.3 °C, respectively). However, large discrepancies between TEX₈₆-derived temperatures and actual growth temperatures (up to 12 °C) were observed at low- and high-growth temperatures of HCA1 and HCE1. All isolates had different ring index values (a common metric of GDGT cyclization) when grown at 25 °C (SI Appendix, Table S2) (40). The ring index values of strains SCM1, HCA1, and HCE1 all showed a linear increase in cyclization of total GDGTs with increasing temperature (Fig. 3).

Fig. 1.

Reconstructed TEX₈₆-derived temperatures of four marine AOA isolates incubated at 25 °C (dashed line). Each bar represents the average of measurements from triplicate incubations and at least duplicate injections. Error bars are the SD among average values of triplicate incubations.

Fig. 2.

Correlation of TEX₈₆-derived temperature (filled circles) with growth temperature of strains SCM1 (A), HCA1 (B), and HCE1 (C). Filled triangles represent the temperature dependence of the growth of strains SCM1 (A), HCA1 (B), and HCE1 (C) (in terms of specific growth rates per day). All plotted data represent the average of measurements from triplicate incubations. Error bars represent the SD of triplicates.

Fig. 3.

Correlations of ring index values with growth temperatures of strains SCM1 (15–35 °C), HCA1 (10–30 °C), and HCE1 (10–25 °C). The linear regression lines of SCM1 (solid line), HCA1 (dashed line), and HCE1 (solid line) are $y=0.076x+1.40\hspace{0.17em}\left(r^2=0.96\right)$, $y=0.082x+0.77\hspace{0.17em}\left(r^2=0.94\right)$, and $y=0.058x+1.31\hspace{0.17em}\left(r^2=0.95\right)$, respectively. Error bars represent the SD of data from triplicate cultures (some error bars are too small to be visible in the figure).

In addition to the expected adaptation to changes in temperature, the observation of abundant and transcriptionally active AOA populations near the oxic–anoxic boundary of OMZs also suggested a capacity to grow in highly O₂-depleted waters (9, 41). Both SCM1 and PS0 were shown to be capable of growing at very low O₂ (0.1–21% of initial headspace O₂ in sealed culture bottles; see SI Appendix, Fig. S5 and Table S3 for experimental details and results at different percentages of O₂). All ammonia (10 μmol) was completely oxidized to nitrite when O₂ was in stoichiometric excess (0.5–21% initial headspace O₂). In accordance with the established reaction stoichiometry of 1 mol NH₃ oxidized per 1.5 mol O₂ consumed (13), ∼15–1,238 μmol of O₂ remained at the end of the incubation (SI Appendix, Fig. S5 and Table S3). In contrast, all headspace O₂ was consumed at 0.1% and 0.2% initial O₂, resulting in residual ammonia and a lower level of nitrite production (∼5.1 and ∼8.2 μmol nitrite, respectively), also consistent with predicted reaction stoichiometry (∼4.0 and ∼7.8 μmol nitrite for 0.1% and 0.2% O₂, respectively) (SI Appendix, Fig. S5 and Table S3). Ammonia oxidation was coupled to growth at all O₂ concentrations. The cell densities of early stationary phase cultures of SCM1 and PS0 were 13–27 times and 25–29 times higher than initial cell densities, respectively (SI Appendix, Table S3).

Because marine AOA demonstrate the ability to grow at very low O₂ in the laboratory and the environment, we next asked whether changes in membrane composition might also be associated with the growth of AOA at low O₂ concentrations. The relative abundances of GDGT-2 and GDGT-3 (SI Appendix, Figs. S1 and S6) increased as initial headspace O₂ concentration decreased (SI Appendix, Fig. S7 and Table S4). Notably, the abundance of GDGT-2 in SCM1 and PS0 showed a significant and continuous increase from 12.4% and 7.7% at 21% O₂ to 16.6% and 27.0% at 0.1% O₂, respectively. In contrast, a 5.2% and 5.7% decrease in the relative abundance of GDGT-1 was observed between 21% O₂ and 0.1% O₂ treatments for SCM1 and PS0, respectively (SI Appendix, Fig. S7 and Table S4). These changes along with changes in GDGT-0 and crenarchaeol abundance were reflected in the increase in ring index values as the O₂ concentration decreased with the exception of the lowest O₂ treatments (SI Appendix, Table S4). Crenarchaeol regioisomer (cren′) (SI Appendix, Fig. S1) was detected in low abundance (<1.3%) in all samples. The relative decrease in GDGT-1, and increase in GDGT-2 and GDGT-3, at low O₂ concentrations resulted in higher calculated TEX₈₆ temperatures in O₂-deficient cultures. Nearly 12 °C of TEX₈₆-derived temperature elevation reconstructed using the Kim et al. (39) calibration equation was observed for strain SCM1 from 21% O₂ to 0.1% O₂ (Fig. 4A). Likewise, the TEX₈₆-derived temperatures of PS0 were 11.4 °C below incubation temperature (26 °C) at 21% O₂, but changed to 9.9 °C above incubation temperature at 0.1% O₂ (Fig. 4B). Although some change in DO is expected with growth, the comparable growth kinetics observed at the higher, but nonlimiting, O₂ concentrations suggest that O₂ concentration primarily influences lipid composition (SI Appendix, Fig. S5).

Fig. 4.

Reconstructed TEX₈₆ temperatures for the total cellular lipids of strains SCM1 (A) and PS0 (B) showing the exponential TEX₈₆-derived temperature increase from 21% O₂ to 0.1% O₂. Dashed lines represent the constant incubation temperature of strains SCM1 (A) and PS0 (B) at 30 °C and 26 °C, respectively. Error bars represent the SD of the mean of triplicate cultures.

Discussion

Although the TEX₈₆ paleothermometer has been applied in diverse depositional settings across more than 100 million y of geological history, the TEX₈₆–SST relationship is strictly a correlation among environmental data. The precise cause of the global correlation and observed deviations within this correlation dataset remain poorly understood (20). Sedimentary GDGTs likely originate from a variety of GDGT-producing organisms, all of which are members of the archaeal domain. Planktonic Thaumarchaeota are considered to be a major source of sedimentary GDGTs, but their populations are often most abundant in subsurface waters, near the base of the photic zone, leaving unknown the causative factor driving the correlations between sedimentary GDGT composition and growth temperature (likely subsurface) (20). Likewise, there is also some contribution from pelagic Euryarchaeota and sediment-dwelling Archaea (42, 43).

Before the availability of pure cultures, mesocosms for North Sea water preincubated at high (27 °C) and low (13 °C) temperatures were used to test the validity of the TEX₈₆ proxy (33). The lipid composition was determined following shifts in incubation temperature. Incubations of the high-temperature mesocosm (preincubated at 27 °C) between 10 °C and 35 °C generated TEX₈₆ values having the same slope as, but different intercepts from, the global calibration. Even more surprisingly, no linear dependence of growth temperature vs. TEX₈₆ values could be found for the low-temperature mesocosm (preincubated at 13 °C) when incubated at the other temperatures (5–35 °C). Instead, a plot of TEX₈₆ vs. temperature had a concave down shape (33). However, when all data were combined, the nonlinearity was less apparent (33).

Our use of pure cultures of marine AOA allowed us to independently examine the influence of growth temperature and O₂ on TEX₈₆, and variation in response among different AOA strains. Notably, despite the close phylogenetic relationships, all four marine AOA isolates differed markedly in their TEX₈₆ values at the same growth temperature, and strains SCM1, HCA1, and HCE1 exhibited distinctly different temperature–TEX₈₆ relationships. The tropical (strain SCM1) and temperate (strains HCA1 and HCE1) marine AOA isolates were initially enriched at 28 °C and 15 °C, respectively (2, 12), temperatures comparable to those (27 °C and 13 °C) applied in the mesocosm study (33). Interestingly, the linear and curved temperature–TEX₈₆ trends for strains SCM1, HCA1, and HCE1 are very similar to those observed in the high- and low-temperature mesocosms, respectively (33). The large offsets between TEX₈₆-inferred temperatures and growth temperatures of these isolates (up to 12 °C) are consistent with the mesocosm experiments. The TEX₈₆ variation among strains at the same growth temperature suggests this index is reporting both variation in SST and variation in biosynthetic response of distinct AOA ecotypes. In addition, the different temperature optima of our AOA isolates suggest that temperature may select for ecotypes having significantly different temperature–TEX₈₆ relationships.

Significant increase in the relative abundance of crenarchaeol with increasing growth temperature was observed with all examined AOA isolates (strains SCM1, HCA1, and HCE1). This suggests the preferential synthesis of crenarchaeol relative to GDGT-1–3 and cren′ could be a common feature for marine AOA in warmer environments, as suggested by the 40 °C temperature optimum of crenarchaeol synthesis (20, 44). Therefore, the ring index equation may be a more suitable proxy for temperature than the TEX₈₆ equation. Indeed, although no linear dependence of growth temperature and TEX₈₆ values was found for some isolates, there was a linear relationship between ring index values and growth temperatures for all of the marine AOA examined in this study. Notably, the temperate isolates (HCA1 and HCE1), having a growth optimum near 25 °C, displayed a similar degree of cyclization of total GDGTs at a given temperature. In contrast, strain SCM1 produced a membrane with a higher ring index than temperate isolates at all temperatures, possibly related to its adaptation to higher-temperature environments. However, further studies are needed to more fully evaluate the utility of ring index values as a complementary temperature proxy.

Alternative factors known to influence membrane composition of archaea include pH and salinity. The GDGT composition of the cultured thermophilic AOA Nitrosocaldus yellowstonii and members of Crenarchaeota vary with pH (20, 45–47). However, for the highly buffered ocean, pH variability is relatively small, and thus would not be expected to significantly bias reconstructed SST. In addition, although salinity has been shown to exert substantial control on the lipid composition of halophilic archaea (48), no significant variation of GDGT distributions was found in a marine mesocosm study where salinity was varied (33). In contrast, the influence of O₂ concentration on lipid composition is relatively unexplored. Thus, the striking response of the marine AOA isolates to reduced O₂ was surprising, demonstrating an increase in TEX₈₆-derived temperatures of up to 21 °C by varying O₂ concentration alone.

O₂ concentrations are known to span a wide range of values with depth in the ocean and in sediment. Suboxia and anoxia prevail in some modern enclosed basins and in restricted waterways present early in Earth history such as the Atlantic Ocean during the Oceanic Anoxic Events (OAE) of the Cretaceous and Jurassic periods (49). Presently, the ocean is also home to several large OMZs where subsurface waters experience varying degrees of suboxia and anoxia. Thus, our observation of a major controlling influence of O₂ on TEX₈₆ values is important for interpretation of environments and sedimentary records influenced by low O₂ or anoxia. Notably, anomalously warm TEX₈₆ patterns of SPM samples have been reported over a wide range of suboxic settings, such as in the permanent OMZ of the Arabian Sea and the Eastern Tropical North Pacific Ocean, and seasonally O₂-deficient regions in coastal upwelling areas (27, 28, 50). In addition, TEX₈₆ values of sediment recovered from the summit of the Murray Ridge seamount that extends into the OMZ of the Arabian Sea were higher than those of sediment recovered from adjacent deeper locations well below the OMZ, and the reconstructed TEX₈₆ temperatures from sediment cores within the OMZ were more than 5 °C higher than the mean annual SST (51).

Although surface waters are not O₂-deficient in these areas, the AOA are of higher abundance in the oxycline and upper OMZ where waters can be oxygen-deficient (9, 10). The AOA living under O₂ limitation would likely be the major contributor to the sedimentary GDGTs collected within the Arabian Sea OMZ (51). Elevated TEX₈₆ values were mainly the result of the relative increase in GDGT-2 that we have now shown to increase under low O₂ concentrations with pure cultures of marine AOA (25, 27, 28). Thus, our study clearly implicates a lipid biosynthetic response as a causative factor in producing anomalously high TEX₈₆ values within low-oxygen environments. In addition, if reduced O₂ concentrations select for a different marine archaeal community, then an altered species composition could also influence the composition of GDGTs preserved in the sediments. However, resolving the relative importance of adaptation vs. selection on the TEX₈₆ record in response to “greenhouse” climates will require a more complete understanding of thaumarchaeal ecophysiology.

Our laboratory findings may also provide an explanation for some unexpected TEX₈₆ sedimentary records. The Jurassic and Cretaceous Periods are known as relatively warm intervals in Earth’s history. One consequence of warmer water is lower DO, because O₂ is less soluble in warm water than in cold water. TEX₈₆-derived temperature records of this time period routinely exceed modern temperatures at equivalent latitudes (22, 29, 52–54). Our results suggest decreased DO concentrations of warm ancient seawater would raise reconstructed temperatures to values somewhat higher than the actual SST. During particular intervals of Earth’s history, such as the many OAEs and Paleocene/Eocene thermal maximum (PETM), widespread dysoxic or even anoxic conditions impinged on the photic zone, as indicated by a sedimentary biomarker (isorenieratane) for photic zone anoxia (31, 49, 53, 55, 56). Although some increase in temperature during OAEs and the PETM is consistent with paleoclimate model simulations and planktonic foraminiferal δ¹⁸O-based SST reconstructions, TEX₈₆ values (as high as 0.95) and TEX₈₆-derived temperatures (up to 43 °C) observed during these low-oxygen events seem anomalously high (31, 53–58). However, because δ¹⁸O-based temperature estimates are not available throughout the black shales of an OAE, there is no independent validation of the exceptionally warm TEX₈₆-inferred temperatures for these climate periods (31, 59, 60).

Excursions in O₂ concentration are also associated with short historic periods of pronounced cooling (53, 54, 61). For example, the TEX₈₆ paleotemperature proxy suggests a cooling of up to 12 °C during the Plenus Cold Event (PCE) of OAE-2 (54). However, the PCE is also marked by a drop in isorenieratane concentrations in the sedimentary record, indicative of reoxygenation of a previously anoxic photic zone (53). Because the steep decrease in TEX₈₆ values within OAEs is difficult to explain with a drastic drop of SSTs only, reoxygenation would offer an explanation for large swings in TEX₈₆ values observed during these events. Thus, together our data now implicate DO as an additional causative factor of anomalously high or low TEX₈₆ values and inferred temperatures.

We recognize that more experiments with AOA strains representing a broader range of ecotypes are necessary to further explore environmental influences on GDGT composition and how these factors relate to physiology. There also remains the question of the biophysical significance of thaumarchaeal membrane compositional changes associated with reduced O₂ concentration. It has been suggested that an increase in cyclopentane rings reduces proton permeability as a result of denser membrane packing (19, 62). This adaptive response may allow AOA to cope with the energy stress of low-O₂ environments, such as the boundary of the OMZs in today’s oceans and during past periods of low O₂ availability, such as the OAE of the mid-Cretaceous. However, apart from the questions of physiological significance highlighted by these studies, the available data do offer an argument in favor of viewing the global calibration datasets of TEX₈₆ index as a compilation of local calibrations (20) and point to the need to reevaluate inferences of paleoclimate from geologic records where conditions differ significantly from the modern and from unique local environments such as those underlying OMZs.

Finally, in addition to the significance of our observations for reinterpreting past climates and oceanic conditions, the adaptive response of these AOA may have relevance to predicting the response of oceanic populations affected by climate change. Predicted warming, deoxygenation, and acidification of the global ocean will certainly have an impact on major biogeochemical systems (63–65). Thus, better resolving the physiological response of marine AOA to these changes is of global significance. Phenotypic plasticity that enhances the adaptive capacity of AOA may mitigate some of these responses with important implications for both adaptation to and reconstruction of change in present and past marine environments, respectively.

Methods

Culture Maintenance and Experimental Setup.

All materials and methods for marine AOA pure culture maintenance and temperature and O₂ growth experiments are described in detail in SI Appendix, Materials and Methods.

Lipid Extraction and Analysis.

Lipids were extracted from 0.22-µm Durapore membrane filters (Millipore Co.) containing early stationary phase cells and analyzed using atmospheric pressure chemical ionization on an Agilent 1100 Series liquid chromatograph coupled to an Agilent ion trap mass spectrometer. For details, see SI Appendix, Materials and Methods.

Calculation of TEX₈₆ Index, TEX₈₆-Derived Temperature, and Ring Index.

The TEX₈₆ values were calculated based on the relative abundances of GDGT-1, GDGT-2, GDGT-3, and Cren′ using the respective peak areas (21):

$$TE{X}_{86}=\frac{\left[GDGT-2\right]+\left[GDGT-3\right]+\left[Cre{n}^{\prime }\right]}{\left[GDGT-1\right]+\left[GDGT-2\right]+\left[GDGT-3\right]+\left[Cre{n}^{\prime }\right]}.$$ [4]

The reconstructed TEX₈₆ temperatures were calculated on the basis of the core-top linear calibration of Kim et al. (39):

$$SST=-10.78+56.2\times TE{X}_{86}\left(r^2=0.94\right).$$ [5]

The ring index was calculated according to Pearson et al. (40):

$$\text{Ring}\hspace{0.17em}\text{index}=\frac{\left[\text{GDGT}-1\right]+2\times \left[GDGT-2\right]+3\times \left[GDGT-3\right]+5\times \left[Cren+Cre{n}^{\prime }\right]}{\left[GDGT-0\right]+\left[GDGT-1\right]+\left[GDGT-2\right]+\left[GDGT-3\right]+\left[Cren+Cre{n}^{\prime }\right]}.$$ [6]