Seasonal Succession and Temperature Response Pattern of a Microbial Community in the Yellow Sea Cold Water Mass

ABSTRACT

Explaining the temporal dynamics of marine microorganisms is critical for predicting their changing pattern under environmental disturbances. Although the effect of temperature on microbial seasonality has been widely studied, the phylogenetic structure of the temperature response pattern and the extent to which temperature shift leads to disruptive community changes are still unclear. Here, we explored the microbial seasonal dynamics in the Yellow Sea Cold Water Mass (YSCWM) that occurs in summer and disappears in winter and tested the temperature thresholds and phylogenetic coherence in response to temperature change. The existence of YSCWM generates strong temperature gradients in summer and confers little temperature change during seasonal transition, thus representing a unique intermediate state. The microbial community of YSCWM is more similar to that in the previous YSCWM in winter than that outside YSCWM. Temperature alone explains >50% of the community variation, suggesting that a temperature shift can induce a nearly seasonality-level community variance in summer. Persistence of most previous winter YSCWM inhabitants in YSCWM leads to conservation in predicted functional potentials and cooccurrence patterns, indicating a decisive role of temperature in maintaining functionality. Evaluation of the temperature threshold reveals that a small temperature change can lead to significant community turnover, with most taxa negatively responding to an elevation in temperature. The temperature response pattern is phylogenetically structured, and closely related taxa show an incohesive response. Our study provides novel insights into microbial seasonality and into how marine microorganisms respond to temperature fluctuations.

IMPORTANCE Microbial seasonality is driven by a set of covarying factors including temperature. There is still a lack of understanding of the details of the phylogenetic structure and susceptibility of microbial communities in response to temperature variation. Through examination of the microbial community in a seasonally occurring summer cold water mass, which experiences little temperature change during seasonal transition, we show here that the cold water mass leads to nearly seasonality-level variations in community composition and predicted functional profile in summer. Moreover, massive community turnover occurs within a small temperature shift, with most taxa decreasing in abundance in response to increased temperature, and contrasting response patterns are observed between phylogenetically closely related taxa. These results suggest temperature as the fundamental factor over other covarying factors in structuring microbial seasonality, providing important insights into the variation mode of the microbial community under temperature disturbances.

INTRODUCTION

Microorganisms are ubiquitously distributed in the ocean and play key roles in driving biogeochemical cycling (1, 2). They show intricate interactions among each other and with their environments (3), determining the direction of energy and nutrient flow (4, 5). Exploring the dynamics of marine microorganisms under fluctuating physicochemical conditions is pivotal for understanding future ecosystem changes (3, 6). Despite the occurrence of irregular events such as storms, eddies, and typhoons, changes in microbial communities in the marine environment show rhythmicity and follow predictable patterns. For example, long-term studies of marine surface water have consistently revealed seasonal variation and annual recurrence in temporal community dynamics (7–10). Such seasonal variability has stimulated broad scientific interest aiming for a predictive understanding of how marine microbial communities will respond to environmental disturbance, particularly to the change of temperature (11).

Seasonality in temporal dynamics of marine microbial communities is reflected by elevation in diversity from summer to winter in the global temperate ocean (12, 13). Compositional shifts have also been observed, such as the summer dominance of Cyanobacteria and SAR11 and the winter prevalence of Thaumarchaeota and some Gammaproteobacteria, despite regional differences in the dominance of specific taxa (14). These changing patterns are regulated by both abiotic effects and interspecific interactions, including temperature, day length (light), seasonal phytoplankton blooms, etc. (9, 15–19). However, it is not clear whether these factors play a combined role or exert influence alone (20, 21). Their covariation, as well as the effect of stochastic events (22), makes it a challenge to determine key factors and tease apart their respective effects.

Temperature has been identified as a key factor underpinning temporal microbial dynamics, as indicated by several long-term studies lasting up to a decade (23–26). However, due to similar seasonal trends, the relative importance of temperature and day length remains controversial. Temperature has been reported as the major factor governing the seasonal transition of microorganisms in a time-series study from the U.S. East Coast (27), whereas day length is thought more important in a study targeting the western English Channel (9). This discrepancy may be attributed to the different ranges of temperature gradient between studies. Indeed, comparison among ocean time-series sites has shown that seasonality in microbial community is more pronounced in high-latitude (e.g., Bermuda Atlantic time series) than in low-latitude (e.g., Hawaii ocean time series) areas (3). These findings suggest that temperature is the major factor structuring microbial seasonality in the temperate ocean. Additionally, temperature is found to play a determinant role in shaping the microbial community structure across the global surface ocean (28), further demonstrating the role of temperature from a spatial perspective. These previous observations raise an interesting question of how microbial communities will develop if temperature remains constant during seasonal transition. Moreover, despite the above-described knowledge, much less is known about the phylogenetic structure of the temperature response pattern and the extent to which temperature variation leads to disruptive community changes. Such information, however, is important for predicting the effect of temperature disturbance on marine ecosystems.

Here, we tackle these knowledge gaps by investigating the seasonal shift of prokaryotic communities in the Yellow Sea (YS) of China, with a specific focus on the Yellow Sea Cold Water Mass (YSCWM; Fig. 1). YSCWM is a seasonal event that occurs in summer and disappears in winter; it is formed locally with the trapping of previous winter waters in the YS central trough (29). In winter, the joint effect of surface cooling and wind stress causes a thorough mixing of the water column, which is characterized by low temperature and high salinity (30). During the transition from winter to summer, surface heating and enhanced river runoff decrease the density (due to the increase of temperature and decline of salinity) of the YS surface water, creating a strong thermocline at a depth of ~10 to 22 m (31), which inhibits vertical mixing in the water column. The water bodies below the thermocline thus escape surface heating and maintain the cold property of winter (32, 33). The formation of a cold water mass in summer generates strong temperature gradients that may induce significant community variations. Meanwhile, the consistently low temperature from winter to YSCWM represents a unique scenario of microbial seasonality with a weak change of temperature compared to other seasonal factors. We hypothesize that the microbial community in YSCWM is more similar to that in the previous winter samples than in summer samples outside YSCWM, which is mainly determined by water temperature. Testing this hypothesis is important to gain new insights into the mechanisms underpinning the temporal microbial dynamics in the temperate ocean.

FIG 1.

Sampling map of the Yellow Sea of China. (A) Sites marked by circle indicate the availability of both summer and winter samples, whereas those marked by square indicate only summer samples. (B) The summer samples were clustered according to geolocations and water temperature reported to indicate the occurrence of Yellow Sea Cold Water Mass (YSCWM). All the surface samples are outside YSCWM. The grouping pattern of middle and bottom samples is shown in panel B. The summer samples inside (yellow), at the margin of (pink) and outside (black) YSCWM are marked with different colors. For sites with more than one color, different depths are grouped differently according to water temperature and geolocation (see details in Table S1). Numbers in parenthesis indicate the number of depth layers in each group. (C) A three-dimensional structure of YSCWM with the temperature contour generated along three transections (yellow lines). The stations marked in blue represent the 34 sites for microbial analysis, while those in dark gray represent stations from the same cruise used only for generation of the temperature contour. BS, Bohai Sea; NYS, north Yellow Sea; SYS, south Yellow Sea; ECS, East China Sea; YE, Yellow River estuary.

We examined the seasonal succession of microbial communities both within and outside YSCWM and traced the winter-to-summer community variation with YSCWM as a unique transitional state. We also provided an assessment of the community-level temperature threshold range and identified the phylogenetic pattern in response to temperature shift. Our study will deepen the understanding of the relationship between seasonal change in environmental factors and microbial communities and provide insights into the impact of temperature disturbance on marine microorganisms.

RESULTS

The prokaryotic community displays a stronger seasonal than spatial variation.

A total of 161 seawater samples were collected at different depths (see Table S1 in the supplemental material) from 34 sites in the YS, covering two seasons (summer and winter; Fig. 1A). To compare the prokaryotic community in different types of samples, 16S rRNA gene amplicon sequencing was performed, yielding 26,341 sequences per sample and 9,199 amplicon sequence variants (ASVs) across all samples. The diversity (Shannon and Chao 1) and evenness of the prokaryotic community show higher values in winter than in summer (P < 0.001) and display an increasing trend from surface to bottom layers (Fig. S1; Table S1). In contrast to alpha diversity, bacterial abundance derived from quantitative PCR shows an inverse seasonal variation trend, with higher values in summer than in winter (P < 0.001; Fig. S2). Archaea are less abundant than bacteria in general and show no significant change in abundance between seasons except at the surface water, where archaea are more abundant in winter.

Ordination and similarity analyses indicate significant differences in prokaryotic community composition among areas, depths, and seasons (Fig. 2A; Table S2). In both seasons, there are clear community variations between the north and south part of the YS, which is reflected by significant distance-decay patterns at all sampling depths (Fig. S3). Depth-related partition is evident in summer but is unseen in winter. Despite this spatial dissimilarity, nonmetric multidimensional scaling analysis (NMDS) shows a more pronounced community variation between seasons than between areas and depths. Temperature is identified as the most important factor in driving such community shifts, followed by phosphate, depth, etc. (Fig. 2A).

FIG 2.

Comparison of prokaryotic community composition and alpha diversity in different seawater samples and the temperature effect. (A) Nonmetric multidimensional scaling analysis of samples across depths and seasons with environmental variables fitted to the plot. (B) Nonmetric multidimensional scaling analysis of summer samples located within the Yellow Sea Cold Water Mass (S_in), at its margin (S_mar), and outside it (S_out) and their corresponding previous samples in winter. The water temperature of each sample was mapped onto the plot. (C) Pairwise Spearman’s rank correlation between Bray-Curtis dissimilarities and differences in temperature across all samples. (D) Shannon, Chao I, and evenness indices in the different habitats. W_in, winter samples formerly inside YSCWM; W_mar, winter samples formerly at the margin of YSCWM; W_out, winter samples formerly outside YSCWM.

High community similarity between YSCWM and the previous winter samples driven by temperature.

To distinguish the community variability generated by YSCWM, the summer samples representing YSCWM are designated according to geolocations and water temperature (<10°C) characteristic of its occurrence (Fig. 1B and C) (32). The former YSCWM samples in winter are also designated for parallel comparisons. Environmental characterization shows lower temperature and higher salinity and nutrient levels in the YSCWM and winter samples than in the summer samples outside the mass (Fig. S4). In terms of community compositions, the summer samples located inside (n = 28), at the margin of (n = 19), and outside (n = 46) YSCWM are significantly separated from each other (Fig. 2B; Table S3). In winter, in contrast, the differences are less distinguishable when comparing the former YSCWM samples with those formerly at the margin of and outside YSCWM. More importantly, the YSCWM community shows a higher similarity to that in former YSCWM in winter than that outside YSCWM (Fig. 2B; Fig. S5), with relatively lower abundance of Alphaproteobacteria, Cyanobacteria, and Bacteroidota but higher abundance of Nitrososphaeria, Thermoplasmatota, and Marinimicrobia (Fig. S6). These changing patterns are also seen in a principal-coordinate analysis (Fig. S7) and supported by the permutational multivariate analysis of variance (PERMANOVA) analysis (Table S3).

As temperature is the major factor governing the overall distribution patterns, we mapped temperature onto the ordination plot. The results show a clear covariation between temperature and community compositions (Fig. 2B). Meanwhile, there is a significant increase in community dissimilarity with increasing difference of temperature (rho = 0.696; Fig. 2C). To identify the contribution of temperature to community variation relative to other environmental factors, multiple regression on distance matrices (MRM) was used. The whole set of measured environmental factors together with water depth and geographic distance explain ~60% of the community turnover in the MRM model (P < 0.001), with temperature being the most important explanatory variable (as evidenced by the partial regression coefficient) compared to PO₄^3–, NH₄⁺, pH, and geographic distance, which also have significant contributions. The MRM model with temperature as the sole variable provides an explanatory power of 57.4%, which accounts for 98.1% of the explanatory power provided by all variables. Water depth explains a minor proportion (2.3%) of the community variation according to the MRM model fitted using all samples. However, as YSCWM mainly occurs at the lower (middle and bottom) water layers in summer, variation partition analysis (VPA) and the partial Mantel test were performed on only the summer samples to address the relative role of depth and temperature (Fig. S8). Both the analyses reveal a higher pure effect of temperature (T|D) than that of depth (D|T), although the VPA shows a high joint effect of these two variables.

A null model analysis was performed to explore mechanisms underpinning the observed geographic pattern. The results show a greater role of determinism relative to stochasticity in the community assembly, with the highest explanatory power provided by homogeneous selection compared to other ecological processes (Fig. S9). Homogeneous selection is found to be the only process governing the community variation in YSCWM and its previous winter samples and explains a higher proportion of community variation in these cold waters than in warm waters.

YSCWM preserves most of the previous winter microbes and provides a unique perspective in ascertaining microbial seasonality.

Having observed a high community similarity between YSCWM and its previous waterbody in winter, we compared enriched and depleted ASVs between seasons both within and outside YSCWM to detail the community changing process over time. For simplicity of comparison and to highlight the difference generated by the cold water mass, samples at the margin of YSCWM were not included in these analyses. About 9.4% of ASVs were differentially distributed between YSCWM and its previous winter samples, a proportion lower than that between summer and winter samples outside the mass (14.9%). The summer samples either within or outside YSCWM possess more depleted rather than enriched ASVs compared with their counterparts in winter. Inside YSCWM, however, the number of seasonally depleted ASVs is more than two times lower than that outside YSCWM (Fig. 3A). A further analysis of the top abundant ASVs (median relative abundance of >0.05% in at least one habitat) shows the persistence of most dominant previous winter inhabitants in YSCWM, but the degree of ASV uniqueness is higher in winter than in summer samples outside the mass (Fig. 3B and C). The overall decrease in ASV relative abundance from winter to summer is in concert with the decreasing trend of the α-diversity indices (Fig. 2D). However, these changing processes are decelerated by the occurrence of YSCWM.

FIG 3.

Changing pattern of ASVs between habitats and the phylogenetic relatedness. (A) Differently distributed ASVs based on the Wilcoxon rank sum test. (B and C) Rank abundance curve showing overlapping of the top abundant ASVs with a median relative abundance of >0.05% in at least one habitat. The relative abundance was squarely transformed. In pairwise habitats, pink ASVs are shared between them, while blue ASVs are unique to each environment. The total number of the abundant ASVs analyzed in each habitat and the number of ASVs unique to each habitat in pairwise comparison are indicated by the bar plot. (D) Nearest taxon index (NTI) in different habitats. (E) The pairwise Spearman’s rank correlation between NTI and temperature.

To examine changes in the extent of phylogenetic clustering in the microbial community during winter-to-summer transition, we calculated the sample-specific nearest taxon index (NTI). The positive NTI values indicate that the communities are phylogenetically more clustered than expected by chance. Moreover, a decrease in NTI values from cold to warm conditions (Fig. 3D) and a negative correlation between temperature and the NTI values (Fig. 3E) are observed.

Conservation of microbial association patterns and predicted functional profiles from previous winter samples to YSCWM.

To explore the ecological implications involved in this unique community transition process, we investigated variations in microbial cooccurrence relationships (Fig. 4A; Fig. S10) and predicted functional potential (Fig. 4B to D). Positive correlations dominate the cooccurrence network in all habitats (67.9 to 93.4%). The network topology reveals similar cooccurrence patterns in YSCWM and its previous winter samples and contrasting patterns between summer and winter samples outside the mass (Fig. 4A; Fig. S10). Lower values of average path length, diameter, modularity, and betweenness centralization and higher values of mean node degree and density are observed in cold than in warm waters.

FIG 4.

Cooccurrence patterns and predicted functional profiles of the microbial community. (A) Network-level topological features in networks from different habitats. (B) Nonmetric multidimensional scaling analysis based on the functional profiles annotated with Tax4Fun2. (C) R² values of the PERMANOVA analysis showing pairwise comparison of the compositional and functional dissimilarities. (D) Differently distributed functional pathways based on the Wilcoxon rank sum test.

Microbial functional potentials were annotated with Tax4Fun2 (34). As expected, the ordination analysis of the predicted functional profiles shows a high similarity among samples from cold water, which are clearly separated from those from warm water (Fig. 4B). It is noticeable that the similarity in functional profiles between YSCWM and its previous winter samples is more pronounced than that in community compositions (Fig. 4C; Table S4). From cold to warm conditions, a general tendency of reduction in functional pathways is observed (Fig. 4D).

Adaption to temperature: a narrow threshold range and phylogenetically structured pattern.

To examine to what extent temperature shift leads to disruptive community turnover and how individual taxa respond to the change in temperature, we evaluated the temperature threshold values using TITAN2 (35). We identified 42.0% and 21.0% of ASVs negatively and positively responding to the temperature increase, respectively. The community-level threshold/changing point is 11.8°C for negative responders (z–) and 16.4°C for positive responders (z+) (Fig. 5A).

FIG 5.

Pattern of microbial response to change in seawater temperature. (A) Community-level sums of taxon-specific indicator scores along the temperature gradient. (B) Mantel correlogram showing the relationship between pairwise phylogenetic (genetic) distance and dissimilarity in temperature response. (C) The number of microbial classes showing positive and negative response to temperature. (D) For classes containing members showing both positive and negative responses to temperature change, the log abundance ratio of positive to negative responders was calculated. Pink and blue indicate a higher abundance of positive and relative responders, respectively.

We then examined whether the microbial response to temperature change is related to phylogeny. Fritz and Purvis’s D test reveals phylogenetically nonrandom responding patterns in both the positive (D = 0.877, P < 0.05) and negative (D = 0.887 P < 0.05) directions. Furthermore, a correlogram analysis reveals the magnitude of response in relation to phylogenetic distance (Fig. 5B). Interestingly, at relatively small genetic distances, negative and/or no correlations are observed between the pairwise similarity in temperature response and genetic distance of ASVs. With increasing genetic distance of 22.8 to 27.2%, positive correlations are observed. After this, the correlations become negative again.

We find a higher number of negative than positive responders to the increase of temperature (the temperature threshold for each taxon at the class level is shown in Fig. S11). A total of 35 classes (plus orders affiliated with Bacteroidia and Cyanobacteriia) are unique to the negative responders, which mainly belong to Dehalococcoidia (phyla Chloroflexi), Dadabacteriia (Dadabacteria), the Pla3 lineage (Planctomycetota), Nitrospinia (Nitrospinota), and the BD2-11 terrestrial group (Gemmatimonadota). In contrast, the only two classes exclusively found in positive responders are represented by Phormidesmiales of Cyanobacteria and Leptospirae of Spirochaetota (Fig. 5C; Fig. S12). A total of 37 classes (plus orders affiliated with Bacteroidia and Cyanobacteriia) are composed of both negative and positive responders (Fig. 5D). Of them, ~65% show a relatively higher abundance in the negative than in the positive groups, including Nitrososphaeria (mainly Marine Group I) and Thermoplasmata (mainly Marine Group II). Lineages typically abundant in the marine surface, e.g., proteobacterial classes and Flavobacteriales of Bacteroidia contain negative and positive responders at almost equal abundance. Overall, although the response of microorganisms to temperature is phylogenetically structured, they are broadly distributed in different taxonomical groups.

DISCUSSION

Our results expand the current view of microbial seasonality by examining the variation of the microbial community in a seasonally occurring cold water mass (YSCWM) in the temperate coastal ocean. The existence of YSCWM is found to induce a clear community separation in summer resembling that derived from seasonality (winter versus summer). The relative similarity in community composition and association pattern and a more conserved predicted functional profile between YSCWM and the previous winter samples suggest that the cold water mass significantly retards the winter-to-summer transitional process, supporting a determinant role of temperature in governing seasonal microbial dynamics. More importantly, we evaluated ranges of the temperature threshold, suggesting that a minor temperature shift is likely to cause disruptive community alternations. The temperature-induced community change is phylogenetically structured and is featured by a substantial number of negative responders, hinting at a changing trend of the marine ecosystem upon temperature disturbances.

There is an inverse trend between microbial abundance and diversity during seasonal transition. Compared with winter, the summer seawater is populated by a more abundant but less diverse community. Previous studies have also reported the increase of prokaryotic abundance in temperate surface oceans during summer (36, 37), which may relate to elevation in alga-derived organic nutrients. The lower species richness and evenness, however, indicate that the increased carbon source may have benefited the growth of only a few taxa, mostly likely opportunistic copiotrophs. This is supported by a recent observation of a link between ecological strategy and temporal dynamics of bacterioplankton communities (38). In winter, the observed higher diversity is in line with many previous reports even at a global scale (12), suggesting that species loss is a general pattern of winter-to-summer transition in the temperate ocean. Additionally, the present study demonstrates through NTI values that seasonal switching preferentially occurs between closely related species. A similar result has also been obtained in a time-serial study at the Pivers Island Coastal Observatory (PICO) site using the minimum entropy decomposition method (27), indicating adaptive divergence between phylogenetically closely related taxa during seasonal dynamics.

Notably, with an intermediate level of diversity ranging in the middle between winter and summer, YSCWM acts as a unique environmental niche and may provide a novel perspective of microbial seasonality. YSCWM is a cold water mass in the Yellow Sea of China that occurs in summer and disappears in winter (32, 33). It features a relatively small change of temperature (1.67 ± 1.18°C in this study) in the same waterbody from winter to summer, thus representing a shelter for cold-adapted taxa from the change of their habitats to warmer conditions. Indeed, we find a higher community similarity between YSCWM and its previous winter samples than that between summer and winter samples outside the mass. YSCWM preserves most of the previous winter-dominant prokaryotes. This indicates that YSCWM retards the seasonal changing process, highlighting the important role of temperature in governing community structure. Temperature typically covaries with day length and sunlight during seasonal switching, making its relative importance in explaining microbial seasonal dynamics controversial (7, 9). However, seawater in YSCWM retains the cold property of winter and meanwhile undergoes normal changes of other seasonal factors, leading to asynchronicity in these typically covarying factors and providing an opportunity for assessing the independent effect of temperature. The MRM model shows that temperature explains over half of the community variation, similar to many previous studies documenting a key effect of temperature (24, 25, 39). This means that temperature is likely the major contributor to the seasonality-level community variation in summer, and may have a more important role in determining microbial seasonality than other seasonally changing forces.

However, some other aspects may also contribute to the observed community variability, attenuating the effect of temperature. First, since YSCWM mainly occurs in the subsurface (middle and bottom) water of the studied area, the effect of temperature may be confounded by that exerted by water depth. To resolve this dilemma, VPA and the partial Mantel test were performed, which revealed a higher pure effect of temperature than water depth. Second, vertical stratification, as an important cause of formation of YSCWM, inhibits vertical water mixing and movement of microorganisms between water layers (31). Thus, the contribution of restricted species emigration and immigration to a distinct community in YSCWM from that outside the mass could not be ruled out. However, the bacterial abundance in the middle and bottom layers representing YSCWM is similar to that outside (mainly surface; Fig. S2), indicating bacterial growth fueled by the supply of surface-derived organic substrates to deep layers. The enhanced primary production in summer may facilitate downward transportation of organic matter and microbes by sinking particles (40). Indeed, the relative abundance of some heterotrophic groups such as Flavobacteriaceae and Alteromonadaceae is similar in summer samples either within or outside YSCWM, with higher values than those in winter samples (Fig. S6). This to a certain extent may contribute to homogeneity of the summer community and to the community differences between YSCWM and its previous waterbody. Our results suggest that temperature is an important factor in maintaining community composition but does not necessarily control microbial abundance.

In addition to community compositions, interspecific interactions and predicted functional profiles also show high consistency between YSCWM and its previous winter samples. A network with increased interconnectivity in cold waters may be explained by the higher microbial diversity compared with the warm environment and by the metabolic dependency of oligotrophic taxa that are always auxotrophic for surviving under resource-depleted conditions (41, 42). Exchange of metabolites such as sugars and amino acids has been suggested as a major driver of species cooccurrence (43), yet, considering that YSCWM mainly represents the subsurface waters, our results are different from the previous report indicating a more connected microeukaryotic community at the surface than at a 30-m depth attributed to the highly variable environmental conditions (44). The possible explanation of these contrasting results is that the low temperature exceeds other abiotic factors in constraining microbial interactions. As network connectivity is a potential positive proxy of community stability (45), it may be inferred that warming would result in a more vulnerable microbial community that is readily disrupted. Caution is needed when interpreting these results, as cooccurrence patterns may not indicate true interactions (22).

Predicting functional profiles from 16S rRNA genes can help us gain insights into the functional capability of microbial communities, although the accuracy of this method has not been well evaluated (22). In this study, consistent functional profiles are observed between YSCWM and its previous winter samples, with the degree of similarity being higher than that of the compositional profiles. Similarly, there have been reports of less change in microbial functional structure under high taxonomic variabilities (46), supporting the idea of functional redundancy (47). The higher functional consistency observed here may reflect the temperature similarity, as environmental conditions have been found to exert more influence on microbial functional traits than on taxonomic compositions across the global ocean (48). Overall, these findings suggest that low temperature sustains high interassociations and drives a more cohesive response of functional traits in the coastal microbial community.

To examine to what extent temperature shift leads to a disruptive community turnover, we evaluated the thresholds of community change for temperature. Environmental threshold has recently been established for microbial communities in soil and lake systems to infer the community-level change points for specific environmental factors (49–52). To our knowledge, we report the first evaluation of temperature threshold for microbial community in the temperate coastal ocean. The result indicates that cold-preferring taxa initiate massive turnover at around 12.0°C during temperature elevation, whereas warmth-preferring taxa experience severe disruption upon a temperature decrease to <16.5°C. Proximity of the lower (11.8°C) and upper (16.4°C) thresholds indicates a narrow temperature niche breadth (4.6°C). This value is slightly lower than a threshold range of 5.7 to 6.7°C for soil fungal communities in response to mean annual temperature (53), indicating that aquatic prokaryotic communities may be more sensitive to temperature fluctuation. However, Wan et al. (52) reported that bacterioplankton communities in a eutrophic lake displayed temperature threshold ranges of 0 to 2.9°C. This disparity may be explained by the different temperature gradients covered: 20.9 to 25.2°C and 1.9 to 27.9°C in Wan et al. (52) and the present study, respectively, indicating dependency of the threshold analysis on the scale of environmental gradients. The large fluctuating magnitude and gradual change of temperature in our samples may have facilitated a more accurate assessment of the threshold values.

In fact, our result of a narrow threshold range for temperature is supported by the observation of clear seasonal community variations under a temperature gradient of ~10°C (from ~7 to 18°C) (9). Wang et al. (54) further demonstrated that warming by only a 3°C increase in temperature could significantly alter the coastal microbial community composition under a 5-day incubation in 20-L carboys. These observations jointly imply a high sensitivity of marine microorganisms to temperature elevation. Beyond this information, we show that a temperature increase would eliminate taxonomically diverse taxa, with the most sharply decreasing their abundance at ~12°C. As this thermal condition does not reach the upper temperature limit for growth of psychrophiles (55), we hypothesize that such an decline in abundance is likely due to the excessive growth of a few copiotrophic taxa, which may result in reduced living spaces and unfavorable interspecies interactions for oligotrophs representing most of the marine microbes (56). Supporting this, Yung et al. (57) has reported significantly increased growth rates of Vibrio at above ~12°C. Vibrio, indeed, shows a higher abundance in summer than winter (data not shown). Thus, it is likely that marine microbial communities, as a whole, are more adapted to resource depletion in the cold environment than to the unsuitable situation under warm conditions. Although the estimated threshold value may need further experimental validations, these findings enhance our understanding of the temperature-derived disturbance for microbial populations.

Accumulating evidence has suggested a linkage between phylogenetic information and microbial response to environmental change (58, 59). This is also the case in the present study, in which a phylogenetically structured temperature response for marine microorganisms is observed, with more pronounced phylogenetic clustering under low-temperature than high-temperature conditions. However, in contrast to the frequently observed positive correlations between pairwise environmental response and phylogeny at relatively small genetic distances (60, 61), indicative of phylogenetic conservation, we show negative and/or negligible relationships for temperature response at short genetic distances, reflecting the unique pattern of adaptation to temperature. Indeed, closely related marine prokaryotes have been reported to show different temperature preferences and occupy different seasonal niches (27, 62–64). The results suggest that the evolution of adaptive divergence among relatives is a key strategy for a microbial community to resist the seasonal temperature shift. This is further supported by our observation of various taxonomical classes containing both negative and positive temperature responders and a recent finding revealing a lack of coherent seasonal response at broader taxonomic levels, such as the class level (62). Although a class-level coherence may not exist, we do observe positive correlations at intermediate genetic distances. However, due to the short read length, the genetic distances calculated cannot inform accurate taxonomy, which needs further exploration of a more detailed relationship between temperature response and taxonomic levels, partially via full-length 16S amplicon sequencing. Finally, most of the taxonomic classes show a negative response, emphasizing the harmful effect of temperature increase on marine microbial ecology.

Conclusion. This study provides a novel scenario of microbial seasonality by examining a winter-to-summer community variation accompanied by little change of temperature. We present solid evidence of a temperature-determined higher similarity in compositional and predicted functional profiles between YSCMW and its previous winter water than that between YSCMW and water outside it. This indicates that a shift in temperature alone is able to cause seasonality-level community variations, with increased temperature resulting in a decrease of taxonomic diversity and community stability. The microbial community exhibits a narrow threshold range for temperature, and close relatives have distinct temperature preferences reflecting an evolutionary strategy to cope with temperature variations. The present study emphasizes the effect of temperature over other covarying factors on microbial seasonality and consolidates the susceptibility of temperate seawater-inhabiting microorganisms to shifts of temperature, in particular, temperature elevation that is unfavorable for most taxa. The temperature response pattern facilitates a predictive understanding of how global warming will impact microorganisms in the marine environment.

MATERIALS AND METHODS

Sampling and environmental characterization.

The Yellow Sea (YS) is a temperate marginal sea of China. Seawater samples in the YS were collected during two cruises of the research vessel (R/V) Dongfanghong 2 using a rosette water sampler. A total of 93 samples were taken during the summer cruise in August to September 2015, and 68 samples were taken during the winter cruise in January 2016. The sampling stations in summer covered both YSCWM and non-YSCWM areas (Fig. 1B). Two or three depths were collected at each site, representing surface (~3 m), middle (according to water depth), and bottom (1 to 2 m above sediment) layers. Seawater sample (1 L) was prefiltered through 3-μm-pore-size filters, and cells were collected on 0.22-μm-pore-size polycarbonate filters (Millipore Corporation, Billerica, MA, USA). Filters were frozen in liquid nitrogen on board and stored at −80°C in the laboratory until nucleic acid extraction. Water depth (2 to 81 m), temperature (1.9 to 27.9°C), and salinity (29.5 to 33.3 PSU) were recorded with a Seabird 911 conductivity-temperature-depth system (Fig. S4). Dissolved oxygen and pH values were measured following previously described methods and have been reported in detail by Zhai (65). The samples from this study have dissolved oxygen concentrations of between 3.25 and 11.27 mg L⁻¹ and pH values between 7.85 and 8.39. Dissolved inorganic nutrients were analyzed using an AA3 autoanalyzer system with the concentrations of dissolved inorganic nitrogen, phosphate, and silicate in the range of 0.01 to 18.32 μmol L⁻¹, 0.01 to 1.11 μmol L⁻¹, and 1.13 to 21.12 μmol L⁻¹, respectively.

DNA extraction and quantitative PCR.

Extraction of total DNA was carried out using a combination of mechanical and chemical cell breaking and phenol-chloroform-isoamyl extraction according to the procedure described previously by Liu et al. (66). DNA quality and concentration were measured using a NanoDrop 2000 spectrophotometer. Bacterial and archaeal abundance were assessed by SYBR green I-based quantitative PCR (qPCR) using the primer sets 338F/518R (67) and 967F/1060R (61), respectively. Triplicate amplifications were performed for each sample in a 20-μL reaction system under the conditions of an initial denaturation at 95°C for 30 s, 40 cycles of 95°C for 5 s, 53°C (bacteria) or 50°C (archaea) for 30 s, and 72°C for 30 s, and a final extension at 72°C for 5 min. Then, 10-fold serially diluted linear plasmids containing a single copy of 16S rRNA gene were amplified for generation of a standard curve. For bacteria, amplification efficiency was 98.3% and R² was >0.999; for archaea, amplification efficiency was 99.8% and R² was >0.999.

Amplicon sequencing and processing.

Primer 515F/806R (68) targeting the V4 hypervariable regions of the prokaryotic 16S rRNA gene was used for gene amplification. This primer set may induce biases during the amplification step, including an underestimation of SAR11 (69, 70), one of the most abundant clades in the surface ocean. The abundance of SAR11 ranges from 0.1 to 20.1% across all samples and displays clear seasonal variations, with higher values in summer than in winter (P < 0.001). This trend is similar to the previously reported seasonal variation in this clade (71). Therefore, the primer set used here should provide reliable patterns of microbial seasonal dynamics. Amplification was performed in a 20-μL reaction system with the following parameters: 95°C for 5 min, followed by 35 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 45 s, with a final extension at 72°C for 10 min. For each sample, triplicate amplifications were performed and were mixed for library preparation. Barcode sequences were ligated to the primer during synthesis before amplification. The amplicon library was paired-end sequenced on an Illumina MiSeq PE250 platform at Biozeron Co., Ltd., Shanghai, China, according to the standard protocols. Sequences were first demultiplexed using the barcode sequences and then trimmed to remove those of low average quality score (<20), of short length (<50 bp), and with any mismatch to barcodes and at most two mismatches to primers. Amplicon sequence variants (ASVs) were generated, and the phylogenetic affiliation of each ASV was assigned against the SILVA database (release 138; https://www.arb-silva.de/) using a confidence threshold of 80% following the pipeline of QIIME 2 (72).

Diversity, environmental correlation, and null model analysis.

Alpha diversity indices (Shannon and Chao I) and Pielou’s evenness were calculated after sequence rarefication, which was conducted to equalize sampling effort. Comparison of community composition between samples was conducted in the package vegan v2.5-7 (73) of the R software v3.6.1 (74) using Bray-Curtis dissimilarity-based nonmetric multidimensional scaling analysis (NMDS) and principal-coordinate analysis (PCoA). Dissimilarity tests were performed using permutational multivariate analysis of variance (PERMANOVA). Environmental variables were fitted to the NMDS ordination plot using the envfit function in the vegan package (73). To detect the distance-decay relationship, Spearman’s rank correlations between pairwise Bray-Curtis dissimilarities and geographic distances (deduced from the latitude and longitude coordinates using the R package geosphere v1.5-10 [75]) were calculated. To investigate the impact of temperature on the microbial distribution pattern, we first mapped the temperature parameter to the NMDS ordination plot and calculated the Spearman’s rank correlation between pairwise Bray-Curtis dissimilarities and differences in temperature. Then, multiple regression analysis on distance matrices (MRM) was conducted using the R package ecodist v2.0.7 (76) to address the influence of temperature relative to other environmental and spatial factors and their joined effect. MRM provides the advantage of investigating the relationship between a response (community) distance matrix and any number of explanatory (environmental and spatial) distance matrices (77). All explanatory distance matrices were standardized with the R package MuMIn v1.43.17 (78) in order to evaluate their relative importance. Collinearity between environmental variables was detected by calculating their Spearman’s rank correlations (when between-variable rho was >0.7, one of the variables was removed).

Variation partitioning analysis (VPA) and the partial Mantel test were conducted to address the relative role of temperature and water depth and their combined effects. The pure effects of temperature (T|D) and water depth (D|T) were tested for significance using the permutation test. These analyses were performed in the R package vegan v2.5-7 (73).

Stegan’s null model analysis (79, 80) was implemented to further investigate the ecological mechanisms underpinning microbial community assembly. This method measures the β-nearest taxon index (βNTI) and the Bray-Curtis-based Raup-Crick metric (RC_bray) as phylogenetic and taxonomic β-diversity metrics, respectively. When βNTI is >2 (heterogeneous selection) or <−2 (homogeneous selection), deterministic processes are prevalent. When |βNTI| is <2 and stochastic processes prevail, RC_Bray is calculated with RC_Bray of >0.95, |RCBray| of <0.95, and RC_Bray of <–0.95 indicating dispersal limitation, drift, and homogeneous dispersal, respectively.

ASV enrichment analysis and phylogenetic relatedness.

ASV enrichment analysis was performed using the Wilcoxon rank sum test based on the average relative abundance of ASVs in different sampling groups (81). P values were adjusted using the FDR method with a threshold value of 0.05. The NTI index was used to link phylogenetic information to the changes of ASVs between habitats, with higher values indicating a higher degree of phylogenetic clustering. The value of NTI is equivalent to −1 times the standardized effect size of mean nearest-taxon distance, which was calculated using the ses.mntd function in the R package picante v1.8 (82) under the taxa.lables null model with 999 randomizations.

Functional prediction and network construction.

In addition to taxonomic composition, the implication of YSCWM on microbial ecology was explored in terms of functional profiles and intertaxa associations. We predicted the functional profiles of microorganisms based on their 16S rRNA gene sequences using Tax4Fun2 (34). Inference of microbial function from the 16S rRNA gene has provided great insights into prokaryotic functional capabilities in various natural environments, although it cannot accurately reflect the functional potential as derived from metagenomics.

Cooccurrence networks were constructed using the R packages igraph v1.2.4.1 (83) and Hmisc v4.3-0 (84). To increase accuracy, the top 500 most abundant ASVs in each group were included. A valid correlation was only considered if the between-ASV Spearman’s correlation coefficient was >|0.7| and the P value was <0.01 (false-discovery rate [FDR] adjusted). Number of nodes and edges, mean node degree, clustering coefficient, average path length, modularity, density, diameter, betweenness centralization, and degree centralization were calculated to interpret the network topological feature. Gephi (https://gephi.org) was used for network visualization.

Threshold of temperature adaption and response pattern to temperature change.

Threshold Indicator Taxa Analysis (TITAN2 v2.4.1) (35) was applied to identify the response of individual taxa to the temperature gradient. To meet the criterion of TITAN2, 2,698 ASVs with a sequence number of >10 and occurring in at least five samples were included. Both positive and negative responders were identified. TITAN2 also allows detection of the community-level changing point (threshold) at which multiple taxa show steep changes in abundance. The filtered sums of indicator species scores (fsum[z]) for positive (z+) and negative (z–) responders were considered the upper and lower thresholds, respectively, for temperature adaption. To examine whether response to temperature is phylogenetically structured, Fritz and Purvis’s D test and a correlogram correlation between pairwise phylogenetic distance and similarity in temperature response were performed.

Data availability.

The raw reads have been deposited in the NCBI SRA database under the accession number PRJNA669577, and deposited in the NODE database under the accession number OEP003523.