Environmental parameters of each river at different phases
Water samples were collected from corresponding water bodies, important environmental parameters including water temperature, pH, Chl-a, total nitrogen, and total phosphorus were summarized in Table 1, and complete environmental parameters were summarized in Table S1. In general, the Chl-a in the pools was higher than that in the corresponding rivers, and the maximum concentration reached 535.68µg/L in JFZ’s pool. Further statistical analyses on the complete parameters of each river at different algal bloom phases indicated that for each river, there was no significant difference amongst the parameters at different phases (indicating as N.S.), except ORP and Chl-a (P<0.05, Table S1). For these two parameters, detailed significances between relevant phases were included.
General statistics of sequencing data and alpha-diversity comparison
General statistics of sequencing data
In this study, a total number of 5,115,068 sequences was pooled for the total 104 samples, and for each sample, the sequencing data was normalized to 18,633 sequences, which were categorized into 8,689 OTUs and 2,998 bacterial species. The cyanobacterial community were collectively extracted from each sample, stratified into 515 OTUs and 191 species, leaving the bacterial communities constituting 8,174 OTUs and 2,807 species.
Rarefaction curves for each sample showed that most samples tended to approach saturation (Fig. S1), and the Good’s coverage ranged from 94.91–99.42% amongst the samples. The two indices indicated that the majority of the bacterial taxa had been extracted from the studied communities (Fig. S1). Considering that the tributary pools were one of the main sources of cyanobacterial communities in rivers, alpha-diversity indices, including Shannon and Chao 1, were analyzed on cyanobacterial and the rest bacterial communities in pools and rivers (Fig. S2).
Alpha-diversity comparisons and statistics
For cyanobacterial communities in pools of HS, JFZ, SH, and XH across the algal bloom phases, the Shannon index maximized at “During AB” phase for JFZ and XH. Although it maximized at “Post AB” and “Before AB” phase for HS and SH, respectively, the diversities at “During AB” phase were also relatively high for both pools (Fig. S2 A1). Whilst Chao 1 index of these cyanobacterial communities showed a concurrent pattern that community diversities minimized at “During AB” and maximized at “Post AB” phase (Fig. S2 A2). For bacterial communities, both Shannon and Chao 1 indicated that community diversity decreased at “During AB” and increased afterwards (Fig. S2 A3 and A4). On the contrary to pools, cyanobacterial communities in all rivers minimized at “During AB” phases according to Shannon and Chao 1 index (Fig. S2 B1 and B2), except for Shannon of HS and JFZ. Similar with pools, bacterial communities in all rivers minimized at “During AB” phases and increased afterwards based upon both indices estimate (Fig. S2 B3 and B4).
Differences of microbial communities on alpha-diversity were further determined amongst different rivers at the same phase as well as individual rivers at the consecutive phases (Fig. 2). For the cyanobacterial communities at each phase, there was in general no significant difference amongst the studied rivers (P>0.05), except for the comparisons between HS and JFZ at “Before AB”, as well as JFZ and SH at “Post AB” phases (P<0.05) (Fig. 2A1-A4). Whereas, community diversities in individual rivers exhibited apparent variances across the consecutive phases (Fig. 2B1-B4). Cyanobacterial communities in HS showed no significant difference across the phases (Fig. 2B1), whilst communities in JFZ, SH, and XH demonstrated significant differences amongst these phases, and especially, communities at “During AB” phase were significantly lower than the rest three phases according to Chao 1 index for JFZ, SH, and XH (Fig. 2B2-B4).
Similarly, for the bacterial counterparts at the same phase, there was in general no significant difference amongst the rivers (P>0.05), except for the comparison between JFZ and XH according to Shannon index (Fig. 2C1-C4). Bacterial communities in individual rivers also showed similar variance with the cyanobacteria across the phases (Fig. 2D1-D4). HS communities had no significant difference amongst the phases (P>0.05) (Fig. 2D1), whilst JFZ, SH, and XH communities showed significant differences, and communities at “During AB” phase were also significantly lower than some of the rest phases in JFZ, SH, and XH, respectively (Fig. 2D2-D4).
OTU comparisons of both cyanobacteria and bacteria in tributary pools and their rivers at corresponding phases were performed (Fig. S3 a1-d4, and A1-D4). In general, for both cyanobacterial and bacterial OTUs, the comparisons either amongst pools and rivers at certain phases (Fig. S3 a1-b4 and A1-B4) or in each individual pool and river at different phases (Fig. S3 c1-d4 and C1-D4), only a small portion of OTUs was shared in common amongst the four compared counterparts. It should be noted that only a small number of cyanobacterial OTUs was classified in pools of JFZ, SH, and XH during the algal bloom (Fig. S3 c2-c4), which supported the alpha-diversity analyses that the cyanobacteria diversities during the algal bloom were significantly lower than the rest phases for JFZ, SH, and XH (Fig. 2B2-B4).
Comparisons of cyanobacterial and bacterial community composition
In general, most diverse cyanobacterial OTUs were assigned to phylogenetic orders of Chloroplast, Cyanobacteriales, and Synechococcales in both rivers and their tributary pools. However, the cyanobacterial community composition exhibited spatial variance amongst these ecosystems. In each pool, the three taxa took turns as the most dominant community, for example, in the tributary pool of HS, the most dominant community was Chloroplast, Synechococcales, and Cyanobacteriales across each of the phases; whilst JFZ had Chloroplast and Cyanobacteriales as the most dominant communities. For the cyanobacteria in rivers, Chloroplast was the most dominant taxa in HS river, but Synechococcales was the most dominant in the rest of three rivers along the algal bloom (Fig. 3A).
For bacterial communities, abundant phylogenetic genus to which most diverse OTUs were assigned were sorted out in each river. Based upon relative abundance of each taxon, the phylogenetic genera were further categorized into groups of AAT, ART, AMT, and CVT, respectively (Fig. 3B-E). Although these rivers had some genera in common in each group (AAT, ART, AMT, and CVT), such as Acinetobacter and Flavobacterium,
the bacterial community composition also exhibited spatial variance amongst the studied rivers. For example, community of hgcI clade was always abundant in HS, JFZ, and SH (Fig. 3B-D), whilst was conditionally varied in XH (Fig. 3E), and its relative abundance was greatly reduced during the algal bloom, especially in XH. Communities of Pseudarcicella and Limnohabitans in HS were always abundant and their relative abundance increased during the algal bloom (Fig. 3B), however, their relative abundances were conditionally varied and noticeably reduced in the rest three rivers during the same phase (Fig. 3C-E). Similarly, Comamonadaceae was always abundant in each river, but its relative abundance clearly reduced during the algal bloom, and Allobaculum was always rare in HS, SH, and XH (Fig. 3B, C, E), but conditionally varied in JFZ (Fig. 3D) that its relative abundance greatly increased during the algal bloom.
Variations of cyanobacterial and bacterial communities were further illustrated through NMDS with PERMANOVA test. Cyanobacterial communities at the “Before AB” phase exhibited significant variation amongst the studied rivers (PERMANOVA, df = 3, F model = 3.220, R2 = 0.305, P = 0.001), however not at the rest three phases (Fig. 4A1-A4). Across the algal bloom phases, cyanobacterial communities in SH and XH showed significant variations (PERMANOVA, df = 3, F model = 2.252 and 2.298, R2 = 0.252 and 0.256, P = 0.011 and 0.02, respectively), whilst those in HS and JFZ did not (Fig. 4B1-B4). For bacterial communities, they only showed significant difference at the “Post AB” phase amongst the studied rivers (df = 3, F model = 1.729, R2 = 0.191, P = 0.039) (Fig. 4C1-C4). However, except for HS, bacterial communities in JFZ, SH, and XH collectively revealed significant discrepancy across the algal bloom phases (df = 3, F model = 1.884, 3.731, and 4.525, R2 = 0.220, 0.359, and 0.404, P = 0.006, 0.001, and 0.001, respectively) (Fig. 4D1-D4).
Significant difference in cyanobacterial and bacterial biomarkers (OTUs)
Above results indicated that the discrepancies of both cyanobacterial and bacterial communities could be more attributed to the difference of algal bloom phases instead of the spatial heterogeneity of rivers. Therefore, the significantly different OTUs of cyanobacterial and bacterial communities were analyzed amongst samples at different phases in each individual river (containing corresponding tributary pools) (Table S2). The significance test was subsequently used for LEfSe analysis (Table S3) to identify those cyanobacterial and bacterial OTUs at different phases contributing to the significance.
For cyanobacteria, more OTUs were identified in JFZ, SH, and XH, and these OTUs were mainly identified at the rest three phases except the “During AB” phase (Fig. 5). Cyanobacterial OTUs at different phases were mainly classified into Chloroplast (OTU2648, OTU7367, OTU8492, OTU6617, OTU5009, and OTU867) and Synechococcales (OTU7163, OTU6821, OTU8339, OTU8374, and OTU6991) in JFZ (Fig. 5B), whilst the counterparts were mainly categorized into Chloroplast (OTU8417, OTU7367, OTU2648, and OTU983, etc.), Cyanobacteriales (OTU1843), and Leptolyngbyales (OTU3191) in SH (Fig. 5C), and were pooled into Chloroplast (OTU8492, OTU 6675, OTU7166, OTU2082, and OTU983, etc.) and Cyanobacteriales (OTU6199) in XH (Fig. 5D), respectively. Additionally, several OTUs, including OTU2648, OTU7367, OTU8417, and OTU867, were found to be widely distributed amongst these three rivers, which indicated that they might share a core cyanobacterial community.
Similarly for bacterial community, more OTUs were identified in JFZ, SH, and XH, however were universally identified across the algal bloom phases (Fig. 6). Members of Rhodobacteraceae (OTU 853 and OTU 5851), Limnohabitans (OTU4177, OTU216, and OTU3910), Porphyrobacter (OTU6265), and Sporichthyaceae (OTU2594) were mainly screened out in JFZ (Fig. 6B), whilst Pseudarcicella (OTU3296), Comamonadaceae (OTU3244), Sporichthyaceae (OTU2594), Limnohabitans (OTU3910), Acinetobacter (OTU8318, OTU2460, OTU2685, OTU5230, OTU2703, OTU5164, OTU483, and OTU2184) were broadly identified in common in SH and XH (Fig. 6C and D). Notably, members of Rhodobacteraceae (OTU853 and OTU5851), Porphyrobacter (OTU6265), and Acinetobacter (OTU2460, OTU2685, OTU2184, and OTU5230) in JFZ, SH, and XH exhibited significantly higher (P<0.05) relative abundances at “During AB” phase than the rest three phases (Fig. 6B-D), which may indicate their relationships with the cyanobacterial communities causing algal bloom. Correlations between the dominant cyanobacterial taxa and the above identified biomarkers were subsequently analyzed (Fig. 7). In general, the results indicated that same bacterial taxa in different aquatic ecosystem showed different correlations with the same cyanobacterial taxa. For example, OTU853 in JFZ was negatively correlated with Chloroplast and Cyanobacteriales along the whole bloom phases (Fig. 7A), however, it showed contrasting correlation patterns with Chloroplast and Cyanobacteriales in XH (Fig. 7C). OTU6265 in JFZ was positively correlated with Cyanobacteriales and Synechococcales for the first three phases, but negatively correlated in the last phase (Fig. 7A), however, it was mostly negatively correlated with these cyanobacterial taxa in SH (Fig. 7B). These discrepancies on correlation analyses indicated that the bacteria-cyanobacteria interaction may additionally be affected by environmental factors.
Correlations of cyanobacterial and bacterial communities with the physiochemical parameters
The VIF analysis (Table S4) on the physiochemical parameters against the cyanobacterial and bacterial communities concluded that all parameters expect DTN, DTP, and COD were selected for cyanobacterial CCA/RDA, whilst all the parameters were used for bacterial CCA/RDA, and analyses were separately performed on microbial communities of each river at different phases (Fig. 8).
In general, cyanobacterial communities in these rivers at different phases were mainly correlated with ORP, pH, TEMP, MPO4−, NO3− and Chl-a, and communities of each individual river had specific correlation with additional parameters (Fig. 8A1-A4). For example, cyanobacteria in JFZ also had close correlation with TP, TN, and NH4+ (Fig. 8A2), and the ones in SH had close correlation with TN, NH4+, and NO2− (Fig. 8A3). Additionally, cyanobacterial communities of these rivers also exhibited variance on correlations with the main parameters, which indicated the discrepancy of the cyanobacterial communities amongst these rivers. For instance, most of the samples of HS across algal bloom were positively correlated with ORP, TN, NH4+, NO3−, MPO4−, TP, TN, and were either not or negatively correlated with the rest main parameters, including pH, TEMP, NO2−, and Chl-a (Fig. 8A1), whilst most samples in JFZ but the ones of “Before AB” showed positive correlation with TEMP, ORP, NH4+, and MPO4−, and negative correlation with TN, NO3−, TP, pH, and Chl-a (Fig. 8A2). Cyanobacterial communities in SH were universally negatively correlated with the main parameters (Fig. 8A3), whilst the ones in XH largely illustrated positive correlation with TEMP, pH, NO3−, and MPO4− (Fig. 8A4). The correlation between XH communities and the main parameters were reflected through RDA according to the DCA analysis.
For bacterial communities in all rivers at different phases, they were mainly correlated with COD, TN, DTN, NO3−, ORP, TEMP, and pH, and communities of each individual river also showed specific correlation with additional parameters (Fig. 8B1-B4). Bacterial communities in HS additionally exhibited close correlation with TP, NO2− and Chl-a (Fig. 8B1), the counterparts in JFZ were with NH4+ and Chl-a (Fig. 8B2), the ones in SH and XH were both with TP, MPO4− and NH4+, whilst XH showed additional correlation with DTP, and NO2− (Fig. 8B3&B4). Interestingly, different from cyanobacteria, the bacterial communities of these rivers demonstrated relatively consistent correlation patterns with the main parameters. For example, communities of “During AB” in these rivers universally showed positive correlation with pH, TEMP, and TP, and negatively with TN, DTN, NO3−, and ORP, except HS, the rest three rivers also collectively showed negative correlations with Chl-a (Fig. 8B1-B4). Furthermore, communities of “During AB” in these rivers consistently exhibited contrasting correlation patterns compared with their counterparts either in “Post AB” or “Before AB” (Fig. 8B1-B4).