Since the 1950s, China has built 87,000 reservoirs and dams to address water shortage and water security, making them an increasingly important source of drinking water (Liu and Yang, 2012; Wang Y. et al., 2021). However, the frequent occurrence of taste and odor (T&O) events in reservoirs recently has posed a huge challenge to tap water treatment plants, while threatening the safety of drinking water resources (Xu et al., 2022). Odor compounds are often misunderstood as an aesthetic issue, but they are also a major indicator of public acceptance of drinking water (Wu et al., 2021b). Due to their low odor thresholds (both < 10 ng/L), 2-methylisocampheol (2-MIB) and trans-1,10-dimethyl-trans-9-decalol (geosmin) are the most common and typical contaminants that affect water quality in reservoirs (Devi et al., 2021). In China, 2-MIB and geosmin are listed as the main control compounds to ensure the safety of drinking water sources (Rong et al., 2018). Given the chemical stability and small molecular weight of 2-MIB and geosmin, it is difficult to remove them through conventional water treatment processes (coagulation, flocculation, filtration, aeration, and disinfection/oxidation by chlorine), and chemical methods of removing odor substances can produce hazardous oxidation by-products (Sagehashi et al., 2005; Xu et al., 2022). Once encountering T&O episodes in drinking water, it is difficult for water treatment to take applicable control measures, resulting in enormous financial damages. It is necessary to identify and monitor the producers of odor substances in advance to ensure the safety of drinking water and reduce economic losses.

As to the producer of 2-MIB and geosmin, there are mainly producers such as cyanobacteria, fungi, myxobacteria and actinomycetes have been reported (Juttner and Watson, 2007), and cyanobacteria was proven as the main source of 2-MIB and geosmin in the aquatic environment (Lee E.S. et al., 2020). Among more than 100 articles published since 1990 on cyanobacteria associated geosmin/2-MIB in drinking water systems, mainly distributed in North America, Australia, Europe, China, Japan, South Korea, Philippines and South Africa (Su et al., 2015; Devi et al., 2021). Cyanobacterial species that have been associated with 2-MIB- or geosmin-related off-flavors include Anabaena, Aphanizomenon, Lyngbya, Oscillatoria, Phormidium, Planktothrix, and Pseudanabaena (Peterson et al., 1995; Sugiura et al., 1997; Zimba et al., 1999; Su et al., 2015; Zhang et al., 2016; Cai et al., 2017; Rong et al., 2018). A total of 132 cyanobacterial strains from 21 genera and 72 cyanobacterial strains from 13 genera produced geosmin and 2-MIB, respectively. The fungi, myxobacteria and actinomycetes that produce geosmin and 2-MIB have been less studied than cyanobacteria (Devi et al., 2021). The higher incidence of geosmin events in the United States, Canada and Australia, and the higher incidence of 2-MIB events in China and Korea suggest that differences in geographical distribution influence the occurrence and frequency of T&O events (Devi et al., 2021). The Yuqiao Reservoir has experienced eutrophication during the last decade, accompanied by the frequent occurrence of bloom events (Huo et al., 2018). The Yuqiao reservoir has been reported to have earthy-musty taste and odor (T&O) issues with a relatively high amount of 2-MIB. Due to the diversity of these microorganisms, it is still necessary to identify the main producer of 2-MIB and geosmin in the specific waterbody.

On the other hand, not all these microorganisms can be cultured in the laboratory, so it becomes critical to use culture-independent molecular tools to track the odor-producers (Cristina Casero et al., 2019). The biosynthesis pathways of 2-MIB and geosmin were elucidated in actinobacteria and cyanobacteria, providing a molecular basis for traceability analysis of odor compounds (Hamano et al., 2002; Cane and Watt, 2003; Ludwig et al., 2007; Komatsu et al., 2008; Wang et al., 2011). In the synthesis of 2-MIB, the generic precursor of monoterpenoids, GPP (geranyl pyrophosphate), is methylated to methyl-GPP, and then 2-MIB synthase cyclizes methyl-GPP to 2-MIB (Oldfield and Lin, 2012; Brock et al., 2013). Geosmin is generated from farnesyl diphosphate (FPP) synthesized by the bifunctional single sesquiterpene cyclase (geosmin synthase) in the presence of Mg²⁺ (Jiang et al., 2006; Jiang et al., 2007). The geosmin synthase gene and the 2-MIB synthase gene are indicators to investigate the producer of geosmin and 2-MIB, which provided a useful tool to screen the producer of geosmin and 2-MIB. Moreover, high-throughput sequencing is a lower-cost and more efficient sequence determination technology compared to traditional Sanger sequencing, and has played an increasingly important role in the life sciences since its introduction in 2005 (Margulies et al., 2006). Applying high-throughput sequencing to odor compounds biosynthetic genes provides a time-saving, high-resolution tool for the diversity of odor-producing microorganisms in freshwater bodies worldwide. High-throughput sequencing, including amplicon sequencing and metagenomic sequencing, could rapidly detect unculturable low-abundance microorganisms in the environment (Liu et al., 2021). Amplicon sequencing is widely used in the study of microbial diversity, especially for toxic or odor-producing cyanobacteria (Cristina Casero et al., 2019; Qiu et al., 2021). However, the identification of species by amplicon sequencing can only reach a genus-level resolution, and the selection of primers and the determination of the number of PCR cycles during sequencing have a large impact on the sequencing results. Currently, the developed geosmin/2-MIB primers have focused on cyanobacteria, thus quantifying the species composition of geosmin/2-MIB-producing cyanobacteria (Devi et al., 2021). These primers could rapidly identify a lot of geosmin/2-MIB-producing cyanobacteria, such as Anabaena sp. Planktothrix sp., Phormidium sp., Oscillatoria sp., Pseudoanabaena sp (Giglio et al., 2008; Giglio et al., 2011; Tsao et al., 2014). This neglected the contribution of actinomycetes and bacteria to geosmin/2-MIB in the environment. In previous studies, the correlation between the qPCR of geosmin/2-MIB synthase gene and geosmin/2-MIB content was not correlated, low correlated, and highly correlated (Su et al., 2013; Tsao et al., 2014; Chiu et al., 2016; Wang et al., 2016; Huang et al., 2018). When the correlation between gene copy numbers and geosmin/2-MIB concentration is low, it indicates that cyanobacteria contribute only a small fraction of geosmin/2-MIB in water. The presence of odor-producing actinomycetes and bacteria in the water is one of the important factors affecting the correlation value. Metagenomic sequencing is performed by randomly interrupting microbial genomes and assembling small fragments into longer sequences (Liu et al., 2021). Metagenomic sequencing has no primer biases. Metagenomic sequencing can provide not only a more accurate species composition (species or strain level) compared to amplicon sequencing, but also functional gene information (Smits et al., 2017). Nevertheless, the application of metagenomic sequencing technology in odor-producer identification and monitoring is still rare, and more research work is needed.

Besides the above, the synthesis and release of odor compounds were influenced by environmental factors. Temperature and light intensity were the major factors that affect the production of 2-MIB and geosmin (Shen et al., 2022). Unfavorable conditions, such as strong light, low and high temperatures, could increase the potential for geosmin/2-MIB synthesis in cyanobacterial cells (Zimba et al., 1999; Wang et al., 2011; Zhang et al., 2016; Cai et al., 2017). Moreover, nitrogen and phosphorus could affect the production of geosmin and 2-MIB. TN : TP < 29 : 1 favored the generation and accumulation of 2-MIB concentration in the Yangcheng Lake (Wu et al., 2021a). Ammonium was essential to promote the production of 2-MIB and geosmin compounds (Perkins et al., 2019). The microbial communities in waters were diverse, and the production of odor compounds might change under the interaction between different microbial species. Microcystis aeruginosa could obviously promote 2-MIB production of Pseudanabaena sp. (Zhang et al., 2020). Identifying key influencing environmental factors could reduce the risk of T&O outbreaks (Wang et al., 2011; Cai et al., 2017; Oh et al., 2017; Alghanmi et al., 2018) and be useful for adaptive environmental management. Further work is still needed to identify the environmental drivers for geosmin and 2-MIB.

In this study, a long-term field investigation was conducted from 2018 to 2021 in the Yuqiao Reservoir, Tianjin. Macrogenomic approaches were used to investigate the main producers of 2-MIB and geosmin and their related functional genes, while quantile regression was used to find the drivers of 2-MIB and geosmin. The results would be helpful to get insight into the problems of T&O and support drinking water resource management.

The Yuqiao Reservoir has experienced eutrophication during the last decade, accompanied by the frequent occurrence of bloom events (Huo et al., 2018). The coexistence of multiple odors and multiple odor-producers has threatened the Yuqiao Reservoir since 2018. Many studies have shown that phytoplankton was important sources of odor compounds with little consideration given to bacteria and actinomycetes, probably because phytoplankton is easier to observe, isolate and culture compared to actinomycetes and bacteria (Wu et al., 2021b). Amplicon sequencing is commonly used to study microbial diversity in the environment, and the main genes marked are 16srDNA, 18S rDNA and internal transcribed spacer (ITS) (Liu et al., 2021). Amplicon sequencing to analyze microbial functional gene diversity is flawed due to the immaturity of comparative databases and primers. Metagenomic sequencing provides more information, not only extending taxonomic resolution to the species or strain level but also providing potential functional information (Xu et al., 2018). However, metagenomic sequencing requires more funding, high sample quality requirements and increased difficulty in data analysis. Streptosporangium sp. is a member of the family Actinomycetaceae (Kemmerling et al., 1993). 2-MIB and geosmin were first identified from the actinomycete and its molecular formula and exact chemical structure were determined successively in the following years (Gerber and Lecheval, 1965; Medsker et al., 1968). Dolichospermum sp., Pseudoanabaena sp., Microcoleu sp., and Planktothricoides sp. have been frequently reported to be odor-producers all over the world (Niiyama et al., 2016; Zhang et al., 2016; Alghanmi et al., 2018; Huang et al., 2018; Kim et al., 2020). Planktothricoides raciborskii and Dolichospermum circinale were not detected by microscopic observation from June to September 2021. Since Planktothricoides raciborskii and Microcoleus pseudautumnalis Ak1609 are benthic cyanobacteria, water samples were collected only 0.5 m below the water surface for fixation (Stielow and Ballantine, 2003; Su et al., 2015). Odor-producing and non-odor-producing strains often coexist in the field (Cai et al., 2017). Changes in the rise and fall of odor-producing and non-odor-producing strains are considered to be the most important factors regulating the concentration of odors in freshwater (Devi et al., 2021). The abundance of Pseudanabaena sp., Microcoleus sp., and Planktothricoides sp. in the microbial community in July was 1.27%, 0.04%, and 0.87%, respectively, while the abundance of Pseudanabaena sp., Microcoleus sp. and Planktothricoides sp. in the 2-MIB producers in the same month was 29.28%, 45.49%, and 25.22%, respectively. It can be speculated that the proportion of 2-MIB-producing taxa was higher in Microcoleus sp. and Planktothricoides sp. than in Pseudanabaena sp. The relationship between odor compounds and microorganisms in water bodies is a widespread concern at home and abroad. The risk of 2-MIB exceeding 15 ng/L in water was as high as 90% when the Planktothrix sp. density was more than 4.0 × 10⁵ cells/L (Su et al., 2015). Phormidium sp. was the 2-MIB-producing microorganism, accounting for 80–95% of the algal cell density, with an estimated 2-MIB production of 0.022 pg/cell (Sun et al., 2013). A strong positive correlation was detected between 2-MIB concentrations and Pseudanabaena sp. cell numbers (Lee et al., 2022). Geosmin concentrations was positively correlated with cyanobacteria (R² = 0.84, p < 0.0001) and not significantly correlated with actinomycetes (R² = 0.01, p = 0.709) (Lee J. E. et al., 2020). Significant correlation between odor-producers cell densities and 2-MIB/geosmin concentrations was acquired according to the Pearson correlation test (p < 0.01) ( Figure 7 ). The concentration of geosmin/2-MIB in the reservoir may be controlled not only by the density of odor-producer cells but also by environmental factors (Zhang et al., 2016).

Different odor-producing microorganisms vary greatly in their ability to produce odor, and external physicochemical and biological factors may affect growth and odor production (Wang et al., 2011; Wang et al., 2015; Zhang et al., 2020). COD_Mn was the driver for both 2-MIB and geosmin, and both odors increased with increasing COD_Mn. Geosmin and 2-MIB are terpenoids produced and secreted by microorganisms (Lovell et al., 1986). In the correlation analysis between functional genes and environmental factors, a significantly positive correlation between COD_Mn and the biosynthetic pathway of secondary metabolites/terpenoid backbone biosynthesis pathway could be found ( Figure 3B ). In previous studies, several empirical models have been developed based on the strong correlation between chlorophyll a or COD and odor compounds (Bowmer et al., 1992; Rosen et al., 1992; Panova and Dimkov, 2008; Kim et al., 2015). COD had a significant effect on the T&O prediction model and was positively correlated with α-β-ionone concentrations (Qi et al., 2012). The odor occurrence of 2-MIB and geosmin was effectively predicted and validated using multiple linear regression and artificial neural network techniques with chlorophyll a and COD as explanatory variables (Sugiura et al., 2004). COD is an important indicator to determine the pollution of organic matter in water. The decay of aquatic plants in the Yuqiao Reservoir led to an increase in COD_Mn, higher COD_Mn plays a stress factor to phytoplankton and significantly stimulated the synthetize of T&O, and made the increase of odor compounds concentration ( Supplementary Figure 1 ). The three temperature treatments could be arranged in descending order as 10 °C, 35 °C, and 25 °C based on geosmin produced by Dolichospermum ucrainica (Wang and Li, 2015). The production of geosmin was increased with incubation temperature (20–30 °C) by Streptomyces roseoflavus (Tung et al., 2005). Streptomyces tendae cultures incubated at 30 and 45 °C for 48 h produced more geosmin than those incubated at 10 and 20 °C (Dionigi and Ingram, 1994). Many studies found that when external conditions were not adapted to the growth of Cyanobacteria and its growth rate was low, more 2-MIB and geosmin were produced by individual algal cells (Zimba et al., 1999; Wang et al., 2015; Zhang et al., 2016). The direct precursors of geosmin and 2-MIB synthesis, GPP and FPP, are intermediate products of the chlorophyll a synthesis pathway. Some experimental evidence suggests that the biosynthesis of geosmin and 2-MIB in cyanobacteria is associated with chlorophyll a synthesis (Giglio et al., 2008; Zhang et al., 2009; Giglio et al., 2011; Wang et al., 2011). Since geosmin and 2-MIB have common precursors with chlorophyll a synthesis in cyanobacterial cells, the synthesis of earthy-musty odorants and photosynthetic pigments is largely a competition for substrates, i.e., GPP and FPP tend to be converted to 2-MIB and geosmin during the phase of reduced photosynthetic activity and pigment synthesis (Giglio et al., 2008; Zhang et al., 2009; Giglio et al., 2011; Wang et al., 2011; Wang and Li, 2015). At lower or higher temperatures, the production of geosmin by Dolichospermum sp. and Streptomyces sp. is elevated. To our knowledge, our study is one of the first applications of quantile regression in long-term studies of odorant substances in reservoirs, a method that can help managers focus on specific factors, focus future monitoring efforts, and critically provide information.

A four-year survey revealed the distribution of 2-MIB/geosmin and the distribution of odor-producers in the Yuqiao Reservoir, and the following conclusions were drawn. Metagenome unmasked that Pseudanabaena sp. dqh15, Microcoleus pseudautumnalis Ak1609, Pseudanabaena limnetica, and Planktothricoides raciborskii were the 2-MIB-producers, while Streptosporangium caverna and Dolichospermum circinale were the geosmin-producers. Metagenome is a rapid, high-precision and high-throughput method for detecting odor-producers. Quantile regression analysis indicated Pseudanabaena sp. and COD_Mn were the best predictors of 2-MIB concentrations, temperature and COD_Mn were the most useful parameters for describing geosmin concentration change. Combining quantile regression and metagenome results showed that COD_Mn was an important driver of odor compounds. Quantile regression has the advantages of not assuming the existence of moment functions, not being affected by anomalous observations, and not making any distributional assumptions. When accurate and complete data are obtained, quantile regression can quickly find the drivers of odor compounds in the reservoir.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/, PRJNA895910.

PQ: Conceptualization, methodology, software, visualization, formal analysis, writing – original draft, data curation. YZ: Investigation. WM: Project administration. GS: Software, validation. YB: Writing – review & editing, supervision. All authors contributed to the article and approved the submitted version.