Limnological parameters, metabarcoding, infrared spectroscopy, and chemometrics for characterizing freshwater cyanobacteria communities

Abstract

Harmful cyanobacterial blooms pose risks to the environment, humans, and animals. Rapid identification of these microorganisms is essential but still relies on costly and time-consuming techniques. This study evaluates the potential of attenuated total reflectance infrared spectroscopy (ATR-FTIR) combined with chemometrics, supported by molecular methods, to detect potentially harmful cyanobacteria. Samples were collected from 14 sites in the Rio Pardo Valley, RS, Brazil, with limnological parameters measured. Spectra (4000–650 cm⁻¹) in triplicate, and 16S rRNA metabarcoding was performed using Dada2, Phyloseq, and the CyanoSeq database. PCA and OPLS analyses were applied to integrate spectral, taxonomic, and environmental data. Twelve cyanobacterial genera were identified. Sphaerospermopsis (Zapomelová et al., 2010) dominated SP14, associated with aluminum, iron, and nitrogen levels. Dolichospermum ((Bornet & Flahault) P. Wacklin, L. Hoffmann & Komárek, 2009) and Wollea (Bornet & Flahault, 1886) correlated with phosphorus and sodium. Planktothricoides (S. Suda & M. Watanabe, 2002) showed 95.15% abundance at SP01. PC1×PC2 plots revealed location-based clustering of spectral data. ATR/OPLS models showed strong correlations with relative abundances derived from metabarcoding data, with RMSECV values from 1.3×10⁻⁵ to 5.3×10⁻⁴ and R² > 0.99. These results support ATR-FTIR and chemometrics as a promising exploratory approach for the assessment of cyanobacterial community composition in freshwater systems.

Key words
Infrared spectroscopy; Harmful cyanobacteria; Metabarcoding amplicon; Chemometrics

INTRODUCTION

Cyanobacteria are photosynthetic bacteria that have been on Earth for ~3.8 billion years. These organisms are found in a diversity of ecosystems being distributed worldwide in freshwater, marine, and terrestrial environments (Mutoti et al. 2022, Zanchett & Oliveira-Filho 2013). Some cyanobacteria can proliferate and lead to potentially cyanobacterial harmful algal blooms (cyanoHABs) that cause severe damage to environmental and public health. The cyanotoxins released are a major threat in waters used for drinking, irrigation, fishing, and recreation as they can lead to morbidity and mortality of many organisms in the environment (Paerl & Barnard 2020, Zepernick et al. 2023). More intense UV radiation, higher temperatures, and increased nutrients favor the growth of cyanoHABs in freshwater and marine ecosystems over other phytoplankton (Paerl & Barnard 2020, Zhang et al. 2023).

As cyanobacteria can cause large economic and environmental impacts, it is important to understand the communities (Huang et al. 2021). There have been several studies on the taxonomic diversity of cyanobacteria in Brazil using both morphological and molecular methods (e.g., Alvarenga et al. 2021, Batista & Giani 2019, Genuário et al. 2018, Sant’anna et al. 2019, Santos et al. 2018, Werner et al. 2018, 2020). More recently, studies incorporating metabarcoding to study microbial communities, including cyanobacteria, have been undertaken in different areas from Brazil (Butarelli et al. 2022, Ceccon et al. 2019, Gerikas Ribeiro et al. 2018, Guedes et al. 2018, Machado-De-Lima et al. 2019, 2021, Rigonato et al. 2018). These studies give crucial information for water management by identifying the species and communities present in waters.

The occurrence of cyanoHABs in water sources results in more complex water treatment processes and an increase in bottled water consumption due to public distrust, despite the higher prices of bottled water compared to tap water (Cook et al. 2023, Geerts et al. 2020, Pindilli & Loftin 2022, Zat & Benetti 2011). In the context of studying a watershed like the Vale do Rio Pardo in southern Brazil, Lake Dourado stands out because of its history of cyanoHAB episodes (Wilges et al. 2021). The odor threshold concentration in treated water can be easily perceived by the local population (Berlt et al. 2021, Wilges et al. 2021).

Lake Dourado is the main drinking water source for Santa Cruz do Sul (RS) and it is extremely important to monitor watersheds and drinking water reservoir to mitigate damage to environmental and public health caused by cyanoHABs. For this reason, technologies should be investigated for environmental monitoring. Combining robust technologies as metabarcoding with clean and affordable technologies, as infrared, could lead to new perspectives and tools for environmental management.

Fourier Transformation Infrared (FTIR) spectroscopy is a fast, sensitive, low-cost, and non-destructive technique that has been used for microbial identification and quantification, mainly for bacterial samples. Studies have identified microbial species comparing the spectra of different sample origins (e.g., human blood and fluids, food, nucleic acids) to the specific spectra signatures of isolated species that have been classified by FTIR analysis (Ayhan et al. 2021, Farouk et al. 2022, Ghosh et al. 2015, Lee et al. 2019, Maity et al. 2013, Wenning & Scherer 2013). However, FTIR-based microbial quantitative determinations coupled with metabarcoding amplicon sequencing have not yet been investigated, emerging as an alternative approach for diversity studies.

In this context, this study aims to identify and quantify the harmful cyanobacterial community through amplicon metagenomic analysis and correlate the taxa and their abundances to the spectra data from ATR-FTIR. Further, our research investigates the use of ATR-FTIR and chemometrics to predict harmful cyanobacteria in the environment.

MATERIALS AND METHODS

Study area and sampling

The Rio Pardo Valley (Rio Grande do Sul, Brazil) is comprised of 23 cities where Santa Cruz do Sul is located. The main water resource for this city is Lake Dourado. The sampling points were chosen to include water samples upstream and downstream of the lake to understand the harmful cyanobacterial community and their variations within the Pardinho and Pardo rivers. At the moment of sampling the locations had no bloom formation, even though, the study was performed to investigate the potential for ATR-FTIR as an exploratory tool for the preliminary assessment of harmful genera in the environment. Water samples were also collected for physical-chemical analysis (aluminum (Al), calcium (Ca), cobalt (Co), iron (Fe), magnesium (Mg), manganese (Mn), potassium (K), sodium (Na), nitrate (NO₃-), total phosphorus (TP), total nitrogen (TN), dissolved oxygen (DO%), orthophosphate not filtered (ONF), oxidation reduction potential (ORP), temperature (T), and pH) (Apha 2017). Specifically for metal concentrations (Al, Ca, Co, Fe, Mg, and Mn), samples were acidified with ultrapure nitric acid to pH < 2 and analyzed by inductively coupled optical emission spectrometry (ICP-OES).

Samples were collected along the Pardo River watershed (Rio Grande do Sul, Brazil) in September 2022 (Figure 1). Two rivers were included for analysis: Pardinho River with 6 sampling points: SP01, SP02, SP03, SP04, SP05, SP06; and Pardo River with 3 sampling points: SP10, SP11, SP12. Pardinho River is upstream of Lake Dourado, and its water drains into the lake. The lake is the main water reservoir for Santa Cruz do Sul. Also, the Pardinho River drains to the Pardo River. One sample was included that passes through Santa Cruz do Sul before it drains into the Pardinho River (upstream of Lake Dourado). This sampling point, SP13, was collected in the stream in Parque da Gruta (Gruta Park). One sampling point was selected with no connection with the rivers and Lake Dourado, the Castelhano stream, SP14, located in the town of Venâncio Aires (Rio Grande do Sul, Brazil). Three sampling points were selected in Lake Dourado, SP07 is located at the water inlet, SP08 in the middle, and SP09 in the opposite site of water inlet.

Figure 1
Sampling points located in Pardo River watershed. Samplings points: SP01 to SP14. Sampling points were numbered from upstream to downstream of the river. Pardinho River drains to the Pardo River and is also the water source to Lake Dourado. Castelhano stream has no connection to the rivers and the lake.

Surface water samples were collected using sterile 1 L high-density polyethylene bottles and placed on ice until storing at 4°C before analysis. Water was filtered through a 0.22 µm mixed cellulose ester ([MCE] Merck) filter for eDNA and FTIR. Filtration was carried out until the filter was clogged so that there was maximum biomass collected. The volume varied from 1 L to 1.5 L. All filters were stored at -20°C before further processing. The vacuum filtration system was sterilized between each collection point and kept in an acid bath of 10% HCl for 15 min. The filters for DNA extraction and FTIR analysis were dried in room temperature in a vacuum desiccator containing silica for 48 h.

Total environmental DNA extraction, PCR, sequencing, and data analysis

For total environmental DNA extraction, the filter was cut and extracted following a modified protocol by Djurhuus et al. (2017) using DNeasy Blood and Tissue Kit (QIAGEN, Germany). The bacterial 16S rRNA gene region (V4-V5) was amplified by polymerase chain reaction (PCR) using the primers 515F (5’ TGYCAGCMGCCGCGGTAA 3’) and 926YR (5’ CCGYCAATTYMTTTRAGTTT 3’) (Parada et al. 2016). Sequencing was conducted using Illumina NovaSeq 6000 with paired-end (2x250bp) sequencing (Novogene, Beijing, China). Sequences were demultiplexed and assigned to specific sample IDs based on their MIDs at Novogene using an in-house bioinformatic pipeline. DADA2 was used to denoise raw sequences into amplicon sequence variants (ASVs) in R v4.0.0 (R Core Team). Briefly, paired-end reads were filtered, trimmed, and merged prior to dereplication and then analyzed for detection and removal of potential chimeras. Non-chimeric sequences were pooled together to define ASVs (Callahan et al. 2016).

Taxonomic assignment of ASVs was based on naïve Bayesian classifying method (Wang et al. 2007) and the cyanobacterial 16S rRNA database CyanoSeq v1.2 (Lefler et al. 2023) as the taxonomic database. Sequences assigned to categories such as “Chloroplast”, “Eukaryota”, “Mitochondria”, “Archaea”, or with missing phylum-level classification (Phylum = NA) were removed to retain only Cyanobacteria. PCoA and bar plot analysis data were generated in R Studio using Phyloseq (Mcmurdie & Holmes 2013) and MicroViz (Barnett et al. 2021) packages. PCoA analyses were conducted with taxa transformed by Hellinger and ordination by Bray-Curtis distances.

Sequences were deposited at the Sequence Read Archive of the National Center for Biotechnology Information (NCBI) and made publicly available under accession number PRJNA1051624. R code used for data analysis, including a full list of R packages, is on GitHub (github.com/flefler/).

ATR-FTIR analysis

The infrared spectra were acquired in triplicate in ATR-mode in the range of 4000-650 cm-1 with four scanning pulses under 70-75 Nm-2 gauge pressure after deposition of the dried filter in ZnSe crystal of UATR accessory (catalog number L 1250050) coupled in Spectrum 400 FTIR/FT-NIR spectrophotometer (Perkin Elmeyer, USA). Three distinct regions of the membrane were positioned on the ATR accessory to generate true spectral replicates. The spectra were analyzed starting from the higher wavenumber region (4000–2600 cm⁻¹), followed by the mid-infrared region associated with protein and amide bands (1700–1500 cm⁻¹), and finally the fingerprint region (1200–650 cm⁻¹).

Statistical and chemometrics

Univariate statistics were applied for sample characterization of water bodies and their respective community composition using the spectra data. To analyze the regions with absorbances, the graphs were generated with ATR-FTIR mean spectra of dried filters from water bodies (4000-650 cm-1).

For the correlation matrix of physical-chemical parameters, the Spearman Correlation Matrix was applied after verification of abnormal distribution of samples through Shapiro-Wilk test. The data were analyzed using Jamovi Software version 2.3 (Jamovi 2022).

Chemometric analysis was carried out in Pirouette Software version 4.0 (Infometrix). Principal Component Analysis (PCA) was applied to two datasets and the results were represented by the PC1xPC2 diagram of scores and respective loadings. The first set was composed by limnological parameters associated with cyanobacterial genera abundances and was preprocessed by autoscaling. Limnological parameter values that were below the limit of quantification received half of this value. The second set contained the spectroscopic data where the replicate absorbances were normalized by amplitude followed by vector normalization. The mean spectra of each sample were then obtained and preprocessed by mean centering or by autoscaling. Additional data transformation through 1st and 2nd derivatives by Savitzky-Goulay algorithm (5 points) was applied only on the data preprocessed by mean centering.

ATR-FTIR spectra were acquired in triplicate for each sampling point as technical replicates to assess instrumental variability. Replicate spectra were not treated as independent observations. Instead, they were used to evaluate spectral consistency and to generate a single representative spectrum per sampling point. Therefore, the statistical unit of analysis corresponds to the 14 independent sampling points, avoiding pseudoreplication.

Orthogonal Partial Least Square (OPLS) regression analysis was applied to correlate the spectroscopic data (technical replicates averaged fromsample, matrix X, independent variables) with the relative abundance of cyanobacterial genera (matrix Y, supervision data, dependent variables). The spectroscopic data were min-max normalized and vector normalized. Leave three-out cross-validation was applied and the quality of the models assessed by the root mean square error of cross-validation (RMSECV) and by the R2 coefficient of determination. The maximum number of latent variables was defined by rule ASTM (2017) E1655-17. Data transformation of matrix X was carried out using 1st and 2nd derivatives (Savitzky-Goulay, 5 points) and applying orthogonal signal correction (OSC) aiming to optimize the merit figures.

RESULTS

Limnological analysis

The limnological data from the water bodies are provided in Table I. Values of boron, copper, ammonium, and zinc were not included because these parameters had values lower than the detection limit. The SP14 sampling point had a different chemical composition than other locations due to much higher levels of calcium, aluminum, iron and magnesium. SP14 is in Venâncio Aires being the outlier sample for comparison.

Table I
Physical-chemical parameters of sampling points.

Sampling point	Al mg L-1	Ca mg L-1	Co mg L-1	Fe mg L-1	Mg mg L-1	Mn mg L-1	K mg L-1	Na mg L-1	NO3- mg L-1	TP mg L-1	TN mg L-1	TN:TP	ONF mg L-1	DO ppm	ORP (mV)	Temp. °C	pH	Sites
SP01	0.31	5.34	0.0005	0.29	1.36	0.01	2.60	4.20	0.4	0.03	0.25	8.33	0.15	9.67	294.3	17.53	6.05	upper
SP02	0.60	5.33	0.0005	0.62	1.69	0.04	2.09	3.43	0.4	0.03	0.70	23.33	0.03	9.43	316.3	18.26	6.05	upper
SP03	0.62	6.65	0.0005	0.58	2.07	0.01	2.7	3.86	0.8	0.03	0.25	8.33	0.03	9.42	326.1	18.41	6.79	upper
SP04	0.76	8.07	0.0005	0.81	2.80	0.03	2.42	4.81	0.4	0.04	0.25	6.25	0.04	9.78	414.0	17.61	6.36	upper
SP05	0.45	8.31	0.0005	0.60	2.91	0.03	3.06	4.69	0.5	0.05	0.25	5.00	0.04	9.66	480.6	18.33	6.20	upper
SP06	0.73	7.69	0.0005	0.85	2.65	0.03	2.93	4.43	0.2	0.04	0.70	17.50	0.14	8.05	288.1	19.35	6.86	upper
SP07	1.08	6.13	0.0005	1.08	2.92	0.03	18.48	7.15	0.5	0.03	0.25	8.33	0.03	9.93	241.5	19.79	7.38	lake
SP08	2.91	3.93	0.0010	2.88	2.27	0.04	2.18	3.76	0.05	0.04	0.25	6.25	0.04	10.05	273.6	18.99	7.11	lake
SP09	2.51	4.24	0.0005	2.63	2.08	0.04	1.39	3.59	0.6	0.03	0.25	8.33	0.04	8.95	301.7	20.23	6.47	lake
SP10	1.07	8.73	0.0005	1.72	2.64	0.10	2.59	6.47	0.6	0.09	0.7	7.78	0.09	10.26	354.6	19.77	5.28	down
SP11	0.83	8.85	0.0010	1.64	2.76	0.11	2.67	5.81	0.9	0.08	0.25	3.13	0.09	9.41	199.9	20.19	5.61	down
SP12	0.84	8.70	0.0005	1.49	2.68	0.08	2.59	6.61	0.05	0.07	0.25	3.57	0.33	8.91	312.6	20.63	6.14	down
SP13	0.27	6.14	0.0005	0.36	2.71	0.04	1.05	4.98	0.5	0.03	0.70	23.33	0.05	9.85	286.2	16.30	6.56	Gruta Park
SP14	3.75	11.56	0.0020	3.85	4.38	0.13	3.24	6.16	0.7	0.09	0.70	7.78	0.13	8.94	379.3	17.58	6.18	Venâncio Aires

The highest potassium concentrations were found at SP07 with 18.48 mg L-1. SP07 was 6 times higher in potassium than the second highest value in SP05 with 3.06 mg L-1. SP07 is in the water inlet of Lake Dourado, where there was a large abundance of aquatic plants.

SP02, SP06 and SP13 had high ratios of TN:TP at 23.33 mg L-1, 17.50 mg L-1 and 23.33 mg L-1, respectively. All these sampling points are located upstream of Lake Dourado. SP13 is located within the city Santa Cruz do Sul, explaining the high ratio of these nutrients that are probably related to water contamination from the city. The same can be said for SP06 that is located downstream of Sinimbu.

Most of the limnological parameters showed an abnormal distribution. Thus, the Spearman test for correlation matrix (Supplementary Material - Table SI) was applied. Aluminum and iron presented the highest positive correlation showing a value of 0.947 (p < 0.001). Calcium had a significant positive correlation of 0.824 (p < 0.001) with total phosphorus. Additionally, TP and TN:TP ratio exhibited a negative correlation of -0.712, which also lacked statistical significance.

Harmful cyanobacteria metabarcoding amplicon analysis

Using environmental DNA analysis, 12 genera of potentially harmful cyanobacteria were identified in the 14 samples (Figure 2). SP02, SP03 (both Pardinho River) and SP09 (Lake Dourado) showed a higher diversity of harmful cyanobacteria than the other locations (Figure 3a, b). These three locations had all 12 genera in their waters and are grouped in Principal Coordinate Analysis (PCoA). The SP02 and SP03 are the points close to Sinimbu and SP09 is the sampling point opposite the water inlet of the lake, with more accumulation of nutrients that are carried to that side of the lake (Lobo et al. 2011, Martini 2018, Wilges et al. 2021).

Figure 2
Relative abundance of harmful cyanobacteria in the 14 samples from the Pardo River watershed. Legend: down – samples downstream of Lake Dourado; lake – Lake Dourado; Gruta Park – stream; upper - samples upstream of Lake Dourado; VA - Castelhano Stream.

Figure 3
Distribution of sampling points according to relative abundances of taxa and their similarities through PCoA analysis. Figure 3a represents sampling points SP09, SP02 and SP03. Figure 3b represents sampling point SP01. Figure 3c represents sampling points SP04 to SP08 and SP10 to SP14.

Planktothricoides (S. Suda & M. Watanabe, 2002) (34.37%), Microcystis (Lemmermann, 1907) (21.17%) and Sphaerospermopsis (Zapomelová et al., 2010) (14.89%) were most abundant in the SP09 sampling point, where cyanoHABs normally occur in the lake. At the moment samples were collected no bloom formation was identified.

SP02 and SP03 sampling points showed the same three genera with highest abundances with Microcystis having the highest percentages of 20.84% and 41.56%, respectively (Figure 3a).

The SP01 sampling point is the first sampling point in Pardinho River. It is located upstream of Sinimbu. It is the most preserved (less impacted) place in the study area and the water flow was intense during sampling. The water column is very shallow with rocks at the bottom of Pardinho River. Planktothricoides presented a relative abundance of 95.15% in this location (Figure 3b).

The other sampling points are grouped together in the PCoA analysis showing high similarity of diversity (Figure 3c). SP10 is more distant from others in this group because this location presented only four genera, Dolichospermum ((Bornet & Flahault) P. Wacklin, L. Hoffmann & Komárek, 2009) (37.50%), Wollea (Bornet & Flahault, 1886) (37.50%), Sphaerospermopsis (20.83%) and Umezakia (M. Watanabe, 1987) (4.17%). This location is unique since there were no Microcystis nor Planktothricoides found.

The sampling points SP07, SP08, SP11, SP13 and SP14 had the same 6 genera consisting of Planktothricoides, Microcystis, Sphaerospermopsis, Dolichospermum, Wollea and Umezakia. Dolichospermum and Wollea were most abundant during these sampling points.

ATR-FTIR analysis

The dehydrated samples generated a similar profile of FTIR spectra (Figure 4a). Silicates and triglycerides were identified in the region from 3600 to 2600 cm-1 (Figure 4b). Proteins from Amide I and Amide II group were identified from 1700 to 1500 cm-1 and nitrates are represented by absorbance around 1384 cm-1 (Figure 4c). The fingerprint region (1200–650 cm⁻¹) showed the greatest spectral variability among samples (Figure 4d) In this region, silicates were identified around 1030 cm-1, carbohydrates near

Figure 4
ATR-FTIR mean spectra of communities from water bodies (4000-650 cm-1). b, c and d = amplified view of regions highlighted in Figure 4a.

1000 cm-1, carboxylic acids around 910 cm-1, carbonates near 875 cm-1 and nitrates around 840 cm-1.

Characteristic spectra absorptions of silicates or Si-O bands in argyle network were observed in all samples, predominant in SP10 and SP12. These two samples are from the Pardo River, where the water is more turbid with more suspended particles compared to samples collected from Pardinho River, Lake Dourado, Gruta Park and Castelhano Stream.

Statistical and chemometrics analysis

A complementary understanding of the similarities between samples was obtained by applying PCA to the sets of limnological parameters, 16SrRNA-metabarcoding data and spectroscopic parameters (Figure 5).

Figure 5
PC1xPC2 diagram (a) and their respective loadings (b) of cyanobacterial genera abundances and respective physical-chemical parameters of sample points. Data pre-processing by autoscaling. PC1xPC2 diagram obtained with spectroscopic data from community spectra of water bodies (4000-650 cm-1) using min-max normalization followed by vector normalization and mean centering (c) or autoscaling (e). d and f represent loadings of first three PCs from c and e, respectively.

PCA of limnological and metabarcoding autoscaled parameters from 14 water body samples was represented with three PCs which explained 58.69% of original information of data. PC1, PC2, and PC3 represent 33.78%, 15.86%, and 9.07%, respectively. PC1 separates Dolichospermum and Wollea (positive semiaxis) from Aphanizomenon (Morren ex Bornet & Flahault, 1886), Anabaena (Bory ex Bornet & Flahault, 1886), Planktothrix (Anagnostidis & Komárek, 1988), Pseudanabaena (Lauterborn, 1915), Amphiheterocytum (Sant’Anna et al., 2019), Microcystis and Limnoraphis (J. Komárek, E. Zapomelová, J. Smarda, J. Kopecky, E. Rejmánková, J. Woodhouse, B.A. Neilan & J. Komárková, 2013), grouped in negative semi-axis (Figure 5b). Planktothricoides, Umezakia and Sphaerospermopsis placed closer to origin of PC1 axis. Planktothricoides and Umezakia are discriminated from Sphaerospermopsis by PC2 (Figure 5b). Three groups of physicochemical parameters are discriminated by PC2. The first, at positive semi-axis, constituted by metals (Al, Ca, Co, Fe, Mg, Mn), TP, TN, NO₂; the middle, next origin, constituted by TN:TP, ORP, T, ONF, and Na. pH, DO% and K, highly correlated formed the third group at negative semi-axis. PC1 concentrated all metals at the positive semi-axis (Figure 5b).

The first three PCs of mean centered spectroscopic data, grouped 96.30% of original data (PC1 = 75.36%; PC2 = 17.39%; PC3 = 3.56%). PC1 of raw data separate the samples in relation to the origin (location), however the PC2 separate the samples in relation to the spectra reads (Figure 5c). PC1 shows relevant positive contributions between 1100-1000 cm-1 and negative contributions between 990-840 cm-1 (Figure 5d). The data transformation by 1st and 2nd derivatives decreased the sample separation in PC1 level.

The preprocessing by autoscaling produced the distribution of scores at PC1 with high correlation with sampling points (Figure 5e) increasing the data compaction to 90.98% at the first three PCs (PC = 53.40%; PC2 = 20.52%; PC3 = 17.05%). The mean positive loadings for such discrimination were associated with ranges of 3000-2500 cm-1, 1550-1450 cm-1 and 1150-1100 cm-1 (Figure 5f). PC2 and PC3 compressed high amount of noise at 4000-3700 cm-1 and 2700-1750 cm-1, becoming these spectral ranges without value to explain the discrimination of scores.

Considering the abundances of genera observed in each sampling point and comparing the distribution of sampling points through the spectra of the community, SP01 is distributed alone on the positive side of PC1, this location had a high abundance of Planktothricoides. This distribution is related to the higher absorbance in the region of 1000 to 1100 cm-1 in which its related to the presence of Si-O-Si (silicates) and C-O (carbohydrates).

The samples SP10, SP11 and SP12 are grouped together on the positive side of PC1 (Figure 5a) showing the same spectra profile. These are located in the Pardo River. The SP10 and SP11 had a higher relative abundance of Dolichospermum and Wollea, and these genera were correlated with total phosphorus and sodium (Figure 5b).

Planktothricoides is strongly related to SP01 sampling point due its high relative abundance of 95.15%. In the PCA, Planktothricoides is related to pH on the negative side of PC1 and PC2. Furthermore, the SP01 is displayed on the negative side of both PC according to its spectra of the community.

Microcystis was highly abundant in SP02, SP03 and SP09 with a strong correlation to nitrates and TN:TP ratio. These samples were distributed on the negative side of PC1 and positive side of PC2, confirming the correlation of the data.

SP14 was distant from other locations. In both PCAs, SP14 is displayed on the positive side of PC1 and PC2, showing a strong relation of the spectra to the levels of Mn, Mg, Fe, Ca, Co, Al, and P, indicating that positive contributions in 1100-1000 cm-1 as well as negative contributions in 990-840 cm-1 are directly related to these parameters (Figure 5b). This location presented the highest level of Al, Ca, Fe and Mg. Sphaerospermopsis had a strong relationship to this location, related to the aluminum, iron, and total nitrogen levels.

The other sampling points were strongly related to the concentration of sodium, orthophosphate, temperature, potassium, dissolved oxygen, and oxidation reduction potential. Dolichospermum and Wollea were present with higher levels of relative abundances within these locations and were closely related to sodium levels. Umezakia had the highest abundances in SP04, SP07 and SP13 and strongly related to potassium and dissolved oxygen parameters. The SP07 location had the highest value of potassium (18.48 mg L-1) almost six times higher (Figure 5b).

OPLS regression resulted from analysis of latent variables supervised by 16S rRNA-metabarcoding of amplicons (genera and relative abundances) to evaluate the potential of ATR/OPLS models to explore correlations between spectral patterns and cyanobacterial community compositionin water bodies. The error values displayed for each genus demonstrated a strong model for the prediction, as the values are below 0.015 for most of the genera, with only Planktothricoides resulting an error value of 0.028 considering one latent variable (Figure 6a). Models with preprocessing by mean centering or autoscaling, as well as without the 1st derivative transformation, had low predictive quality. This performance was also observed in a model with the 1st derivative transformation, but without the orthogonal signal correction component. The addition of two or three OSC components did not produce an important improvement in the merit figures that would justify its application.

Figure 6
Merit figures of ATR/OPLS prediction models of harmful cyanobacteria genera and relative abundances. Data min-max normalized followed by vector normalization, transformed by 1st derivative and one component of orthogonal signal correction. a = RMSECV versus number of latent variables; b = vector regression profile of respective models with highlight (c) for range at 1250-650 cm^-1.

After the orthogonal correction, all ATR/OPLS models resulted in adequate performance for the estimation of abundances from the 12 genera of harmful cyanobacteria. It reached values of coefficient of determination (R2) between 0.98 and 0.999 with the first latent variable. Using the maximum number of 6 latent variables allowed by the ASTM (2017) E1655-17 for 42 spectra set, the RMSECV values were between 1.3⋅10-5 – 5.3⋅10-4 with R2 values above 0.99999. The results support the potential of ATR/OPLS models as a promising exploratory approach for monitoring water sources. The reliability is guaranteed as the method was developed based on robust and reliable information from metagenomic analysis.

The spectral range from 1100 to 650 cm-1 contains the spectral regions with the highest weights to correlate the latent variables with the relative abundance data of the cyanobacterial genera. These regions are the same for the 12 genera detected (Figure 6b, c). However, the weights vary according to the amplitude of variation of the relative abundances: the higher the quantified relative abundance value for a given genus, the greater the weight of the range in question in the regression vector.

DISCUSSION

Geographic characterization and physical-chemical analysis of sampling points

The Resolution # 357/2005 (Conama 2005) sets the guidelines for levels of organic and inorganic matter in freshwater systems of Brazil. For each individual substance it should be under the limit established in the resolution and monitored by public organizations. In this study, we found aluminum and iron above the range established by such regulation. Aluminum should be below 0.1 mg L-1, but in this study the concentrations varied from 0.27 mg L-1 to a maximum of 3.75 mg L-1. Iron concentration also exceeded the maximum concentration of 0.3 mg L-1 for all samples, varying from 0.36 mg L-1 to 3.85 mg L-1.

The residues and effluents that are discharged in River Pardinho came from cattle farming, agriculture, mineral extraction and domestic effluents. For this reason, the total phosphorus, iron and lead normally present higher concentrations in the river being above the national levels established (Kotzian et al. 2003). The highest concentration for aluminum and iron was found in Castelhano Stream (Venâncio Aires, RS).

According to same Resolution (Conama 2005), pH should range between 6.0 to 9.0. In this study, lower values were observed for sampling points SP10 and SP11 located in the Pardo River, with 5.2 and 5.6, respectively. Purper et al. (2010) also found low values of pH in the Pardo River. The study collected in 9 different sampling points along the river and found one sampling point with pH 5.25. The sampling point P7 in Purper et al. (2010) study was collected very close to SP10 and SP11 comparing to our study. They found a pH value for P7 equal to 6.10, even though still above Resolution, the pH value is considering low.

Harmful Cyanobacterial metabarcoding analysis

In the Pardinho River, SP01 is adjacent to a dense forest with few rural properties and human activities. According to Kotzian et al. (2003), locations as the SP01, characterized by irregular landform and hard access, still have natural forests preserved in the region. The location had the highest relative abundance of Planktothricoides (95.15%) and this location was the least diverse compared to other sampling points. They presented that the low diversity of species in this area is due to a shallow water column, rocky bottom, low nutrient availability and the presence of rapids in the river causing turbulent flows. The Pardo River discharges into the Jacuí River that discharges into the Guaíba Lagoon and ultimately the Patos Lagoon. The Patos Lagoon Estuary is the main connection in southernmost Brazil of freshwater systems to the marine environment. A study monitoring the phytoplankton communities in the Patos Lagoon from 1993 to 2012 reported the occurrence of Microcystis, Anabaenopsis (V.V. Miller, 1923), Aphanizomenon, Dolichospermum, Planktothricoides, Pseudanabaena and Sphaerospermopsis (Haraguchi et al. 2015).

A study by Werner & Laughinghouse (2009) focusing on coiled ‘Anabaena’ (=Dolichospermum and Sphaerospermopsis) around the state of Rio Grande do Sul found Dolichospermum flos-aquae ((Bornet & Flahault) P. Wacklin et al., 2009) in Guaíba Lagoon and Sphaerospermopsis torquesreginae ((Komárek) Werner et al., 2012) in Patos Lagoon (Werner et al. 2012). Both lagoons receive inflow from the Jacuí River.

De Carvalho et al. (2008) also reported the presence of Dolichospermum, Microcystis and Anabaenopsis in a freshwater ecosystem in Rio Grande do Sul. They studied the coastal Lake Violão in Torres and found anabaenopeptin F, anabaenopeptin B, microcystin-LR, and microcystin-RR. In that study, cyanotoxins were directly measured; however, in the present study, toxin production was not assessed, and taxonomic identification is used solely as an indicator of potential ecological relevance.

Werner et al. (2018) found 33 species of cyanobacteria in a pond from an environmental preservation area in Triunfo city (RS, Brazil), including species of Microcystis, Planktothrix and Pseudanabaena that were found in the present study.

Dolichospermum, Microcystis, Aphanizomenon and Pseudanabaena have also been found in other areas of southern Brazil, e.g. Alagados Reservoir (Paraná, Brazil) (Calado et al. 2017).

Several of these genera are reported in the literature to include strains capable of producing cyanotoxins, which has motivated their classification as potentially harmful cyanobacteria in environmental monitoring programs. These toxigenic taxa were identified among 17 species of cyanobacteria with the most toxic being Raphidiopsis raciborskii ((Woloszynska) Aguilera et al., 2018); R. raciborskii was also reported by Werner et al. (2020) in samples from a lake in Canoas, Rio Grande do Sul. Both Aphanizomenon and Raphidiopsis (F.E. Fritsch & M.F. Rich, 1929) are reported in the literature as cylindrospermopsin-producing genera, justifying the need for monitoring water bodies where these taxa occur due to their potential detrimental effects on the environment. It is important to emphasize, however, that the presence of these genera in the current study does not confirm toxin production.

In Joanes Reservoir located in northeastern Brazil (Bahia, Brazil), 15 cyanobacteria genera were described (De et al. 2019). Microcystis, Pseudanabaena, Planktothricoides, and Dolichospermum were observed, as were also found in this study. Microcystis along with Aphanocapsa (Nägeli, 1849), were the most abundant genera in their study. The Microcystis abundance was related to the high levels of aluminum, different than what was found in this study, where Microcystis was related to TN:TP ratio and nitrates. Aphanocapsa was not observed in our samples.

Dolichospermum was dominant in deep systems with lower levels of nutrients in waters studied by Soares et al. (2013). The genus was related to low concentrations of total phosphorus when distributed as monospecific dominance, which was also observed for other waters in South America (Almanza et al. 2019). In the present study, Dolichospermum was highly abundant in SP08, SP10 and SP11. These locations are characterized by deep waters compared to other locations and the concentration of total phosphorus was higher in SP10 and SP11 in present study. The highest relative abundance of Dolichospermum was found in S08. This sampling point is located in the main drinking water reservoir for Santa Cruz do Sul. Dolichospermum is reported in the literature as a genus that can include strains capable of producing cylindrospermopsins, anatoxins, saxitoxins and microcystins (Kramer et al. 2022, Pearson et al. 2016). Dolichospermum, Microcystis and Raphidiopsis are the most prevalent toxic bloom-forming and potentially toxigenic taxa in freshwaters of Argentina and Brazil (Aguilera et al. 2018). These findings reinforce the importance of monitoring cyanobacterial community composition, even in the absence of confirmed toxin measurements.

ATR-FTIR analysis and chemometrics

The spectra of community obtained from dried filters are not only composed of microbial organic matter. Therefore, it is necessary to also consider the presence of organic matter from plants and animals as well as inorganic components from soil particles suspended in the water column to attribute the spectra bands correctly.

Landry & Tremblay (2012) studied organic dissolved matter from freshwater systems finding amides and carbohydrates. The bands between 1660-1575 cm-1 were associated to the presence of amides (proteins). The study also cited the presence of aromatic groups and carbohydrates in those bands. The presence of carboxylate group was described to the bands between 1415-1375 cm-1.

Bands at 1065-1045 cm-1 represent absorptions of carbohydrates, being displayed as the groups C-O and C-H. The bands at 1125-1090 cm-1 are related to OH and C-O-C of carbohydrates, aliphatic ester and ethers, sulfates, and sulfonic acids (Landry & Tremblay 2012). In our study, the bands from 1000 to 1100 cm-1 were the main reason for the separation of samples on the positive side of PC1 in the PCA.

Can et al. (2019) studied the extracellular polymeric substances of cyanobacteria from cultures of Spirulina maxima (Setchell & N.L. Gardner) Geitler. These substances are important for cyanobacteria during stress and bloom formation. ATR-FTIR was used to analyze bands between 1000 to 1100 cm-1 described as stretching of C-O-C etheric bonds of polysaccharides, following Landry & Tremblay (2012) and Volkov et al. (2021).

The band attribution from 1500 to 1700 cm-1 in biological systems is assigned to Amida I (proteins), Amida II (proteins) and carboxylic acid (Volkov et al. 2021). This range matches with Can et al. (2019) study where bands from 1622 to 1530 were related to carboxylic acid and amide group respectively.

The OPLS method showed good results to give information about the harmful cyanobacteria from environmental samples. Commonly the classification of taxa is carried out using algorithms of classification as Partial Least Squares-Discriminant Analysis (PLS-DA). PLS-DA was applied followed by Hierarchical Cluster analysis (HCA) to discriminate bacteria strains in the study of Lee et al. (2019). The analysis was based on the spectra of isolated genomic DNA where PLS-DA data were generated to see the discrimination and compare the results to the 16S rDNA sequences from the selected strains. HCA was used to generate the dendrogram with the separation of strains and there was similar taxonomy compared to the 16S rRNA phylogeny tree. The study confirmed that FTIR data used in PLS-DA analysis were successful to discriminate species and this can be a tool for further studies to identify bacteria in food products. In the present study, the abundances for each genus were obtained through the Dada2 pipeline, the PLS applied was the orthogonal correction only supervised by molecular data.

The application of the method proposed here is limited to freshwaters with limnological characteristics similar to those investigated. Water bodies with differentiated characteristics should be re-analyzed by 16S rRNA amplicon analyses. The results should be included in the pre-existing data for new ATR/OPLS modeling or analyzed separately in a new dataset, developing a less robust but more specific model. This stepwise approach allows the method to be adapted to different freshwater contexts while maintaining methodological consistency.

Within this framework, it is important to consider the time and cost associated with the different approaches used for cyanobacterial monitoring. Traditional light microscopy enables rapid visualization and taxonomic identification and can be performed within minutes by trained personnel; however, it is limited by operator subjectivity and reduced resolution for cryptic or morphologically similar taxa. In contrast, metabarcoding provides high taxonomic resolution and comprehensive community profiling but requires specialized infrastructure, higher costs, and longer processing times, often ranging from several days to weeks. The ATR-FTIR approach combined with chemometrics occupies an intermediate position, offering rapid spectral acquisition within minutes and relatively low operational costs once instrumentation is available, while still relying on molecular data for model development and validation. Therefore, ATR-FTIR should be regarded as a complementary and preliminary screening tool rather than a replacement for microscopy or genomic analyses.

It is also important to emphasize that cyanotoxins were not directly measured in this study. Consequently, the classification of harmful cyanobacteria is based solely on the taxonomic identification of genera reported in the literature to include toxigenic strains, rather than on confirmed toxin production. In addition, ATR-FTIR spectra were normalized prior to chemometric analysis, which removes absolute intensity information and precludes direct estimation of ambient concentrations. As a result, the proposed models establish correlations between spectral patterns and relative abundances derived from metabarcoding data and should be interpreted as exploratory tools rather than quantitative predictors of toxin presence or concentration.

CONCLUSIONS

The biodiversity of harmful cyanobacteria still needs to be investigated in southern Brazil. Even when there is no evidence of harmful cyanobacteria blooms in the environment, it is important to understand the diversity of taxa for further studies of impacts and water quality management of freshwater bodies. This study identified twelve genera of potentially harmful cyanobacteria from different water bodies located in the Rio Pardo Valley in southern Brazil, providing important information on the distribution of these taxa.

In addition, it is important to investigate the environmental characteristics (e.g., physical-chemical parameters) to analyze how these organisms are affected by these conditions. Some physical-chemical parameters (e.g., aluminum, iron, total phosphorus) were over the permitted range in natural water resources. This study is an alert to monitor and investigate the quality of water as two samples analyzed are from waterbodies used for human consumption.

As the monitoring of water bodies for cyanobacterial abundances still depends on time demanding and expensive technologies, this study presented an exploratory model using ATR/OPLS to correlate spectral patterns with cyanobacterial community composition in environmental samples. FTIR coupled with chemometrics represents a rapid and sensitive tool to explore relationship between spectral data and metagenomic amplicons results, supporting the characterization of cyanobacterial community diversity in freshwater systems. Considering the positive results obtained in this study, FTIR combined with chemometric can be regarded as an analytical technological innovation for preliminary screening of cyanobacterial communities, with low environmental impacts. This approach can be used for preliminary assessment of water bodies to investigate community composition prior to further analyses using amplicon metagenomics.

SUPPLEMENTARY MATERIAL

Table SI.

Table SI.

Acknowledgements

This work was supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) - code 001, Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul (FAPERGS) – protocol 21/2551-0002139-0 and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) – protocol 306216/2022-1.

• Data availability
  Not applicable.

References

• AGUILERA A, HAAKONSSON S, MARTIN MV, SALERNO GL & ECHENIQUE RO. 2018. Bloom-forming cyanobacteria and cyanotoxins in Argentina: A growing health and environmental concern. Limnologica 69: 103-114. http://dx.doi.org/10.1016/j.limno.2017.10.006.
  » https://doi.org/10.1016/j.limno.2017.10.006
• ALMANZA V, PEDREROS P, DAIL LAUGHINGHOUSE H, FÉLEZ J, PARRA O, AZÓCAR M & URRUTIA R. 2019. Association between trophic state, watershed use, and blooms of cyanobacteria in south-central Chile. Limnologica 75: 30-41. http://dx.doi.org/10.1016/j.limno.2018.11.004.
  » https://doi.org/10.1016/j.limno.2018.11.004
• ALVARENGA DO, ANDREOTE APD, BRANCO LHZ, DELBAJE E, CRUZ RB, VARANI AM & FIORE MF. 2021. Amazonocrinis nigriterrae gen. nov., sp. nov., Atlanticothrix silvestris gen. nov., sp. nov. and Dendronalium phyllosphericum gen. nov., sp. nov., nostocacean cyanobacteria from Brazilian environments. Int J Syst Evol Microbiol 71: 004811. http://dx.doi.org/10.1099/ijsem.0.004811.
• APHA - AMERICAN PUBLIC HEALTH ASSOCIATION. 2017. Standard Methods For the Examination of Water and Wastewater. American Public Health Association, 23rd ed., Baird RB, Eaton AD & Rice EW (Eds), 1516 p.
• ASTM – ADVANCING STANDARDS TRANSFORMING MARKETS. 2017. E1655-17 - Standard Practices for Infrared Multivariate Quantitative Analysis. Available at: https://standards.globalspec.com/std/4383950/astm-e1655-17
  » https://standards.globalspec.com/std/4383950/astm-e1655-17
• AYHAN K, COŞANSU S, ORHAN-YANIKAN E & GÜLSEREN G. 2021. Advance methods for the qualitative and quantitative determination of microorganisms. Microchem J 166: 106188. http://dx.doi.org/10.1016/j.microc.2021.106188.
  » https://doi.org/10.1016/j.microc.2021.106188
• BARNETT D, ARTS I & PENDERS J. 2021. microViz: an R package for microbiome data visualization and statistics. J Open Source Softw 6: 3201. http://dx.doi.org/10.21105/joss.03201.
  » https://doi.org/10.21105/joss.03201
• BATISTA AMM & GIANI A. 2019. Spatiotemporal variability of cyanobacterial community in a Brazilian oligomesotrophic reservoir: The picocyanobacterial dominance. Ecohydrol Hydrobiol 19: 566-576. http://dx.doi.org/10.1016/j.ecohyd.2019.03.002.
  » https://doi.org/10.1016/j.ecohyd.2019.03.002
• BERLT MMG, SCHNEIDER RCS, MACHADO EL & KIST LT. 2021. Comparative assessment of the degradation of 2-methylisoborneol and geosmin in freshwater using advanced oxidation processes. Environ Technol 42: 3832-3839. http://dx.doi.org/10.1080/09593330.2020.1743371.
  » https://doi.org/10.1080/09593330.2020.1743371
• BUTARELLI ACA, FERREIRA LSS, RIYUZO R, DALL’AGNOL HMB, PIROUPO CM, DA SILVA AM, SETUBAL JC & DALL’AGNOL LT. 2022. Diversity assessment of photosynthesizers: comparative analysis of pre-cultivated and natural microbiome of sediments from Cerrado biome in Maranhao, Brazil. Environ Sci Pollut Res Int 29: 77359-77374. http://dx.doi.org/10.1007/s11356-022-21229-3.
  » https://doi.org/10.1007/s11356-022-21229-3
• CALADO SLM, WOJCIECHOWSKI J, SANTOS GS, MAGALHAES VF, PADIAL AA, CESTARI MM & SILVA DE ASSIS HCD. 2017. Neurotoxins in a water supply reservoir: An alert to environmental and human health. Toxicon 126: 12-22. http://dx.doi.org/10.1016/j.toxicon.2016.12.002.
  » https://doi.org/10.1016/j.toxicon.2016.12.002
• CALLAHAN BJ, MCMURDIE PJ, ROSEN MJ, HAN AW, JOHNSON AJ & HOLMES SP. 2016. DADA2: High-resolution sample inference from Illumina amplicon data. Nat Methods 13: 581-583. http://dx.doi.org/10.1038/nmeth.3869.
  » https://doi.org/10.1038/nmeth.3869
• CAN HK, GURBUZ F & ODABAŞI M. 2019. Partial characterization of cyanobacterial extracellular polymeric substances for aquatic ecosystems. Aquat Ecol 53: 431-440. http://dx.doi.org/10.1007/s10452-019-09699-z.
  » https://doi.org/10.1007/s10452-019-09699-z
• CECCON DM, FAORO H, LANA PDC, SOUZA EM & PEDROSA FO. 2019. Metataxonomic and metagenomic analysis of mangrove microbiomes reveals community patterns driven by salinity and pH gradients in Paranagua Bay, Brazil. Sci Total Environ 694: 133609. http://dx.doi.org/10.1016/j.scitotenv.2019.133609.
  » https://doi.org/10.1016/j.scitotenv.2019.133609
• CONAMA - CONSELHO NACIONAL DO MEIO AMBIENTE. 2005. Resolução No 357 - Classificação dos corpos de água e diretrizes ambientais para o seu enquadramento, bem como estabelece as condições e padrões de lançamento de efluentes. Available at: https://www.gov.br/mma/pt-br
  » https://www.gov.br/mma/pt-br
• COOK KV, BEYER JE, XIAO X & HAMBRIGHT KD. 2023. Ground-based remote sensing provides alternative to satellites for monitoring cyanobacteria in small lakes. Water Res 242: 120076. https://doi.org/10.1016/j.watres.2023.120076.
  » https://doi.org/10.1016/j.watres.2023.120076
• DE CARVALHO LR, PIPOLE F, WERNER VR, LAUGHINGHOUSE IV HD, DE CAMARGO AC, RANGEL M, KONNO K & SANT’ANNA CL. 2008. A toxic cyanobacterial bloom in an urban coastal lake, Rio Grande do Sul state, Southern Brazil. Braz J Microbiol 39: 761-769. http://dx.doi.org/10.1590/S1517-838220080004000031.
  » https://doi.org/10.1590/S1517-838220080004000031
• DE JMAA, DA LUZ LD & DE AGUIAR JUNIOR TR. 2019. Environmental factors driving the dominance of the harmful bloom-forming cyanobacteria Microcystis and Aphanocapsa in a tropical water supply reservoir. Water Environ Res 91: 1466-1478. http://dx.doi.org/10.1002/wer.1141.
  » https://doi.org/10.1002/wer.1141
• DJURHUUS A ET AL. 2017. Evaluation of Filtration and DNA Extraction Methods for Environmental DNA Biodiversity Assessments across Multiple Trophic Levels. Front Mar Sci 4: 314. http://dx.doi.org/10.3389/fmars.2017.00314.
  » https://doi.org/10.3389/fmars.2017.00314
• FAROUK F, ESSAM S, ABDEL-MOTALEB A, EL-SHIMY R, FRITZSCHE W & AZZAZY HME. 2022. Fast detection of bacterial contamination in fresh produce using FTIR and spectral classification. Spectrochim Acta A Mol Biomol Spectrosc 277: 121248. http://dx.doi.org/10.1016/j.saa.2022.121248.
  » https://doi.org/10.1016/j.saa.2022.121248
• GEERTS R ET AL. 2020. Bottle or tap? Toward an integrated approach to water type consumption. Water Res 173: 115578. http://dx.doi.org/10.1016/j.watres.2020.115578.
  » https://doi.org/10.1016/j.watres.2020.115578
• GENUÁRIO DB, SANT’ANNA CL & MELO IS. 2018. Elucidating the Cronbergia (cyanobacteria) dilemma with the description of Cronbergia amazonensis sp. nov. isolated from Solimões river (Amazonia, Brazil). Algal Res 29: 233-241. http://dx.doi.org/10.1016/j.algal.2017.11.034.
  » https://doi.org/10.1016/j.algal.2017.11.034
• GERIKAS RIBEIRO C, LOPES DOS SANTOS A, MARIE D, PEREIRA BRANDINI F & VAULOT D. 2018. Small eukaryotic phytoplankton communities in tropical waters off Brazil are dominated by symbioses between Haptophyta and nitrogen-fixing cyanobacteria. ISME J 12: 1360-1374. http://dx.doi.org/10.1038/s41396-018-0050-z.
  » https://doi.org/10.1038/s41396-018-0050-z
• GHOSH SB, BHATTACHARYA K, NAYAK S, MUKHERJEE P, SALASKAR D & KALE SP. 2015. Identification of different species of Bacillus isolated from Nisargruna Biogas Plant by FTIR, UV-Vis and NIR spectroscopy. Spectrochim Acta A Mol Biomol Spectrosc 148: 420-426. http://dx.doi.org/10.1016/j.saa.2015.03.104.
  » https://doi.org/10.1016/j.saa.2015.03.104
• GUEDES IA, RACHID C, RANGEL LM, SILVA LHS, BISCH PM, AZEVEDO S & PACHECO ABF. 2018. Close Link Between Harmful Cyanobacterial Dominance and Associated Bacterioplankton in a Tropical Eutrophic Reservoir. Front Microbiol 9: 424. http://dx.doi.org/10.3389/fmicb.2018.00424.
  » https://doi.org/10.3389/fmicb.2018.00424
• HARAGUCHI L, CARSTENSEN J, ABREU PC & ODEBRECHT C. 2015. Long-term changes of the phytoplankton community and biomass in the subtropical shallow Patos Lagoon Estuary, Brazil. Estuarine Coastal Shelf Sci 162: 76-87. http://dx.doi.org/10.1016/j.ecss.2015.03.007.
  » https://doi.org/10.1016/j.ecss.2015.03.007
• HUANG IS, PIETRASIAK N, GOBLER CJ, JOHANSEN JR, BURKHOLDER JM, D’ANTONIO S & ZIMBA PV. 2021. Diversity of bioactive compound content across 71 genera of marine, freshwater, and terrestrial cyanobacteria. Harmful Algae 109: 102116. http://dx.doi.org/10.1016/j.hal.2021.102116.
  » https://doi.org/10.1016/j.hal.2021.102116
• JAMOVI. 2022. The Jamovi Project - Computer Software. Versão Version 2.3. Available at: https://www.jamovi.org
  » https://www.jamovi.org
• KOTZIAN HB, PEREIRA D & MARQUES DMLDM. 2003. Zoneamento Ambiental da Bacia do Rio Pardinho (RS, BRASIL). In: XV Simpósio Brasileiro de Recursos Hídricos. Available at: https://files.abrhidro.org.br/Eventos/Trabalhos/154/445.pdf.
• KRAMER BJ, HEM R & GOBLER CJ. 2022. Elevated CO(2) significantly increases N(2) fixation, growth rates, and alters microcystin, anatoxin, and saxitoxin cell quotas in strains of the bloom-forming cyanobacteria, Dolichospermum. Harmful Algae 120: 102354. http://dx.doi.org/10.1016/j.hal.2022.102354.
  » https://doi.org/10.1016/j.hal.2022.102354
• LANDRY C & TREMBLAY L. 2012. Compositional differences between size classes of dissolved organic matter from freshwater and seawater revealed by an HPLC-FTIR system. Environ Sci Technol 46: 1700-1707. http://dx.doi.org/10.1021/es203711v.
  » https://doi.org/10.1021/es203711v
• LEE J, AHN MS, LEE YL, JIE EY, KIM SG & KIM SW. 2019. Rapid tool for identification of bacterial strains using Fourier transform infrared spectroscopy on genomic DNA. J Appl Microbiol 126: 864-871. http://dx.doi.org/10.1111/jam.14171.
  » https://doi.org/10.1111/jam.14171
• LEFLER FW, BERTHOLD DE & LAUGHINGHOUSE IV HD. 2023. Cyanoseq: A database of cyanobacterial 16S rRNA gene sequences with curated taxonomy. J Phycol 59: 470-480. http://dx.doi.org/10.1111/jpy.13335.
  » https://doi.org/10.1111/jpy.13335
• LOBO E, VOOS J & ABREU E. 2011. Utilização de um protocolo de avaliação rápida de impacto ambiental em sistemas lóticos do sul do Brasil. Cad Pesqui 23: 18-32. https://doi.org/10.17058/cp.v23i1.4726.
  » https://doi.org/10.17058/cp.v23i1.4726
• MACHADO-DE-LIMA NM, FERNANDES VMC, ROUSH D, VELASCO AYUSO S, RIGONATO J, GARCIA-PICHEL F & BRANCO LHZ. 2019. The Compositionally Distinct Cyanobacterial Biocrusts From Brazilian Savanna and Their Environmental Drivers of Community Diversity. Front Microbiol 10: 2798. http://dx.doi.org/10.3389/fmicb.2019.02798.
  » https://doi.org/10.3389/fmicb.2019.02798
• MACHADO DE LIMA NM, MUÑOZ-ROJAS M, VÁZQUEZ-CAMPOS X & BRANCO LHZ. 2021. Biocrust cyanobacterial composition, diversity, and environmental drivers in two contrasting climatic regions in Brazil. Geoderma 386: 114914. http://dx.doi.org/10.1016/j.geoderma.2020.114914.
  » https://doi.org/10.1016/j.geoderma.2020.114914
• MAITY JP, KAR S, LIN CM, CHEN CY, CHANG YF, JEAN JS & KULP TR. 2013. Identification and discrimination of bacteria using Fourier transform infrared spectroscopy. Spectrochim Acta A Mol Biomol Spectrosc 116: 478-484. http://dx.doi.org/10.1016/j.saa.2013.07.062.
  » https://doi.org/10.1016/j.saa.2013.07.062
• MARTINI FA. 2018. Avaliação da eficiência do sistema Algae Turf Scrubber (ATS®) no tratamento de águas eutrofizadas e na produção de biomassa perifítica (Tese de Mestrado). Programa de Pós-graduação em Tecnologia Ambiental, Universidade de Santa Cruz do Sul. Available at: https://repositorio.unisc.br/jspui/handle/11624/2278.
• MCMURDIE PJ & HOLMES S. 2013. phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One 8: e61217. http://dx.doi.org/10.1371/journal.pone.0061217.
  » https://doi.org/10.1371/journal.pone.0061217
• MUTOTI M, GUMBO J & JIDEANI AIO. 2022. Occurrence of cyanobacteria in water used for food production: A review. Phys Che Earth A/B/C 125: 103101. http://dx.doi.org/10.1016/j.pce.2021.103101.
  » https://doi.org/10.1016/j.pce.2021.103101
• PAERL HW & BARNARD MA. 2020. Mitigating the global expansion of harmful cyanobacterial blooms: Moving targets in a human- and climatically-altered world. Harmful Algae 96: 101845. https://doi.org/10.1016/j.hal.2020.101845.
  » https://doi.org/10.1016/j.hal.2020.101845
• PARADA AE, NEEDHAM DM & FUHRMAN JA. 2016. Every base matters: assessing small subunit rRNA primers for marine microbiomes with mock communities, time series and global field samples. Environ Microbiol 18: 1403-1414. http://dx.doi.org/10.1111/1462-2920.13023.
  » https://doi.org/10.1111/1462-2920.13023
• PEARSON LA, DITTMANN E, MAZMOUZ R, ONGLEY SE, D’AGOSTINO PM & NEILAN BA. 2016. The genetics, biosynthesis and regulation of toxic specialized metabolites of cyanobacteria. Harmful Algae 54: 98-111. http://dx.doi.org/10.1016/j.hal.2015.11.002.
  » https://doi.org/10.1016/j.hal.2015.11.002
• PINDILLI EJ & LOFTIN K. 2022. What’s It Worth? Estimating the Potential Value of Early Warnings of Cyanobacterial Harmful Algal Blooms for Managing Freshwater Reservoirs in Kansas, United States. Front Environ Sci 10: 805165. http://dx.doi.org/10.3389/fenvs.2022.805165.
  » https://doi.org/10.3389/fenvs.2022.805165
• PURPER M, HANSEN MAF, COSTA ABD, KAUFMANN RC, WETZEL AP, KIRST A & LOBO EA. 2010. Avaliação da qualidade das águas subterrâneas da bacia hidrográfica do Rio Pardo, sistema aquífero Guarani, RS, Brasil. Rev Jovens Pesqui 1: 39-48. https://doi.org/10.17058/rjp.v0i1.2241.
  » https://doi.org/10.17058/rjp.v0i1.2241
• RIGONATO J, KENT AD, GUMIERE T, BRANCO LHZ, ANDREOTE FD & FIORE MF. 2018. Temporal assessment of microbial communities in soils of two contrasting mangroves. Braz J Microbiol 49: 87-96. http://dx.doi.org/10.1016/j.bjm.2017.04.008.
  » https://doi.org/10.1016/j.bjm.2017.04.008
• SANT’ANNA CL, GAMA WA, RIGONATO J, CORREA G, MESQUITA MCB & MARINHO MM. 2019. Phylogenetic connection among close genera of Aphanizomenonaceae (Cyanobacteria): Amphiheterocytum gen. nov., Cylindrospermopsis and Sphaerospermopsis. Algal Res 37: 205-214. http://dx.doi.org/10.1016/j.algal.2018.11.026.
  » https://doi.org/10.1016/j.algal.2018.11.026
• SANTOS KRDS, HENTSCHKE GS, ANDREOTE APD, LAUGHINGHOUSE IV HD, BALLOT A, NOVELO E, FIORE MF & SANT’ANNA CL. 2018. Polyphasic characterization of newly isolated Anabaenopsis (Cyanobacteria) strains from tropical Brazil and Mexico. Phytotaxa 367: 1-12. http://dx.doi.org/10.11646/phytotaxa.367.1.1.
  » https://doi.org/10.11646/phytotaxa.367.1.1
• SOARES MCS, HUSZAR VLM, MIRANDA MN, MELLO MM, ROLAND F & LÜRLING M. 2013. Cyanobacterial dominance in Brazil: distribution and environmental preferences. Hydrobiologia 717: 1-12. http://dx.doi.org/10.1007/s10750-013-1562-1.
  » https://doi.org/10.1007/s10750-013-1562-1
• VOLKOV DS, ROGOVA OB & PROSKURNIN MA. 2021. Organic Matter and Mineral Composition of Silicate Soils: FTIR Comparison Study by Photoacoustic, Diffuse Reflectance, and Attenuated Total Reflection Modalities. Agronomy 11: 1879. http://dx.doi.org/10.3390/agronomy11091879.
  » https://doi.org/10.3390/agronomy11091879
• WANG Q, GARRITY GM, TIEDJE JM & COLE JR. 2007. Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Appl Environ Microbiol 73(16): 5261-5267. https://doi.org/10.1128/AEM.00062-07.
  » https://doi.org/10.1128/AEM.00062-07
• WENNING M & SCHERER S. 2013. Identification of microorganisms by FTIR spectroscopy: perspectives and limitations of the method. AMB 97: 7111-7120. http://dx.doi.org/10.1007/s00253-013-5087-3.
  » https://doi.org/10.1007/s00253-013-5087-3
• WERNER VR & LAUGHINGHOUSE IV HD. 2009. Bloom-forming and other planktonic Anabaena (Cyanobacteria) morphospecies with twisted trichomes from Rio Grande do Sul State, Brazil. Nova Hedwigia 89: 17-47. http://dx.doi.org/10.1127/0029-5035/2009/0089-0017.
  » https://doi.org/10.1127/0029-5035/2009/0089-0017
• WERNER VR, LAUGHINGHOUSE IV HD, FIORE MF, SANT’ANNA CL, HOFF C, SANTOS KRS, NEUHAUS EB, MOLICA RJR, HONDA RY & ECHENIQUE RO. 2012. Morphological and molecular studies of Sphaerospermopsis torques-reginae (Cyanobacteria, Nostocales) from South American water blooms. Phycologia 51: 228-238. https://doi.org/10.2216/11-32.1.
  » https://doi.org/10.2216/11-32.1
• WERNER VR, MARTINS MD & DA SILVA FO. 2018. Cyanobacteria from a Brazilian subtropical freshwater water body. Braz J Bot 41: 901-921. http://dx.doi.org/10.1007/s40415-018-0502-8.
  » https://doi.org/10.1007/s40415-018-0502-8
• WERNER VR, TUCCI A, SILVA LMD, YUNES JS, NEUHAUS EB, BERTHOLD DE & LAUGHINGHOUSE IV HD. 2020. Morphological, ecological and toxicological aspects of Raphidiopsis raciborskii (Cyanobacteria) in a eutrophic urban subtropical lake in southern Brazil. Iheringia, Sér Bot 75: e2020018. http://dx.doi.org/10.21826/2446-82312020v75e2020018.
  » https://doi.org/10.21826/2446-82312020v75e2020018
• WILGES HG, BEN DA COSTA A, MACHADO ÊL, BERLT MMG, SOARES J, HOPPE M, RIZZETTI TM, SANCHEZ-BARRIOS A & SCHNEIDER RDCDS. 2021. Monitoring of 2-Methylisoborneol and Geosmine in a Constructed Lake to Public Catchment in Southern Brazil. Interfaces Cient - Saúde Ambient 8: 279-293. http://dx.doi.org/10.17564/2316-3798.2021v8n3p279-293.
  » https://doi.org/10.17564/2316-3798.2021v8n3p279-293
• ZANCHETT G & OLIVEIRA-FILHO EC. 2013. Cyanobacteria and Cyanotoxins: From Impacts on Aquatic Ecosystems and Human Health to Anticarcinogenic Effects. Toxins 5: 1896-1917. https://doi.org/10.3390/toxins5101896.
  » https://doi.org/10.3390/toxins5101896
• ZAT M & BENETTI AD. 2011. Remoção dos compostos odoríferos geosmina e 2-metilisoborneol de águas de abastecimento através de processos de aeração em cascata, dessorção por ar e nanofiltração. Eng Sanit Ambient 16: 353-360. http://dx.doi.org/10.1590/S1413-41522011000400006.
  » https://doi.org/10.1590/S1413-41522011000400006
• ZEPERNICK BN, WILHELM SW, BULLERJAHN GS & PAERL HW. 2023. Climate change and the aquatic continuum: A cyanobacterial comeback story. Environ Microbiol Rep 15: 3-12. http://dx.doi.org/10.1111/1758-2229.13122.
  » https://doi.org/10.1111/1758-2229.13122
• ZHANG Y, WHALEN JK, CAI C, SHAN K & ZHOU H. 2023. Harmful cyanobacteria-diatom/dinoflagellate blooms and their cyanotoxins in freshwaters: A nonnegligible chronic health and ecological hazard. Water Research 233: 119807. https://doi.org/10.1016/j.watres.2023.119807.
  » https://doi.org/10.1016/j.watres.2023.119807

Edited by

• Handling editor
  Vasco Azevedo

Data availability

Not applicable.

Publication Dates

• Publication in this collection
  28 Aug 2026
• Date of issue
  2026

History

• Received
  19 Apr 2025
• Accepted
  27 Mar 2026