Comparison of Two Bacterial Characterization Techniques for the Genomic Analysis of River Microbiomes

Abstract

This study compares the feasibility of matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry with whole genome sequencing (WGS) for identifying bacteria in river surface water samples. We collected samples from four rivers in the Dominican Republic and used both techniques to characterize bacterial profiles. MALDI-TOF demonstrated high precision, with 86.2% similarity to WGS results, except for a few discordant cases due to database limitations. MALDI-TOF provided cost-effective and rapid identification, making it a promising alternative to WGS in resource-constrained regions. In particular, good effectiveness of MALDI-TOF in identifying bacteria with a high probability of being resistant to antibiotics was observed, which allows this technology to be used in the monitoring processes of this type of microorganism for their rapid, accurate, and low-cost identification. We found this technology to be advantageous for environmental bacterial profiling, with potential applications in understanding waterborne pathogenic bacteria. Our findings underline the relevance of MALDI-TOF in microbiology and its potential to expand its capabilities in bacterial identification and protein profiling.

1. Introduction

Regarding analytic methods, humanity has permanently moved from classic chemical and physical analysis to instrumental and more accurate methods. For these latter ones, new updates have been coming out to facilitate them while maintaining or upgrading the accuracy of results [1]. In the case of identification of species in environmental samples, sequencing has become the standard due to the gradual increase in its availability as well as its high accuracy [2]. However, in the last decade Matrix-Assisted Laser Desorption Ionization-Time of Flight mass spectrometry (MALDI-TOF) has been explored as a possible alternative method with lower cost and time [3]. This technique has been used to profile bacterial proteins from cell extracts [2]. The procedure provides a unique spectral mass (a signature graph for each microorganism, also called their fingerprint) of the microorganisms that can identify their species with an accuracy higher than 90%, making this procedure ideal for clinical microbiology [4]. It is important to mention the latter because bacterial identification is essential to develop effective antimicrobial decisions. In addition, this identification method has great potential in water samples analysis due to the capacity to process greater amounts of samples at lower cost and time, a major advantage when characterizing a microbiome [5]. For this reason, in the last decade a number of studies regarding surface water in rivers have been published [6,7].

On the other hand, genomic analysis of complex environmental samples is becoming an essential tool for understanding the evolutionary history and functional and ecological biodiversity, the only downside being the high-cost and long time requirements of the procedure [8,9]. In addition, the widespread use of DNA sequencing in recent decades has played a pivotal role in accurately identifying bacterial isolates and discovering novel bacteria [2]. DNA barcoding utilizes standardized species-specific genomic regions (DNA barcodes) to generate vast DNA libraries in order to identify unknown specimens [10]. This methodology is essential, particularly in bacteria with unusual phenotypic profiles, slow-growing or uncultivable bacteria, and culture-negative infections [11]. An excellent example of this in bacteria is the 16S ribosomal RNA [12].

The main objective of this article is to compare the viability of the MALDI-TOF technique as an alternative to the Whole Genome Sequencing technique in surface water river samples. The objective is to determine how close the accuracy of mass spectrometry is to the accuracy of a high-precision and high-cost method in DNA sequencing, as the funds devoted to research in the Dominican Republic are rather limited. Studying environmental bacterial profiles is crucial for public health, as many pathogenic bacteria have been identified in its rivers [6].

2. Materials and Methods

2.1. Sampling

Sampling was performed in 2019 for the Isabela River, Ozama River, Yaque del Norte River, and Yaque del Sur River from Santo Domingo in the Dominican Republic. The points sampled for these rivers were, for Isabela: 18.510111, −69.913111 (Point A), 18.513805, −69.898666 (Point B) and 18.696638, −70.158861 (Point C) see Figure 1; for the Ozama: 18.526683, −69.860283 (Point A), 18.518502, −69.895071 (Point B) and 18.552950, −69.822415 (Point C) see Figure 1; for the Yaque del Norte river: 19.067232, −70.864517 (Point A), 19.153600, −70.644250 (Point B), 19.454433, −70.716350 (Point C), 19.589842, −71.059632 (Point D), 19.706249, −71.499045 (Point E), 19.829313, −71.647867 (Point F) see Figure 2; for the Yaque del Sur river: 18.911396, −71.014936 (Point A), 18.398545, −71.185798 (Point B), 18.299459, −71.172602 (Point C), 18.252541, −71.140495 (Point D) see Figure 3. For the sampling method, we took one sample and two replicas of each point in a time-lapse of six hours, one replica every two hours. These waters are not treated, all samples were taken directly from the river, and proper sewage and wastewater systems are not yet implemented.

2.2. Bacterial Isolation

Following the recommendations from [13], 1 mL, 10 mL, and 50 mL aliquots were subjected to vacuum filtration in triplicate using a nitrocellulose membrane (0.22 $\mathsf{\mu }$m pore size) from Simsii, INC. Bacteria retained on the membranes were subsequently cultured in two different media: (a) MacConkey agar supplemented with imipenem (4 $\mathsf{\mu }$g/mL) and (b) MacConkey agar supplemented with cefotaxime (8 $\mathsf{\mu }$g/mL). Then, the samples were incubated at 37 °C for a duration ranging from 16 to 48 h. In addition, isolates were obtained by streaking the samples onto chromogenic culture media (ChromAgar™ Orientation, France) using the streak plate method. Finally, pure isolates were preserved at −70 °C in 25% glycerol.

2.3. Bacterial Classification and Identification

Microbial identification followed the CLSI guidelines with some adaptations using MALDI-TOF (matrix-assisted laser desorption ionization-time of flight) technology [12,14]. A BioTyper® 3.1 software from Bruker Daltonics, Germany, equipped with the MBT 6903 MPS library released in 2019, was utilized to ensure reproducible results. The manufacturer’s recommended MALDI BioTyper Preprocessing Standard Method and the MALDI Biotyper MSP Identification Standard Method were adjusted and employed. Isolated colonies were cultured on blood agar at 35 °C for 24 h. Approximately 0.1 mg was inoculated onto a sample carrier for each new culture using the complete cell transfer protocol. Subsequently, the samples were coated with 1 $\mathsf{\mu }$L of a matrix solution (10 mg/mL) composed of $\alpha$-cyano-4-hydroxycinnamic acid in a mixture of 50% acetonitrile and 2.5% trichloroacetic acid. The coated samples were then allowed to dry at 25 °C for 20 min. Identification was performed in triplicate for each sample.

2.4. Genomic DNA Extraction from Isolates

Genomic DNA was extracted from colonies incubated in TSB at 35 °C for 24 h. A 4 mL aliquot of the culture was centrifuged at 8000× g for 2 min to initiate the extraction process. The resulting cell pellet was then subjected to the DNeasy Blood & Tissue kit (Qiagen, Hilden, Germany) with the following adaptations: the bacterial pellet was resuspended in a modified lysis buffer consisting of 20 $\mathsf{\mu }$L proteinase K, 200 $\mathsf{\mu }$L of TSB, and 100 $\mathsf{\mu }$L of Qiagen’s ATL buffer. This suspension was incubated at 56 °C for 10 min. Next, 50 $\mathsf{\mu }$L of absolute ethanol was added, followed by a 3-min incubation at room temperature. Subsequently, the extraction protocol continued according to the manufacturer’s recommendations. The obtained DNA was suspended in 50 $\mathsf{\mu }$L of Qiagen’s TE buffer. The integrity of the extracted DNA was assessed by running it on a 1% agarose gel stained with SYBR Green, and the gel was electrophoresed at 100 V for 60 min.

2.5. Genome Sequencing, Assembly, and Analysis

The construction of sequencing libraries involved the following steps:

(I)

The genomic DNA was fragmented randomly through sonication.

(II)

The fragmented DNA was subjected to end polishing, A-tailing, and ligation with Illumina sequencing full-length adapters. This was followed by PCR amplification using P5 and indexed P7 oligos.

(III)

The PCR products, which represented the final construction of the libraries, were purified using the AMPure XP system from Beckman Coulter Inc., located in Indianapolis, IN, USA.

To ensure the quality and size distribution of the sequencing libraries, a quality control assessment was performed using an Agilent 2100 Bioanalyzer from Agilent Technologies, based in CA, USA. Additionally, the libraries were quantified by real-time PCR to meet the criteria of 3 nM.

An Illumina NovaSeq 600 platform was utilized for sequencing the whole genomes, employing the PE 150bp strategy. The sequencing process was conducted at the America Novogene Bioinformatics Technology facility Co., Ltd.

Whole genomes were sequenced using the Illumina NovaSeq 600 platform employing the PE 150 strategy. The sequencing process was performed at America Novogene Bioinformatics Technology Co., Ltd., located in Sacramento, CA, USA. The genomes were assembled using the Assembly HiSeq Pipeline [13], which utilizes various quality control tools such as FastQC for read quality analysis and visualization [15], AdapterRemoval v2 for adapter removal [16], and KmerStream for k-mer distribution computation [17]. The assemblers Edena V3 [18] and Spades 3.9.1 [19] were used for genome graph construction, and Unicycler [20] was employed to optimize and integrate the assemblies. Whole-genome annotation was performed using RAST [21] and Prokka [22]. Individual plasmid sequences within the genome assemblies were predicted and reconstructed using MOB-recon [23]. Assembly quality metrics were computed using QUAST [24], and genome phylogenetic affiliation was confirmed using JSpeciesWS web tools [25]. In addition, tools for filtering fasta sequences according to taxonomy were used for the low identity samples, Autometa 1.0 [26] and MaxBin 2.0 [27] were used. The genome shotgun projects have been deposited in the DDBJ/ENA/GenBank databases.

To gain insights into the genomes, the RAST annotation server [21] was used to identify the subsystem of each genome, providing information on genes associated with various functions, including virulence, pathogenicity, plasmids, and antibiotic resistance. Pathogenicity and virulence analyses were conducted using the PathogenFinder [28] and VirulenceFinder [29] tools from the Center for Genomic Epidemiology. Plasmid identification was performed from the same center using the PlasmidFinder [30] tool. Resistomes were identified using the CARD database Resistance Gene Identifier tool [31].

For the analysis of multi-resistant genomes, the Resistance Genes Identifier (RGI) with the CARD protein database and ResFinder-4.0 were employed to predict resistance genes [31]. Plasmid detection was carried out using the MOB-suite and PlasmidFinder-2.1 [30]. Pathogenicity classification was performed using PathogenFinder-1.1 [28], and virulence factors were determined using VirulenceFinder-2.0 [29]. The serotypes of the E. coli genomes were determined using SerotypeFinder-2.0 [32], and the number of mobile elements was assessed using MobileElementFinder [33].

3. Results

We found similarities between several results, specifically, a similarity of 72.41%.

Table 1. In this table, the samples’ Identification Codes with their respective results for both methods are presented. It is important to mention that with the sequencing technique, it was possible to detect the subspecies of certain samples.

Sample Code	MALDI-TOF Result	JSpeciesWS Result
INTEC BC5 1.1	Enterobacter cloacae	Enterobacter cloacae subsp. Cloacae SMART_901
INTEC BI4 1.1	Enterobacter kobei	Enterobacter kobei 35730
INTEC AC6 1.1	Escherichia coli	Escherichia coli SQ2203
INTEC BI10 1.1	Escherichia coli	Escherichia coli 50816743
INTEC BC4	Escherichia coli	Escherichia coli KOEGE 40 (102a)
INTEC BC8	Escherichia coli	Eascherichia coli O32:H37 str. P4
INTEC AI11 1.1	Acinetobacter baumannii	Acinetobacter baumannii ABBL129
INTEC BI5	Acinetobacter baumannii	Acinetobacter baumannii BR097
INTEC BI9	Acinetobacter baumannii	Acinetobacter baumannii NIPH 67
INTEC AI6	Acinetobacter baumannii	Acinetobacter baumannii UH6507
INTEC AI12	Acinetobacter baumannii	Acinetobacter baumannii BR097
INTEC AI10	Acinetobacter baumannii	Acinetobacter baumannii BR097
INTEC BI15	Acinetobacter baumannii	Acinetobacter baumannii NIPH 67
DC2	Escherichia coli	Pseudomonas monteilii
DC8	Escherichia coli	Escherichia coli
DC10	Acinetobacter pitii	Achromobacter xylosoxidans
EC4	Klebsiella pneumoniae	Klebsiella pneumoniae
EC7	Klebsiella pneumoniae	Klebsiella pneumoniae
FC5	Acinetobacter baumanii	Acinetobacter baumanii
FC7	Acinetobacter baumanii	Acinetobacter baumanii
YNP1-2	Bacillus licheniformis	Acinetobacter pittii
YNP2-2	Klebsiella pneumoniae	Klebsiella pneumoniae
YNP5-3	Pseudomonas aeruginosa	Pseudomonas aeruginosa
YSP2-1	Enterobacter bugandensis	Enterobacter mori
YSP3-2	Serratia marcescens	Salmonella enterica
YSP4	Raoultella ornithinolytica	Raoultella ornithinolytica
YSP4-2	Salmonella spp.	Klebsiella pneumoniae
YSP5	Escherichia coli	Klebsiella pneumoniae
YSP6-2	Klebsiella variicola	Salmonella enterica

shows each sample’s results for the MALDI-TOF and Sequencing techniques. The results that differed between both methods are highlighted and the reasons for these discordances are discussed. It was found that several of the sequences in the Genomics method did not have enough similarity to be assigned to a concrete species correctly, as they had a similarity lower than 98.9% in the phylogenetic analysis performed through JspeciesWS (which is the minimum similarity this tool catalogs as acceptable). Because of this, these sequences were filtrated by taxonomy utilizing MaxBin 2.0 [27] and Autometa 1.0 [26] following the authors’ recommendations.

After filtrating the sequences, only YNP1-2, YSP3-2, YSP4-2, and YSP6-2 were different, as the rest of the results came out to coincide with the MALDI-TOF results, upgrading the similarity percent to 86.2%, which is a much more respectable percentage considering the difference in price and time per sample in the Dominican Republic (see

Table 2. Table indicating cost and time needed to identify the bacteria per sample.

Technique	Cost per Sample	Time per Sample
MALDI-TOF	$7.5 USD	2 days
WGS	$120.00 USD	21 days

).

The contamination in the sequences might have been a manipulation mistake. After the filtering with MaxBin 2.0 and Autometa 1.0, we could see that half of the results were fixed and coincided with MALDI-TOF. The filtered sequences had over 99% similitude with the database sequences.

4. Discussion

This research was conducted over three years using two different techniques to identify bacteria: required culture-based (MALDI-TOF) and whole genome sequencing (WGS). It is essential to mention the number of organisms that each technique covers per sample, as both MALDI-TOF and WGS use one per sample; hence MALDI-TOF is a more efficient and cost-effective option for bacteria identification compared to WGS. As described by [2], many new packages allow scanning for particular antimicrobial resistance proteins inside bacteria using MALDI-TOF [4]. This is very limited at the moment, as MALDI-TOF is a culture-based technique and requires at least eighteen hours for microbe isolation and eighteen to twenty-four more hours for growth to produce enough proteins to be read by MALDI-TOF. However, this technology will eventually allow fast identification of ARGs based on expressed proteins in the cytoplasm [34].

The samples were retrieved and processed successfully. It was possible to correctly identify most of the bacteria through the MALDI-TOF methodology. A contamination problem might be related to a manipulation mistake detected in the phylogenetics assessments, causing some results to differ from the MALDI-TOF results. Nonetheless, it was possible to see the contamination due to JspeciesWS providing a range of acceptable similitude percentages between the submitted fasta files (the samples) and the sequences from the database (reference sequences), as well as other calculated features such as the size and the GC% from the files being too high.

The final results of the MALDI-TOF technique were found to have 86.2% similarity with the results obtained from the whole-genome phylogenetic analysis from JspeciesWS, which is very close to the results described by [35]. We hypothesize that the four results that differed between them could have been caused by the MALDI-TOF database requiring improvement [12]. It was stated this way because the data analyzed by the JspeciesWS tool is continuously updated and improved [25]; due to this, it is important to mention that the high cost of the genomics methodologies justifies itself with its frequent updating and high specificity.

The post-filtering results prove the high efficacy of the MALDI-TOF technique in river water bacterial identification and that it can be used as a low-cost alternative to whole genome sequencing in the Dominican Republic. We expect more improvements in the future that could make it possible to characterize whole microbiomes without the need to invest in nucleic acid sequencing.

5. Conclusions

It was in this study proved that MALDI-TOF can be a low-cost alternative to Whole Genome Sequencing for identifying bacteria sampled from surface river water. This finding is of great importance for developing countries such as the Dominican Republic, in which hiring an external company for sequencing samples represents lower-efficiency approach in terms of both price and time. Our findings can make it possible for these countries to conduct related studies with this major resource facility.

Finally, it is important to state that more uses are being standardized for the MALDI-TOF technique, such as identifying proteins related to antibiotic resistance, and the databases are being improved as well, expanding the scope of MALDI-TOF in microbiology even more.