Microbiota identified from preserved Anopheles

Colonized mosquitoes were obtained from the Botha de Meillon Insectary, National Institute for Communicable Diseases (NICD), Johannesburg, South Africa. Two Anopheles species were used in this study: An. arabiensis (MBN colony) and An. funestus (FUMOZ colony). The MBN colony has mosquitoes from Mamfene, KwaZulu-Natal, South Africa, while the FUMOZ colony has mosquitoes from southern Mozambique [96, 97]. Mosquitoes (20 female and 20 male per species per repeat; three biological repeats) were collected between 0- and 24-h post-emergence (here forth called fresh samples). Additionally, 1 ml of each species’ larval rearing water was collected (three biological repeats).

Field-caught An. arabiensis were collected between June and August 2019 from Mamfene. These samples had been preserved on silica in microcentrifuge tubes and were retrieved from the departmental archive. Species and Plasmodium-infection status were confirmed using multiplex polymerase chain reaction (PCR) [98‐100] and enzyme-linked immunosorbent assay (ELISA) [101], respectively, as part of a departmental Sterile Insect Technique (SIT) project. Samples were retrieved after 8, 12 and 16 weeks of preservation.

As mosquito density varies per season, field-collected mosquitoes were low in numbers because June, July and August are the winter months in South Africa (collection numbers are highest in summer and lowest in winter [102]). Mosquito density may have also been exacerbated by the extensive drought in South Africa, which probably left mosquitoes without local breeding pools ([103]: filter between June and August 2019 to view the lack of rainfall). Therefore, only four females were retrieved per time-period, and bacterial identification was performed using culturomics due to the low cost for the limited number of samples. Additionally, as this sample size was low, it was not used to represent the wild mosquito population, and comparisons between field-collected and colonized mosquitoes could not be made.

Colonized mosquitoes were preserved on silica or in RNAlater^® (20 females and 20 males per species, repeat, preservative, and preservation time-period; three biological repeats). Mosquitoes were preserved for 4 and 12 weeks per preservative. Prior to preservation, mosquitoes were immobilized at − 20 °C for 2 min. For silica preservation, mosquitoes were placed individually in 1.5-ml microcentrifuge tubes containing approximately five silica beads (silica gel blue self-indicator (copper sulphate-based), B&M Scientific, South Africa; cat no. CSGB0002) and were separated from silica using a piece of paper (Fig. 1a).

For RNAlater^® preservation, mosquitoes were surface sterilized in 70% ethanol (v/v), left to dry at room temperature on a sterile piece of Kimwipe^® (Kimberly-Clark, TX, USA; cat no. 34155), and individually submerged in 0.5 ml of RNAlater^® solution (Qiagen, Germany; cat no. 76106) in 1.5-ml microcentrifuge tubes (Fig. 1b). Samples preserved in RNAlater^® were stored according to the manufacturer’s instructions (samples preserved for 4 weeks were stored at 4 °C, while samples preserved for 12 weeks were placed in RNAlater^® overnight at 4 °C and subsequently stored at − 80 °C). For each species, a set of fresh (non-preserved) samples was collected for comparison. All supplies (silica beads, microcentrifuge tubes, separating paper, etc.) were sterilized prior to use, and swabs of these supplies were cultured to ensure they were not contaminated.

Midgut dissections of fresh samples were performed aseptically per protocol by the WHO [104], where 4 µl of phosphate-buffered saline (PBS) was used per midgut; 20 midguts were pooled per species, sex, preservative, preservation time-period, and repeat.

Due to desiccation, mosquito abdomens were shrivelled and brittle. Therefore, mosquitoes were surface sterilized twice in 70% ethanol (v/v), and abdominal segments I to V were dissected and placed in sterile microcentrifuge tubes containing 4 µl of PBS; 20 abdominal segments were pooled per sex, species, preservative, preservation time-period, and repeat. Abdominal segments were then homogenized using a TissueLyser II, followed by centrifugation. Negative controls were set up per group of samples using PBS were carried through during downstream analyses. Additionally, abdomens tore apart easily when mosquitoes were submerged in RNAlater^® and midguts could not be isolated. Thus, mosquitoes were removed from solution and blotted on tissue paper to remove excess RNAlater^®, followed by the adapted dissection method used for silica-preserved samples.

Each homogenate (10 µl) was plated on individual selective media agar plates (Table 1). Inoculates were plated and aerobically incubated for a minimum of 16 h at 37 °C. Plates without observable colonies were re-incubated for 16 h to account for slow-growing bacteria. Following incubation, isolates were distinguished morphologically, and distinct colonies were selected for repeated sub-culture by re-inoculation on fresh primary agar plates and incubation for a minimum of 16 h at 37 °C. Plates without observable colonies were re-incubated for 16 h to account for slow-growing bacteria. Negative controls were set up for each plate type, and during subsequent incubation periods, to ensure plates were not contaminated during incubation. Each plate type was also inoculated with PBS negative controls from dissections.

Each colony was placed directly on an individual spot on a 96-spot reusable MALDI-TOF target plate (Bruker Daltonics, Wissembourg, France; cat no. 8280800). Each spot was covered with 1 μl of α-Cyano-4-hydroxycinnamic acid (HCCA) matrix (Bruker Daltonics, Wissembourg, France; cat no. 8255344) diluted in standard solvent (50% acetonitrile: 47.5% water: 2.5% trifluoroacetic acid, Sigma-Aldrich, Lyon, France; cat no. 19182). The matrix was allowed to dry at room temperature, and the target plate was placed in the MALDI Biotyper^® with benchtop microflex™ LT/SH mass spectrometer (Bruker Daltonics, Germany). A bacterial test standard (Bruker Protein Calibration Standard I, Bruker Daltonics, Wissembourg, France; cat no. 8255343) was used according to the manufacturer’s instructions to control for loading and matrix. Spectra were compared with the MBT 7854 MSP Library database installed on the computer (Bruker Daltonics, Wissembourg, France; ref no. 182903). An isolate was identified when spectra had a log score value ≥ 1.9 [81]. Every unidentified isolate was tested successively, where a portion of the same colony was placed on a new spot on the target plate and identified as described.

To supplement culturomics, midguts of fresh mosquitoes and abdomens of preserved mosquitoes (preserved on silica and in RNAlater^® for 4 and 12 weeks) were dissected and pooled (20 female and 20 male per species, preservative, preservation time-period, and repeat; three biological repeats) in sterile PBS as described. Additionally, 1 ml of each species’ larval rearing water was collected (three biological repeats). Bacterial DNA was extracted using the QIAamp^® DNA Microbiome Kit according to the manufacturer’s instructions (Qiagen, Germany; cat no. 51704). Negative controls were set up during each stage of DNA extraction, and extraction was also performed on PBS negative controls from abdominal dissections. Prior to sequencing, DNA quality and purity were measured using the NanoDrop™ 2000c spectrophotometer (Thermo Scientific, MA, USA; cat no. ND-2000C). Due to the high cost of NGS and the high number of negative controls, only experimental samples were sequenced. Nonetheless, negative controls were assessed using the NanoDrop™ 2000c spectrophotometer and were cultured to ensure no contamination.

Samples were sent to Macrogen Europe (Amsterdam, The Netherlands) for 16S rRNA gene sequencing targeting the V3-V4 regions with universal primers, Bakt 341F and Bakt 805R [110]. The Illumina MiSeq system was used to perform paired-end sequencing, and the Fast Length Adjustment of SHort reads (FLASH version 1.2.11) program was used to assemble reads [111]. Pre-processing (denoising) and clustering of sequences were performed with the CD-HIT-OTU and rDnaTools programs [112, 113]. Diversity analyses and taxonomy assignments were performed with the Quantitative Insights Into Microbial Ecology (QIIME) program (see Additional file 1 for OTUs with taxonomy assignment) [114].

According to the MALDI-TOF MS MBT 7854 MSP Library, several bacteria are genetically indistinguishable from one another. Thus, the matching hints section of the library was used to compare MALDI-TOF MS and NGS results (see Additional file 2). Accordingly, indistinguishable bacteria were grouped, and a single bacterium was used to represent indistinguishable bacteria (Table 2). Representative bacteria were chosen based on which indistinguishable bacteria were also present in NGS results.

For fresh and field-collected samples, results are presented as accumulative data across replicates. This is because sample sizes for these groups were lower than the sample sizes of preserved mosquitoes due to the inclusion of preservatives and preservation time-periods. There was also a contrast in microbial composition between fresh and preserved samples identified by culture-independent techniques, which may be attributed to potential contamination. Thus, when comparing group, for example, sex (female or male) irrespective of species (An. arabiensis or An. funestus), condition (fresh or preserved), preservative (silica or RNAlater^®), preservation time-period (4 weeks or 12 weeks) or technique (culture-dependent or culture-independent), only commonly recurring bacteria (bacteria that appeared in at least 50% of replicates in the groups being compared) were reported.

Species richness, bacterial diversity, relative abundance, and evenness were calculated. Species richness was measured as the number of operational taxonomic units (OTUs) per sample. Indices were reported per replicate, and the mean index per group was calculated (Additional file 3). Diversity was measured using the Shannon–Wiener diversity index (H): the higher the value of H, the more diverse the community [118]. To estimate relative abundance as a measure of species dominance, Simpson’s reciprocal index (1/D) was calculated. 1/D measures the probability that a randomly selected species is the dominant species, where a score of one indicates the community is dominated by a single species [119]. E was used to estimate evenness, which ranges from zero to one with zero signifying no evenness and one signifying complete evenness [120].

Statistical analyses were performed at a 95% confidence interval assuming a 5% level of significance using STATA/IC version 16.1. As the data were not normally distributed (as per Shapiro–Wilk tests), non-parametric statistical analyses were performed. Two-sample Wilcoxon rank-sum (Mann–Whitney) tests were used to determine if diversity indices differed between (i) fresh females and males; (ii) fresh An. arabiensis and An. funestus; (iii) fresh mosquitoes and mosquitoes preserved for 4 weeks; (iv) fresh mosquitoes and mosquitoes preserved for 12 weeks; (v) mosquitoes preserved for 4 weeks and mosquitoes preserved for 12 weeks; (vi) silica preservation and RNAlater^® preservation; and, (vii) culture-dependent and culture-independent techniques.

Anopheles arabiensis were predominantly colonized by Proteobacteria, irrespective of technique, followed by Bacteroidetes (Fig. 2a). While culture-dependent results estimated Elizabethkingia as the dominant genus in both An. arabiensis sexes, culture-independent results identified Serratia as the dominant genus (Fig. 2b–e).

Culture-dependent results identified Serratia fonticola as a commonly recurring bacterium, but S. fonticola was not identified by NGS. Thus, it is likely that S. fonticola identified by culturomics was actually Serratia oryzae because S. oryzae was identified as a commonly recurring bacterium by NGS and S. oryzae cannot be detected by the MBT 7854 MSP Library. Results of a pairwise alignment (performed using National Center for Biotechnology Information’s (NCBI) Basic Local Alignment Search Tool for nucleotide sequences (BLASTn) [121]: http://​www.​ncbi.​nlm.​nih.​gov/​BLAST/​) of 16S rRNA sequences revealed 96.75% sequence similarity between S. fonticola (accession number: CP011254.1) and S. oryzae (accession number: NR_157762.1). Meanwhile, a pairwise alignment between S. fonticola and Serratia liquefaciens (accession number: UGYL01000001.1) and S. fonticola and Serratia marcescens (accession number: CP063354.1; the other Serratia species identified by NGS in this study), revealed 94.03% and 85.53% sequence similarity, respectively. Therefore, S. oryzae was used to represent S. fonticola throughout this study.

Aeromonas hydrophila, E. anophelis and S. oryzae were identified as common bacteria between female and male An. arabiensis and their larval rearing water, irrespective of identification technique (Fig. 3). Staphylococcus epidermidis was common between both sexes and the larval rearing water in culture-dependent results, but only in males and the larval rearing water in culture-independent results. Raoultella ornithinolytica was common between both sexes in culture-dependent results, but only in females in culture-independent results. Some bacteria were identified in a sex- and/or technique-dependent manner (see Additional file 4 for an overview of replicates).

Proteobacteria remained the dominant phylum in preserved An. arabiensis (Additional file 5). Elizabethkingia anophelis and S. oryzae were recurring bacteria that were common in silica-preserved and RNAlater^®-preserved An. arabiensis (Table 3; see Additional file 6 for replicate details). Irrespective of condition, E. anophelis and S. oryzae were recurring bacteria that were common between females and males. Enterobacter cloacae was specific to females and S. epidermidis was specific to males (Table 3; see Additional file 7 for replicate details). These sex-specific bacteria were also recurring bacteria in RNAlater^®-preserved (and not silica-preserved) samples. Overall, E. anophelis and S. oryzae were common between fresh and preserved female and male An. arabiensis.

Field-collected An. arabiensis were predominantly colonized by Firmicutes (Fig. 4a). The dominant genus in field-collected An. arabiensis was Staphylococcus (Fig. 4b). Although Firmicutes was the dominant phylum in samples preserved for 8 and 12 weeks, samples preserved for 16 weeks were predominated by Proteobacteria (Fig. 4c). Staphylococcus epidermidis and Staphylococcus hominis were common between samples preserved for 8 and 12 weeks while samples preserved for 16 weeks did not have bacteria in common with samples preserved for 8 and 12 weeks (Fig. 4d).

Anopheles funestus were predominantly colonised by Proteobacteria, irrespective of technique, followed by Bacteroidetes (Fig. 5a). While both identification techniques agree that Serratia is the dominant genus in females, culture-dependent results represent Aeromonas as the dominant genus in males and culture-independent results represent Elizabethkingia as the dominant genus in males (Fig. 5b–e).

Aeromonas hydrophila was identified as a common bacterium between An. funestus females and males and their larval rearing water, irrespective of identification technique (Fig. 6). Serratia oryzae was common between both sexes in culture-dependent results but common in both sexes and the larval rearing water in culture-independent results. Staphylococcus epidermidis was common between both sexes and the larval rearing water in culture-dependent results but only in males and the larval rearing water in culture-independent results. Some bacteria were identified in a sex- and/or technique-dependent manner (see Additional file 8 for an overview of replicates).

Proteobacteria remained the dominant phylum in preserved An. funestus (Additional file 5). Serratia oryzae was a recurring bacterium that was common in silica-preserved and RNAlater^®-preserved An. funestus (Table 4; see Additional file 9 for replicate details). Irrespective of condition, S. oryzae was a recurring bacterium common between females and males. Enterobacter cloacae was specific to females and S. epidermidis was specific to males (Table 4; see Additional file 10 for replicate details). These sex-specific bacteria were also recurring bacteria in RNAlater^®-preserved (and not silica-preserved) samples. Overall, S. oryzae was a common bacterium between fresh and preserved An. funestus.

Collectively, An. arabiensis and An. funestus were predominantly colonized by Proteobacteria (Additional file 11), irrespective of sex, condition or identification technique (see Additional file 12 for replicate details). Both species were also predominantly colonized by bacteria belonging to the Elizabethkingia and Serratia genera. Elizabethkingia anophelis and S. oryzae were recurring bacteria common between both species irrespective of sex, and E. cloacae was dominant in An. funestus. Enterobacter cloacae was specific to females, while S. epidermidis was specific to males, irrespective of species.

Males of both species had higher bacterial species richness than females irrespective of species, condition or identification technique (Table 5). Nonetheless, An. arabiensis and An. funestus had overall comparable species richness.

There were no significant differences amongst diversity indices between female and male mosquitoes (Additional file 13, 6A–C). There were also no significant differences amongst H and E indices between An. arabiensis and An. funestus. There was a significant difference in 1/D between species (P = 0.0209), where An. arabiensis had a higher 1/D than An. funestus (Additional file 13, 6D–F).

Diversity indices were comparable between fresh mosquitoes and mosquitoes preserved for 4 weeks (Additional file 14, 7A–C), fresh mosquitoes and mosquitoes preserved for 12 weeks (Additional file 14, 7D–F), and mosquitoes preserved for 4 and 12 weeks (Additional file 14, 7G–I). There were no significant differences amongst diversity indices between preservatives (Additional file 14, 7J–L). There were, however, significant differences in diversity indices between identification techniques. Culture-independent techniques estimated higher H (P = 0.0200) and 1/D (P = 0.0053) indices than culture-dependent techniques (Fig. 7A-B). Culture-dependent techniques estimated a higher E index (P = 0.0053) than culture-independent techniques (Fig. 7C).