Quantity of mix buffer appropriate for mosquito head homogenization and MS submission
Five Ae. aegypti (Bora) and five An. coluzzii (Dkr) mosquitoes were used to determine the appropriate quantity of mix buffer to add to the mosquito head for protein extraction before sample homogenization and MALDI-TOF MS submission. The visual comparison of MS spectra according to the volume of mix buffer used indicated a high similarity per species (Figure S1A-B). The mean CCI values from mosquito head MS spectra were elevated, ranging from 0.77 to 0.84, and were not significantly different, per species, whatever the volume of mix used (p>0.05, Kruskal Wallis test) (Figure S1C). The analysis of the MS profiles and CCI values indicated a good reproducibility of head MS spectra independently of the mix buffer volume used for sample homogenization. Interestingly, a slight decrease in MS profile peak intensity was noticed when 40µL of mix buffer was used for both species tested (Figure S1A-B). Then, the addition of 20µL or 30µL of mix buffer seems more adapted. Here, to limit experimental variables, the volume of 30µL, as used for legs homogenization, was chosen to add to the head.
Consequences of homogenization procedures and mosquito body part on MS profiles
Heads, thoraxes (without wing) and legs from Ae. aegypti (Bora) and An. coluzzii (Dkr) were homogenized using either TL, MP or PP method prior to submission to MALDI-TOF MS. The three body parts from 10 specimens per species were tested per homogenization mode. A total of 180 samples generated 720 MS spectra with high intensity, independently of the mosquito body part and homogenization methods. The MS profiles were visually reproducible per body part for each species (Figure 2). Interestingly, the MS patterns seem also species specifics and body part specifics. Cluster analysis using two specimens per species and per homogenization method revealed that all samples from the same mosquito species clustered on the same branch (Figure 3). Samples were grouped per body part for each species, reflecting spectra reproducibility. For each body part, the intertwining of spectra, independently of the homogenization mode, underlined that the homogenization method seems not to impact MS spectra. This cluster analysis suggested that the primary determinant for the MS profiles was the species, followed by the body part, with a singularity of legs MS spectra compared to heads and thoraxes.
A CCI-based analysis confirmed the reproducibility of MS spectra per body part and per species independently of the homogenization mode (Figure 4). Effectively, the mean CCI values of each body part were comparable among homogenization mode for both species. However, higher mean CCI values were obtained for thoraxes, followed by legs and finally by heads for both species. The comparisons of mean CCI values showed significant differences between thoraxes and legs (p<0.0001, Mann-Whitney test), thoraxes and heads (p<0.0001, Mann-Whitney test), and legs and heads (p<0.042, Mann-Whitney test) from Ae. aegypti (Bora). Similarly, significant differences in mean CCI were obtained for An. coluzzii (Dkr), between thoraxes and legs (p<0.0001, Mann-Whitney test), thoraxes and heads (p<0.0001, Mann-Whitney test), and legs and heads (p<0.009, Mann-Whitney test). These results underlined a decrease in MS spectra reproducibility from thoraxes to legs and heads. Interestingly, the low mean CCI-values obtained for pairwise comparisons of MS spectra from two distinct body part for both species, ranging from 0.23 ± 0.06 (mean ± SD) to 0.47 ± 0.10, confirmed that these MS profiles are body part specific (Figure 4).
Efficiency of mosquito identification according to body part and homogenization modes by MS
The MS spectra used for MSP dendrogram analysis were included as reference MS spectra to create the DB1 (Table 1, Additional file 2). Then, each body part (legs, thoraxes and heads) from eight specimens per species (Ae. aegypti (Bora) and An. coluzzii (Dkr)) homogenized by MP, PP or TL, corresponding to a total of 144 samples, were submitted, in quadruplicate, to MALDI-TOF MS and queried against DB1 (Figure 1A). All the samples were correctly classified at the species and body part levels (Figures 5A and 5B). Except for the MS spectra from three An. coluzzii (Dkr) head samples, highly relevant identification scores were obtained (LSVs ≥ 2.0), independently of the homogenization mode used. The elevate LSVs indicated the high quality and reproducibility of MS spectra.
To assess the performances of MALDI-TOF MS for mosquito identification according to homogenization mode, LSVs were compared for each body part and species (Figure 5A-B). No significant difference (p>0.05, Kruskal Wallis test) was noticed between the homogenization modes per body part, except for legs from An. coluzzii (Dkr) (p=0.02, Kruskal Wallis test, Figure 5B). Although LSVs from the legs of An. coluzzii (Dkr) obtained with the automatic mode (TL) were significantly lower than those from MP (p=0.01, Mann Whitney test), identification scores remained highly relevant (LSVs>2.2), preventing misidentification risk. LSVs from legs of An. coluzzii (Dkr) were not different between homogenisation modes.
Interestingly, the comparison of the LSVs per homogenization mode, independently of the body part, revealed no significant differences (p>0.05, Kruskal Wallis test) for both species (Figure S2A-B). Conversely, significant different LSVs were obtained among the body part for Ae. aegypti (Bora) (p<0.001 Kruskal Wallis test) and An. coluzzii (Dkr) (p=0.028, Kruskal Wallis test), independently of the homogenization mode used ((Figure S2C-D). Paired comparisons revealed a significant better matching against DB1 of MS spectra from thoraxes compared to legs (p<0.01, Mann Whitney test) or to heads (p<0.001, Mann Whitney test) from Ae. aegypti (Bora). For An. coluzzii (Dkr), LSVs from thoraxes were also significantly higher than heads (p=0.023, Mann Whitney test). These results indicated that the higher LSVs were obtained with MS spectra from thoraxes followed by legs and heads, confirming the data obtained on MSP dendrogram or CCI analyses.
Duration of sample processing according to body part and homogenization mode
To determine which homogenization method is the more advantageous, the time required for samples processing per homogenization mode was measured and estimated for larger specimen collection as complementary criteria. Then, heads, legs and thoraxes from five Ae. aegypti (Bora) were ground with MP, PP or TL by two experimenters and processing duration were recorded. For both manual modes, sample homogenizations were quicker for heads than for legs, than for thoraxes (Supplementary Table S1). Among the manual grinding modes, the PP method was more than 1.5 less time consuming than MP. However, except when the number of samples to process was very low (i.e., less than five), the automatic sample homogenization mode with TL was largely more rapid than both manual methods, independently of the body part. TL apparatus allows processing from 1 to 48 samples in only 3 minutes, whereas 1 to 36 or until 58 minutes were estimated to grind the same quantity of samples with PP or MP, respectively (Figure 5C). TL was then the faster method for sample homogenization, independently of the manipulator, the number of samples to process, or the body part selected. The automatic procedure seemed to be the more appropriate method for samples homogenization and was used for the successive experiments.
Consequence of mosquito blood meal on MS profiles according to body parts
To assess whether mosquito blood feeding status could affect MS profiles and subsequent mosquito identification, heads, thoraxes and legs from Ae. aegypti (Bora) and An. coluzzii (Dkr) collected kinetically 2, 6, 12, 24, 48 and 72 hours post-engorgement were analysed by MALDI-TOF MS (Figure 1B). MS spectra from heads, thoraxes and legs of not engorged Ae. aegypti (Bora) and An. coluzzii (Dkr) specimens, as well as MS spectra from human blood provided for mosquito meals, were used as the control for MS profile comparisons (Figure S3).
MS spectra of high intensity were obtained for the twenty specimens per species and body part tested at each time point. At 72 hours post-blood feeding, only 10 specimens were tested. The visual comparison of the 660 MS spectra using Flex Analysis v3.4 revealed that, for the vast majority of the samples (upper than 80% of the samples), no apparent change was noticed compared to respective body part and species from unfed specimens (Table S2). In the samples in which MS profile changes were observed, these modifications corresponded to the apparition of MS peaks at about 7,568 m/z and 15,138m/z (Figure S3). These two MS peaks, also present in MS profiles from human blood, were considered as blood contaminants of the mosquito MS spectra. These foreign MS peaks were found in all body parts and 2 to 48 hours post-feeding. Interestingly, the intensity of peaks corresponding to human blood signature decreased with the increasing delay post-blood feeding (Figure S3). This observation could likely be attributed to the digestion process of blood meal. However, this blood signature was more frequently found in the thorax samples (Table S2).
Identification of engorged mosquitoes by MS
To assess the consequences of blood engorgement on the identification of mosquitoes, MS spectra from the 660 samples were queried against DB1 (Figure 1B). The proportion of correct and relevant (LSVs ≥ 1.8) identification reached 96.5% (n=637/660) for MS spectra from both species independently of the body part and delay post-feeding (Figure 6). Among the 23 samples identified without relevant LSVs (i.e., <1.8), 11 belonged to thoraxes of Ae. aegypti (Bora), and 12 from An. coluzzii (Dkr) distributed in heads (n=1), legs (n=3) and thoraxes (n=8). The detection of MS peaks from blood origin was visible in half of them (n=12/23), all from thorax.
The comparison of LSVs between MS spectra with and without blood foreign peaks for each mosquito species revealed a significant decrease in matching scores (p<0.001, Mann-Whitney test) only for MS spectra from Ae. aegypti (Bora) (Figure S4A-B). Nevertheless, the proportion of correct and relevant (LSVs ≥ 1.8) identification for mosquito MS spectra with or without blood foreign peaks remained high, reaching 90.6% (n=116/128) and 97.9% (n=521/532), respectively. Regarding mosquito body parts, MS spectra from thoraxes obtained significant lower LSVs for Ae. aegypti (Bora) (p<0.001, Mann-Whitney test, Figure S4C) and An. coluzzii (Dkr) (p<0.004, Mann-Whitney test, Figure S4D). Although blood foreign MS peaks seem to affect more thoraxes match scoring, correct and relevant LSVs (>1.8) were obtained for the large majority of thoraxes spectra with human blood MS peaks (80.3%, n=49/61).
Impact of mosquito origin on the identification and LSV distribution
To assess whether MS spectra variations occurred for specimens from the same species but from distinct geographical origins, MS spectra from 4 distinct mosquito species, laboratory-reared or field collected, were queried against DB2 and DB3. The median LSVs against DB2 were 2.21, 2.32 and 2.36 for heads, legs and thoraxes, respectively, irrespective of the species analysed. The distribution of LSVs varied significantly between body parts (p=0.002, Kruskal Wallis test), with the heads obtaining the lowest scores (Figure S5). The proportion of correct and relevant (LSVs>1.8) identification against DB2 ranged from 79.0% for heads to 83.9% for legs (Table 2). The query of these MS spectra against the DB3, upgraded with MS spectra from field specimens, did not improve significantly (p>0.05, Chi-square tests), the proportion of correct and relevant identification. Conversely, LSVs obtained per body part per field species were significantly improved between DB2 and DB3 for nearly all paired comparisons (p<0.05, Mann-Whitney test, Figure 7). Interestingly, misidentification concerned mainly MS spectra from An. gambiae s.l., underlining the difficulty to classify specimens from species complex.