In Vitro Assay

A High Throughput Method to Test Cry Toxin Activity

Cry toxins have been found to have activity towards insects of different phylogenetic orders. In short, a functional Cry protein is only active when specific binding occurs to the gut membrane of the target insect. In more detail, these receptors have been found in the gut epithelium cells, where the Cry proteins polymerize binding to its membrane receptor forming pores. After the formation of these ion channels, the osmotic pressure is disturbed resulting in cell death of the target host^1,2.
Although a number of Cry toxins have been (partly) characterized, a toxin specific to Varroa destructor has yet to be found. For this reason, we developed an in vitro toxicity test in order to be able to efficiently screen Cry toxins on their activity against V. destructor. From the V. destructor, Brush Border Membrane Vesicles (BBMV's) were made and loaded with the fluorophore 6-carboxyfluorescein to test the pore formation ability of Cry proteins, see Figure 1a. As a proof of principle, BBMV's from the gut of Tenebrio molitor were made and loaded with 6-carboxyfluorescein to test the pore formation ability of Cry3Aa, which is known to be toxic to T. molitor larvae^3,4.

BBMV's are composed of phospholipid double layers originating from midgut membrane cells. This membrane therefore should also contain the membrane proteins of the midgut membrane. After incorporation of 6-carboxyfluorescein into the BBMV's the presence of an active Cry toxin can principally be detected as followed: the Cry toxins will bind to their highly specific receptors⁵ and consequently create holes into the phospholipid double membrane layer, see Figure 1b. After this event, the fluorophores leak out of the vesicles due to diffusion, which can be measured as an increase in fluorescence, see Figure 1c. This method has been proven to work for the activity of Cry3Aa on Leptinotarsa decemlineata⁶. Rausell et al. only demonstrated how the magnitude of the fluorescence increase links to toxicity, but this method has its limitations. BBMV's have the disadvantage of being unstable and undergoing degradation⁷. In other words, measurements on BBMV's can be influenced by spontaneous breaking of the vesicles. In our project, we took this into account by relating toxicity activity to a kinetic value.

The increase of fluorescence due to fluorophore leaking can be attributed to dequenching Dequenching refers to any process which increases the fluorescence intensity of a given substance. . 6-Carboxyfluorescein is a self-quenching fluorophore. At a high concentration 6-carboxyfluorescein forms dimers. Because of dimer-dimer and dimer-monomer interactions, the fluorescence of a carboxyfluorescein molecule decreases drastically⁸. The vesicles created in our study contain a high concentration of fluorophores, which results in self-quenching behavior. When the fluorophores leak out of the vesicles, the local concentration of fluorophores decreases. Hence, dequenching occurs and the fluorescence increases.

Analysis of Brush Border Membrane Vesicles

To visualize the presence of BBMV's, images were made with the use of TEMTransmission Electron Microscopy . This was done in collaboration with the iGEM team from Delft

Figure 2 shows the vesicles - round shapes with sizes between the 100 nm and 300 nm. To gain more information about the composition of the samples that are visualized above, dynamic light scattering was performed. The dynamic light scattering results (Figure 3) give information about the size distribution of the particles that are present in the solution.

Looking at both transmission electron microscopy results and dynamic light scattering results, the T. molitor samples have a large population of vesicles that vary their size around 200 nm. Looking at the V. destructor samples, the population around this size is smaller. To learn more about the proteins in these samples, click here.

Fluorophore Leaking Experiments

The hypothesis is that due to addition of a detergent, the phospholipid bilayer will rearrange. This process will subsequently give the fluorophores the chance to escape their enclosed space and disperse in the solution. The dispersion of the fluorophores into the solution would then cause an increase in fluorescence as described previously.

To determine whether these BBMV's were capable of releasing fluorophores and whether this release could be measured, the following was done: to BBMV's containing carboxyfluorescein, 1 volume percent of SDS (sodium dodecyl sulphate) was added to disturb their stability. Meanwhile the fluorescence was measured over time.

Looking at Figure 4, after addition of SDS (detergent), an increase in fluorescence is indeed visible. However, the same increase in fluorescence can be observed over time with vesicles that are in an environment without SDS. Two possible explanations for this phenomenon are that BBMV's are already porous or are not stable under the circumstances during measurement. Both processes would result in an increase in fluorescence over time. In order to compare induced fluorophore leaking from standard fluorophore leaking, the speed of the carboxyfluorescein leaking should be calculated and compared.

Obtaining Kinetic Values

To determine whether a specific Cry protein can induce fluorophore release, Cry3Aa was isolated from B. thuringiensis and tested on BBMV's obtained from T. molitor. Activated Cry3Aa gives a clear band on a gel of a 68 kDa protein after SDS-PAGE was performed, as described by Rausell et al⁶. The SDS-PAGE gel of the protein extract from B. thuringiensis, shown in Figure 5, shows such a band of approximately this size.

T. molitor BBMV's were incubated with and without Cry3Aa and the fluorescence was measured over time(Figure 6). From this graph it can be concluded that the presence of Cry3Aa results in a faster increase in fluorescent and therefore in a faster fluorophore release. To determine how large the role of induced leaking is, the kinetic properties of both measurements have to be compared. In order to compare these properties, it should be known which kinetic rules the fluorophore leaking follows. Plotting the natural logarithm of the concentration of fluorophores entrapped in vesicles against time, gives a straight line, (Figure 6). The concentration of fluorophores entrapped in vesicles is expressed as the difference of the maximum fluorescence and the fluorescence at a certain time point. The observed straight line in this graph tells us that first order reaction rules can be applied on the measurement⁹. This means that equation 1 can be applied to the obtained data.

The reaction rate constant, k_f, determines how fast the reaction goes. This constant can be used to compare one reaction with another. To find this constant, we plot equation 2 directly through the obtained data. The equation is also shown in Figure 6.

For both samples, 6 technical replicates were measured. Representative measurements of each sample are shown in Figure 6. All data can be found here. For all measurements, the reaction rate constants were determined as explained above and are shown in Figure 7. This figure shows that in the presence of Cry3Aa, the k_f values are higher and thus the fluorophores are released faster. Overall, fluorophore leaking in the presence of Cry3Aa is accelerated with a factor of 1.59 compared to fluorophore leaking in the absence of Cry3Aa. Unless one of the other few proteins still present in the sample would unexpectedly interfere with the experiment, Cry3Aa induces leakage of carboxyfluorescein from T. molitor BBMV's.

A library of Cry3Aa proteins with mutations in the three different binding sites was made to create a Cry toxin that is potentially toxic to V. destructor. When one of these mutant Cry3Aa proteins is toxic to V. destructor, the toxin should induce leakage of carboxyfluorescein from V. destructor BBMV's, and therefore have a higher k_f value than the negative control. The influence of Cry3Aa on T. molitor vesicles can be used as a positive control. The process of optimization of the positive control can be found in the notebook. It should be noted, however, that the relationship between the effect of the Cry toxin on vesicles and the Cry toxin on the whole organism might be different for each organism. To know with certainty whether a certain Cry toxin is lethal for V. destructor, an in vivo test should be performed.

Toxin Engineering

As mentioned above, no specific toxin to Varroa mites is known. To overcome this limitation, we tried to engineer one. We used the Cry3Aa toxin as a starting point.This toxin is known to be effective against a number of Coleoptera as the Colorado potato beetle Leptinotarsa decemlineata and the mealworm T. molitor. This toxin was also shown to be active against 4 different types of mites (Acarus siro L., Tyrophagus putrescentiae, Dermatophagoides farinae Hughes, and Lepidoglyphus destructor)¹. This engineering focused on two different strategies:

1. Modification of the previously described motifs within the Cry3Aa structure –Random mutagenesis of the binding sites, and
2. Finding specific binding motifs to the receptors within the mite gut – Phage Display

Binding Site Mutagenesis

Design

In 1996, Rajamohan et al. demonstrated that mutations in the binding sites of Cry toxins can both decrease and enhance the specificity of a toxin towards its target organism². Knowing this, the 3D structure of the Cry3Aa toxin has been analysed. Three putative binding sites have been identified. These are shown in in the following table.

To mutate these binding sites, the plasmid pSB1A3_Cry3Aa_TEV_HIS has been used as a template. This plasmid contains the gene encoding wildtype Cry3Aa toxin (B. thuringiensis var. tenebrionis, Taxonomy ID: 1444). Its expression is regulated by the arabinose inducible promotor pBAD/araC (Bba_I0500). From this plasmid, three different plasmids were created with randomized sequences within the binding sites. By this we expect that each individual colony harbouring this plasmid containing an individual variants, in other words, a library of Cry3Aa proteins with randomly binding sites to be screened.

Screening

The mutants have been screened following the hightroughput procedure developed and described above. In short, random colonies were picked from each of the binding mutants (n=144) for toxin expression and from these (n=24) were selected for activity testing. As a negative control, cell free extract from empty expression strain, BL21(DE3) has been used. The in vitro toxicity assay results for these 24 tested mutants compared to the Cry3Aa expressing strain and the negative control are shown in Figure 8. Figure 9 shows a heatmap of the tested mutants' activities.

From these results, it can be concluded that the third binding site (amino acids 410-416) seems to be a good candidate for future engineering and specificity adaptation of this particular Cry toxin. Due to the relatively high deviation in reaction speed for the toxins 3.3.3 and 3.3.7, these should not be taken into account, as they are rather inconclusive. This leaves us with one proper candidate – mutant Cry3.3.5. This mutant has both a higher mean reaction velocity compared to the reference value and a relatively narrow range of measured relative k-values obtained for the 9 tested samples. Furthermore, the negative control shows that other proteins present in the cell free extract or cell debris (e.g membrane residue) interfere with our reading. Nevertheless, our preliminary results show that the mutant revokes a stronger response than the negative control, suggesting a specific binding to the Varroa vesicles.

Phage Display

Due to difficulty to obtain mites, the phage display experiments were set up in a model organism where a Cry toxin has already been described: T. molitor (Cry3Aa). We expected to prove the applicability of this technique for specific protein engineering towards toxicity for V. destructor. We used the filamentous bacteriophage M13 and two different libraries; 7-mer flanked by a pair of cysteine residues (The Ph.D.™-C7C Phage Display Peptide Library) and the 12-mer (The Ph.D.™-12 Phage Display Peptide Library).

Display of M13KE-PIII Libraries

Two different strategies for phage display were used, in vivo (conventional phage display) and in vitro (panning). For the in vitro display, a modified biopanning protocol, the Biopanning and Rapid Analysis of Selective Interactive Ligands (BRASIL) method, was used.

in vivo

First as a proof of concept feeding experiments were performed with fluorescein on mealworms and mites (Figure 10). In short, the mites and mealworms were fed with either a piece of carrot soaked with fluorescein or a bee larvae soaked with fluorescein. After washing with PBS buffer, the insect were ground up and fluorescence was measured.

After seeing that the mites and mealworms could take up fluorescein, the phage libraries were fed to T. molitor and V. destructor and the bound phages recovered for further use. Due to a lack of mites, only two rounds of mite in vivo display could be performed. For the mealworms, the titer (the amount of plaque forming phages per mL) decreased drastically per round Figure 12. Thus, a second run with propagation of phages between the round has been performed.

During this second round, an unexpected incident caused us to stop the in vivo phage display. While tittering a phage library obtained from a mealworm, no plaques with the expected morphology (blue, lysogenic) showed up on the /plate. Instead, a M13 plaques, several lytic virus were recovered. The experiments have thus been stopped due to a risk of contamination in the samples. To investigate the source of the contamination, all possible sources have been looked at: Used buffer, used top-agar, and the plates. The source of the unexpected plaques turned out to be the mealworm gut. Kelly et al. found that phages can be found in mealworms naturally³. When using guts of mealworms that have not been fed any phages prior to the process, still a lot of plaques of different kinds could be seen on the titer plates. From this, we could conclude that there are naturally some bacteriophages that E. coli can be a host for in the gut of T. molitor. While those phages could be of interest for general bacteriophage studies, or further specific binding studies, they do interfere with our phage display experiments in two ways:

1. Knowing that phages are already inside of the gut (in relatively high amounts) they might block the receptors for our phage libraries.
2. As the phages and their properties are not known, working with them might not be safe As there was a shortage of living mites for the in vivo display anyway, we decided to switch to an in vitro phage display method.

Nevertheless, the results obtained up to then in the mite phage display experiments have been analyzed and the following binding site motif was found:

in vitro

For an in vitro assay, vesicles of the T. molitor midgut and V. destructor have been produced, according to the protocol. In short, cells were lysed, lipids were separated and then left for reassembly. These vesicles should have no phages attached on the outside, due to the way they were prepared. The only chance of having phages that are naturally occurring in mealworms or Varroa mites was thus that they had been trapped inside of the vesicle during the preparation steps and so relatively low. Unfortunately, the titer of phages bound to the vesicles prepared from V. destructor decreased dramatically after the first round of display and dropped to 0 after the second round. This indicates that the phages did not bind to the vesicles in a way that is strong enough to keep them attached during the separation step of vesicles with attached phages from phages that have not bound. The reasons for this might be investigated and the protocol used for this phage display might still be optimized.

Varroa Isolates

Staining Isolates

To provide an alternative to currently available pesticides, we wanted to adapt Cry toxins to kill V. destructor. Cry toxins, also known as Bt toxins, have been used commercially for decades and can limit the environmental impact of pesticide use¹. Cry toxins are produced by the insect pathogen B. thuringiensis. Their mode of action can make them highly specific to certain species of insects.

While over 800 Cry toxins are known, research on the workings of these toxins mainly focuses on conventional pests such as the Colorado potato beetle, cabbage moths and mosquitoes. Field experiments on non-target invertebrates show that AcariThe scientific name for mites and ticks. are not significantly affected by Bt crops Transgenic crops which express Cry toxins, such as Bt cotton or Bt maize.³, but these use only a few known Cry toxins. Other studies have shown that there are strains of B. thuringiensis which are harmful to Acari. While most of these strains or toxins were tested against herbivorous species of mites^4,5,6,7, some are known to kill species of ticks^8,9,10,11. In fact, two strains have been discovered which show toxicity to V. destructor^12,13! Unfortunately, they did not identify a particular Cry toxin.

To find our own Cry toxin, we adapted a procedure from Rampersad et al.¹⁴ and Alquisira-Ramírez et al¹². We gathered over 800 dead mites from the Dutch foundation "De Duurzame Bij"¹⁵, Bennekom, The Netherlands.

Our method of sample preparation selected for spore formation, but not the B. thuringiensis rod-shaped morphology. Therefore, 106 colonies from the mites were investigated with brightfield microscopy. They were stained with Coomassie Brilliant Blue to make Cry toxins more visible. Five isolates were identified as Bacillus-like species. Some of them, including B. thuringiensis HD-350, are shown in Figure 15.

In Vivo Toxicity

We wanted to test the strains we identified for their toxicity to V. destructor. Protein extracts and SDS-PAGE were performed. The toxicity assay was based on the assay used by Alquisira-Ramírez et al.¹², but we faced one major problem: the assay, including the negative control, killed all of the Varroa mites. An alternative would be to use an assay where we dip honeybee larvae in protein extracts from the strains; a Varroa mite which feeds on the larvae should then ingest any toxins. Unfortunately, Varroa mites tend to share feeding sites and use them repeatedly¹⁶, which means that not all mites would consume the protein extract.

A sample-size calculation¹⁷ was performed in R with the following code to assess the feasibility of such an assay.
> p=0.66
> p0=0.5
> alpha=0.05
> beta=0.1
> (n=p*(1-p)*((qnorm(1-alpha)+qnorm(1-beta))/(p-p0))^2)

Here, p is the alternative hypothesis, p0 the null hypothesis, alpha the type I error and beta the type II error. The values of these variables were based on the following assumptions:

• The null hypothesis is 50% mite mortality. Natural mite mortality is very high in the laboratory.
• Linea Muhsal estimated during feeding experiments that 3 out of 18 mites ingested the fluorophores she had dipped honeybee larvae in.
• If an effective Cry toxin is ingested, we expect to see 100% mite mortality. This is an optimistic assumption.
• 16% of mites eat a toxin which causes 100% mite mortality, so if there is a toxin, we should see 66% mite mortality.

With these assumptions, a sample size of 75 was calculated. We were generally able to gather a maximum of 40 mites per week. Gathering enough mites to test all our samples was unfeasible. Therefore, pursuit of this type of toxicity assay was cancelled in favour of further development of an in vitro assay.

16s rRNA sequencing and SDS-PAGE results

We sequenced the 16s rRNA for all five isolates with Bacillus-like morphology. A search of NCBI’s 16S ribosomal RNA sequences database resulted in the following alignments, each with approximately 99% identity of their best match. The isolates and their matches are displayed in Table 1.

While it narrows down the number of candidates, 16s rRNA sequencing is not a reliable method to distinguish between Bacillus species¹⁸. Therefore, an SDS-PAGE gel of the protein extracts was prepared to further analyse them. This gel is shown in Figure 16; Bacillus subtilis and Bacillus thuringiensis were included as controls.

Proteomics analysis of isolate V82

Based on the previous results, isolate V82 had the most potential to be a putative mite pathogen. It is closely related to a known insect pathogen, Lysinibacillus sphaericus, which is known to produce a 100 kDa mosquitocidal toxin¹⁹. Our isolate also showed a strong band in the 100 kDa region (Figure 16). A new SDS-PAGE gel was prepared with independently grown V82 cultures and bands were cut out for peptide extraction, so the 100 kDa protein could be analysed with LC-MS/MS. Negative control bands were cut out approximately ~1 cM above the 100 kDa bands. MaxQuant²⁰ was used to analyze the data.

Analysis of the LC-MS/MS results is shown in Figure 18, a volcano plot. The x-axis plots the protein abundance ratio between the 100 kDa band and its respective control band, while the y-axis plots the p-value of this difference in protein abundance. Therefore, a protein in the left half of the plot is more abundant in the control bands, while the opposite goes for proteins in the right half of the plot. The higher it is on the y-axis, the more significant this difference is.

Contaminations such as keratins and the trypsin used to digest the samples should not differ significantly between the samples. This is also shown in Figure 18, as most contaminants (the orange data points) are present in the bottom half of the volcano plot. The relative label-free quantification (LFQ) intensity of the significant proteins shown in Figure 18 was very low, in all cases lower than the LFQ intensity of trypsin. From a biological perspective, this makes sense; there is no reason for an organism to produce mass amounts of alanine tRNA ligase, which was one of the significant peptide matches. However, this and the fact that only 3% of peptides could be matched indicates that the 100 kDa protein was not present in the prepared Lysinibacillus sequence database. Therefore, the peptides were also matched against a Bacillus database. If the 16s rRNA sequence was wrong, this could be a more suitable database. However, the expanded database yielded similar results. Genomic DNA was extracted from V82 and sent for sequencing so the LC-MS/MS data could be run against the isolate's own DNA. This genomic DNA was also analyzed with the Toxin Scanner.

References

References for in vitro essay

1. Alejandra Bravo, Sarjeet S. Gill, Mario Soberón. Mode of action of Bacillus thuringiensis Cry and Cyt toxins and their potential for insect control. Toxicon. 2007 Mar 15; 49(4): 423–435. ↩

2. Hideo Ohkawa,Hisashi Miyagawa,Phillip W. Lee. Pesticide Chemistry: Crop Protection, Public Health, Environmental Safety. Wiley-VHC Verlag GmbH & Co KGaA Weinheim. 2007. p.193 ↩

3. Jeff Fabrick, Cris Oppert, Marce´ D. Lorenzen, Kaley Morris, Brenda Oppert, and Juan Luis Jurat-Fuentes. A Novel Tenebrio molitor Cadherin Is a Functional Receptor for Bacillus thuringiensis Cry3Aa Toxin. The Journal of Biological Chemistry VOL. 284, NO. 27, pp. 18401–18410, July 3, 2009. ↩

4.Oppert B1, Martynov AG, Elpidina EN.Bacillus thuringiensis Cry3Aa protoxin intoxication of Tenebrio molitor induces widespread changes in the expression of serine peptidase transcripts. Comp Biochem Physiol Part D Genomics Proteomics. 2012 Sep;7(3):233-42. ↩

5.C. Hofmann, H. Vanderbruggen, H. Höfte, J. van Rie, S. Jansens, H. Van MallaertSpecificity of Bacillus thuringiensis δ-endotoxins is correlated with the presence of high affinity binding sites in the brush border membranes of target insect midguts Proc. Natl. Acad. Sci. U. S. A., 85 (1988), pp. 7844–7848 ↩

6. C. Rausell, I. García-Robles, J. Sánchez, C. Muñoz-Garay, A.C. Martínez-Ramírez, M.D. Real, A. Bravo Role of toxin activation on binding and pore formation activity of the Bacillus thuringiensis Cry3 toxins in membranes of Leptinotarsa decemlineata (Say) Biochim. Biophys. Acta, 1660 (2004), pp. 99–105 ↩

7. G. Schulthess, S. Compassi, D. Boffelli, M. Werder, F. E. Weber, and H. Hauser. A comparative study of sterol absorption in different small-intestinal brush border membrane models. Journal of Lipid Research Volume 37, 1996, pp 2405-2419 ↩

8. Raymond F. Chen, Jay R. Knutson. Mechanism of fluorescence concentration quenching of carboxyfluorescein in liposomes: Energy transfer to nonfluorescent dimers. Analytical Biochemistry Volume 172, Issue 1, July 1988, Pages 61-77 ↩

9. Huang., Y. Tippmann, J. Becker, T. Kinetic Modeling of Hop Acids during Wort Boiling. International Journal of Bioscience, Biochemistry and Bioinformatics, Vol. 3, No. 1, January 2013 ↩



References for Toxin engineering

1. Erban, T., Nesvorna, M., Erbanova, M. et al. Exp Appl Acarol (2009) 49: 339. doi:10.1007/s10493-009-9265-z ↩

2. Rajamohan F, Alzate O, Cotrill JA, Curtiss A, Dean DH. Protein engineering of Bacillus thuringiensis δ-endotoxin: Mutations at domain II of CryIAb enhance receptor affinity and toxicity toward gypsy moth larvae. Proceedings of the National Academy of Sciences of the United States of America. 1996;93(25):14338-14343. ↩

3. Kelly D, Ayres M, Lescott T, Robertson J, Happ G. J. A Small Iridescent Virus (Type 29) Isolated from Tenebrio molitor: a Comparison of its Proteins and Antigens with Six Other Iridescent Viruses. Gen. Virol. 42(1):95-105 doi:10.1099/0022-1317-42-1-95 ↩

 

References for Varroa isolates

1. Lacey, L. A., Grzywacz, D., Shapiro-Ilan, D. I., Frutos, R., Brownbridge, M., & Goettel, M. S. (2015). Insect pathogens as biological control agents: back to the future. Journal of invertebrate pathology, 132, 1-41. ↩

2. de Maagd, R. A., Bravo, A., & Crickmore, N. (2001). How Bacillus thuringiensis has evolved specific toxins to colonize the insect world.TRENDS in Genetics, 17(4), 193-199. ↩

3. Marvier, M., McCreedy, C., Regetz, J., & Kareiva, P. (2007). A meta-analysis of effects of Bt cotton and maize on nontarget invertebrates.science, 316(5830), 1475-1477. ↩

4. Ahmed, N., Wang, M., & Shu, S. (2016). Effect of commercial Bacillus thuringiensis toxins on Tyrophagus putrescentiae (Schrank) fed on wolfberry (Lycium barbarum L.). International Journal of Acarology, 42(1), 1-6. ↩

5. Chapman, M. H., & Hoy, M. A. (1991). Relative toxicity of Bacillus thuringiensis var. tenebrionis to the two‐spotted spider mite (Tetranychus urticae Koch) and its predator Metaseiulus occidentalis (Nesbitt)(Acari, Tetranychidae and Phytoseiidae). Journal of Applied Entomology, 111(1‐5), 147-154. ↩

6. Dunstand-Guzmán, E., Peña-Chora, G., Hallal-Calleros, C., Pérez-Martínez, M., Hernández-Velazquez, V. M., Morales-Montor, J., & Flores-Pérez, F. I. (2015). Acaricidal effect and histological damage induced by Bacillus thuringiensis protein extracts on the mite Psoroptes cuniculi. Parasites & vectors, 8(1), 1. ↩

7. Erban, T., Nesvorna, M., Erbanova, M., & Hubert, J. (2009). Bacillus thuringiensis var. tenebrionis control of synanthropic mites (Acari: Acaridida) under laboratory conditions. Experimental and Applied Acarology, 49(4), 339-346. ↩

8. Fernández-Ruvalcaba, M., Peña-Chora, G., Romo-Martínez, A., Hernández-Velázquez, V., de la Parra, A. B., & De La Rosa, D. P. (2010). Evaluation of Bacillus thuringiensis pathogenicity for a strain of the tick, Rhipicephalus microplus, resistant to chemical pesticides. Journal of Insect Science, 10(1), 186. ↩

9. Hassanain, M. A., Garhy, M. E., Abdel-Ghaffar, F. A., El-Sharaby, A., & Megeed, K. N. A. (1997). Biological control studies of soft and hard ticks in Egypt. Parasitology research, 83(3), 209-213. ↩

10. Neethu, K. B., Priji, P., Unni, K. N., Sajith, S., Sreedevi, S., Ramani, N., ... & Benjamin, S. (2015). New Bacillus thuringiensis strain isolated from the gut of Malabari goat is effective against Tetranychus macfarlanei. Journal of Applied Entomology. ↩

11. Zhioua, E., Heyer, K., Browning, M., Ginsberg, H. S., & LeBrun, R. A. (1999). Pathogenicity of Bacillus thuringiensis variety kurstaki to Ixodes scapularis (Acari: Ixodidae). Journal of medical entomology, 36(6), 900-902. ↩

12. Neethu, K. Alquisira-Ramírez, E. V., Paredes-Gonzalez, J. R., Hernández-Velázquez, V. M., Ramírez-Trujillo, J. A., & Peña-Chora, G. (2014). In vitro susceptibility of varroa destructor and Apis mellifera to native strains of Bacillus thuringiensis. Apidologie, 45(6), 707-718. ↩

13. Tsagou, V., Lianou, A., Lazarakis, D., Emmanouel, N., & Aggelis, G. (2004). Newly isolated bacterial strains belonging to Bacillaceae (Bacillus sp.) and Micrococcaceae accelerate death of the honey bee mite, varroa destructor (V. jacobsoni), in laboratory assays. Biotechnology letters, 26(6), 529-532. ↩

14. Rampersad, J., & Ammons, D. (2005). A Bacillus thuringiensis isolation method utilizing a novel stain, low selection and high throughput produced atypical results. BMC microbiology, 5(1), 1. ↩

15. This is the homepage of "De Duurzame Bij". They want to stop treating Varroa with pesticides. Loosely translated, "Duurzame Bij" means "sustainable bee". ↩

16. Kanbar, G., & Engels, W. (2005). Communal use of integumental wounds in honey bee (Apis mellifera) pupae multiply infested by the ectoparasitic mite Varroa destructor. Genetics and Molecular Research, 4(3), 465-472. ↩

17. Chow, S. C. (2011). Sample size calculations for clinical trials. Wiley interdisciplinary reviews: Computational statistics, 3(5), 414-427. ↩

18. Bavykin, S. G., Lysov, Y. P., Zakhariev, V., Kelly, J. J., Jackman, J., Stahl, D. A., & Cherni, A. (2004). Use of 16S rRNA, 23S rRNA, and gyrB gene sequence analysis to determine phylogenetic relationships of Bacillus cereus group microorganisms. Journal of clinical microbiology, 42(8), 3711-3730. ↩

19. Berry, C. (2012). The bacterium, Lysinibacillus sphaericus, as an insect pathogen. Journal of invertebrate pathology, 109(1), 1-10. ↩

20. Cox, J., & Mann, M. (2008). MaxQuant enables high peptide identification rates, individualized ppb-range mass accuracies and proteome-wide protein quantification. Nature biotechnology, 26(12), 1367-1372. ↩