Coffee berry borers were dissected from infested coffee fruits collected in the field in Hawai’i and Mexico and from a colony continuously reared for ca. 15 years on an artificial diet at the United States Department of Agriculture – Agricultural Research Service laboratory in Beltsville, Maryland. Hawaiian specimens were dissected from infested C. arabica berries collected at: (1) Holualoa, Hawai’i (19°38′26″ N; 155°57′19″ W; 44 meters above sea level – masl); (2) two locations in Wood Valley, K’au, Hawai’i (19°16′15″ N; 155°29′19″ W; 795 masl; and 19°15′52″ N; 155°28′00″ W; 638 masl); and (3) Dole Plantation, Waialua, Oahu (21°33′51″ N; 158°04′26″ W; 146 masl). Coffee berry borers from Mexico were dissected from infested C. canephora fruits collected at Finca San Antonio, Cacahoatán, Chiapas (15°00′20″ N; 92°09′12″ W; 562 masl).

Adult female coffee berry borers (ca. 2 mm long) were sterilized by sonicating in 90% ethanol for 30 s in sterile 1.5 ml Eppendorf tubes placed in a floating foam tube rack placed inside a Branson 2210 ultrasonic cleaner (Branson Ultrasonics Corp., Danbury, CT, United States), followed by soaking in 3.5% sodium hypochlorite for 5 min (no sonication). Insects were then washed twice in nuclease-free water and blotted on a sterile paper towel in the laminar flow hood. Twenty heads from each sampling site (Table 1) were dissected from surface-sterilized adult females and macerated with sterile disposable plastic mini-pestles in 2 ml nuclease-free culture tubes containing 1 ml of nutrient broth (NB; Sigma-Aldrich, St. Louis, MO, United States). The reason for using heads is based on having observed an egg grooming behavior in which adult females smear recently oviposited eggs using her mouthparts (Vega et al., 2017), suggesting they may secrete a substance over the egg. Twenty eggs from each sampling site were cultured with or without surface-sterilization (Table 1). Eggs were sterilized for 1 min by placing in a mixture of 0.5% sodium hypochlorite and 70% ethanol. The sterilizing solution was drawn off with a pipette and replaced with nuclease-free water. Sterilized eggs were left whole or were macerated with sterile disposable plastic mini-pestles in 2 ml nuclease-free culture tubes containing 1 ml of nutrient broth (NB; Sigma-Aldrich, St. Louis, MO, United States). Whole non-sterilized eggs and female larvae (n = 10) from Mexico (Table 1) were directly transferred to NB (see below). Frass (Table 1) was collected with a sterilized metal spatula directly from the surface of artificial media from the Beltsville colony. The effectiveness of the sterilization protocol was assessed by placing a subsample of the nuclease-free water (used to wash the samples after removing the sterilizing solution) on the surface of nutrient agar (NA; Sigma-Aldrich, St. Louis, MO, United States) and observing whether any growth ensued, in which case the samples were discarded.

Samples (macerated sterile heads, whole sterile or non-sterile eggs, macerated sterile eggs, whole non-sterile larvae, and frass) were transferred to 4.5 ml sterile, nuclease-free culture tubes containing 2 ml NB and incubated at 28°C and 225 rpm on a rotary shaker for 48 h. The metagenomic (mixed taxa) NB cultures were archived at −80°C by combining 500 μl of each culture with 125 μl of 80% glycerol in nuclease-free deep-well plates. Ten microliters of each original culture were diluted in 50 μl of molecular-grade water (nuclease-free water) for polymerase chain reaction (PCR) screening analysis. The remainder of the 48 h NB cultures were used for dilution plating on NA (100 μl of each dilution on NA; 24 h at 28°C), to obtain single colony isolates. A total of 938 isolates were obtained: 378 from Mexico; 284 from the Beltsville colony; and 276 from Hawai’i. Single colony isolates were assigned a general morphotype based on macroscopic colony features, picked with a sterile toothpick, transferred to nuclease-free 96-well, deep-well plates containing 500 μl NB and incubated at 28°C and 225 rpm on a rotary shaker for 24 h. NB cultures of the single colony isolates were archived at −80°C (as described previously) in the Sustainable Agricultural Systems Laboratory Collection (USDA, ARS, Beltsville, MD, United States). Ten microliters of each original culture were diluted in 50 μl of nuclease-free water for PCR screening analysis.

Single colony NB dilutions were used as template for PCR amplification of the16S rRNA gene and of the methylxanthine N1-demethylase gene, ndmA. Primers used for 16S rRNA amplification were 27F (5′- AGA GTT TGA TCM TGG CTC AG-3′) and 1392R (5′- ACG GGC GGT GTG TAC A-3′), and for ndmA were CBBcdmF (5′- TGG CAT CCC GTW TGT ACY GT-3′) and CBBcdmR (5′- CTT GKA TAA CRA TTC GCA ACC-3′) (Ceja-Navarro et al., 2015). Twenty-five microliter PCR reactions contained 3 μl of diluted NB isolate culture, 0.4 μM of each primer and 1X AmpliTaq Gold PCR Master Mix (6.25 U AmpliTaq Gold DNA polymerase, 2.5 mM MgCl₂, and 200 μM of each dNTP; Applied Biosystems/Life Technologies, Carlsbad, CA, United States). PCR amplification was conducted in an Eppendorf Mastercycler Gradient thermal cycler (Eppendorf, Westbury, NY, United States). The amplification program consisted of an initial 5 min denaturation step at 95° C, followed by 35 cycles at 95°C for 30 s, 54°C for 30 s, and 72°C for 1 min; and a final extension step at 72°C for 10 min. Following amplification, PCR products were electrophoresed on 1% agarose gels containing 1X TAE and 1X GelStar Nucleic Acid Gel Stain (Lonza Rockland Inc., Rockland, ME, United States), and documented using UV transillumination to detect 16S rRNA and ndmA bands.

Polymerase chain reaction products from the 16S rRNA and ndmA amplifications were diluted 1:3 with nuclease-free water and direct-sequenced using the BigDye Terminator v3.1 Cycle Sequencing Kit according to the manufacturer’s protocol (Applied Biosystems/Life Technologies, Carlsbad, CA, United States). Separate sequencing reactions were run for the respective forward and reverse primers in each set (27F and 1392R, CBBcdmF, and CBBcdmR). Sequencing reactions were purified using the Performa DTR Ultra 96-Well Plate Kit (EdgeBio, Gaithersburg, MD, United States), resuspended in HiDi Formamide (Applied Biosystems/Life Technologies, Carlsbad, CA, United States) and analyzed on an ABI Prism 3130 Genetic Analyzer (Applied Biosystems Inc., Foster City, CA, United States). Sequences were quality checked and aligned using the DNAStar suite of software (DNAStar, Madison, WI, United States), and preliminary identification made using the Basic Local Alignment Search Tool (BLAST) and GenBank nucleotide data bank from the National Center for Biotechnology Information, Bethesda, MD, United States¹.

Fifty representative single colony isolates were selected for Illumina genomic library preparation using a ranking based on at least one of the following: (1) preliminary 16S rRNA BLAST identification; (2) surface-sterilization or no surface-sterilization; (3) type of insect tissue cultured; (4) sampling site; (5) colony morphotype; (6) whole or macerated tissue; (7) presence of ndmA amplicon; or (8) BLAST identification of ndmA sequence. It is important to note that ndmA was not present in all the isolates when this selection was made, as our interest was focused at casting a wide net representative of as many characteristics as possible. Single colony isolates from glycerol stocks were grown overnight in NB at 28°C and 225 rpm. Cells were pelleted at 10,000 relative centrifugal force for 1 min, and genomic DNA extracted using the Qiagen DNeasy DNA extraction protocol for bacterial cultures (Qiagen, Germantown, MD, United States). Genomic DNA quality was assessed by nanodrop 260/280 ratio, and quantity was measured using a Qubit fluorometer (Thermo Fisher Scientific Inc., Waltham, MA, United States). Libraries were prepared from 1 ng of genomic DNA from 50 coffee berry borer-associated single colony isolates using the Nextera XT DNA Library Prep Kit (Illumina, Inc., San Diego, CA, United States). The libraries were sequenced on a NextSeq by bi-directional multiplex-indexed Illumina sequencing (Nextera XT Index Kit, Illumina, Inc., San Diego, CA, United States) using a NextSeq 500 High Output Kit (150 cycles; Illumina, Inc., San Diego, CA, United States).

The quality of the reads was checked by the FASTQC program². The 50 isolates (4 paired-end libraries for each isolate) were assembled using the SPAdes assembler v3.14.0 (-t 60, -m 500, –careful, -k 17,21,31,41,51) (Bankevich et al., 2012). The read coverage (read mapping back to the contigs) was calculated using Bowtie2 (Langmead and Salzberg, 2012) and the BBtools suite³. Gene prediction on the contigs was performed using the software program Prodigal v2.6.3 (Hyatt et al., 2010). The gene-encoded proteins were annotated by blasting against the refseq of bacterial proteins at the NCBI using the BLASTP program. Gene-encoded proteins were also blasted against known caffeine break-down genes: KAF4561170_ndmA, KAF4561167 ndmB, KAF4561164_ndmC, KAF4561163_ndmD, KAF4561162_ndmE, KAF4561168_cafR, KAF4561173_cafT, KAF4561166_cafP, KAF4561172 frmA, KAF4561161_frmB, KAF4561171_VOC where only high E-values close to E 0.0 were considered to be significant matches. 16S rRNAs were predicted using RNAmmer 1.2 and tRNAs were predicted by tRNAscan-1.4. Bacterial identification was based on the top BLAST hits with the 16S rRNA followed by an analysis of the average nucleotide identity (ANI; Chun et al., 2018; Table 1) between the draft genome generated in this study with the genome of the associated top BLAST. Values > 95% were used as the species boundary following proposed minimal standards by Chun et al. (2018). The Orthologous Average Nucleotide Identity Tool (OrthoANI; Lee et al., 2016; Chun et al., 2018) software was downloaded from⁴.

The 50 bacterial strains sequenced in this study were inoculated from frozen glycerol stocks into a modified M9 minimal medium containing, per liter, 12.8 g Na₂HPO₄⋅7H₂O, 3 g K₂HPO₄, 2.76 g NH₄Cl, 0.5 g NaCl, 493 mg MgSO₄, 55.5 mg CaCl₂, and 2.5 g caffeine (J. T. Baker, Avantor Performance Materials, Inc., Center Valley, PA, United States) or in nutrient broth (6 g/L peptone and 3 g/L yeast extract; VWR International LLC, Radnor, PA, United States) containing 2.5 g/L caffeine. Aliquots were removed periodically from the cultures and mixed 1:1 (vol:vol) with methanol to stop cell growth and metabolism. Concentrations of caffeine and metabolites were then determined by HPLC as described previously (Summers et al., 2012) using a Shimadzu Prominence series HPLC equipped with a photodiode array. Metabolites were identified based on their retention times and UV spectra when compared with authentic standards.

Assembly statistics for 50 bacterial strains in 14 genera (Acinetobacter, Bacillus, Delftia, Enterococcus, Erwinia, Klebsiella, Kosakonia, Lactococcus, Leuconostoc, Ochrobactrum, Paenibacillus, Pantoea, Pseudomonas, and Stenotrophomonas) isolated from coffee berry borer eggs, frass, head, and larvae are presented in Supplementary Table 2. Based on bioinformatics analysis, 21 bacterial species (Acinetobacter sp. S40, S54, S55, Bacillus aryabhattai, Delftia lacustris, Erwinia sp. S38, S43, S63, Klebsiella oxytoca, Ochrobactrum sp. S45, S46, Pantoea sp. S61, Pseudomonas aeruginosa, P. parafulva, and Pseudomonas sp. S30, S31, S32, S37, S44, S60, S75) contain genes homologous to those involved in caffeine N-demethylation (Table 1); 35 bacterial species (Acinetobacter sp. S40, S54, S55, B. aryabhattai, B. cereus group; Bacillus sp. S29, S70, S71, S72, S73, D. lacustris, Erwinia sp. S38, S43, S59, S63, K. oxytoca, Kosakonia cowanii, Ochrobactrum sp. S45, S46, Paenibacillus sp. S28, Pantoea sp. S61, S62, P. aeruginosa, P. parafulva, Pseudomonas sp. S30, S31, S32, S37, S44, S60, S75, Stenotrophomonas sp. S39, S41, S48, S49) contain genes that may be involved in C-8 oxidation (Supplementary Table 3); and nine bacterial species (Bacillus cereus group, Enterococcus sp. S52, S53, S76, S77, Lactococcus sp. S47, S64, and Leuconostoc sp. S50, S51) did not have any genes known for caffeine breakdown function (Supplementary Table 3).

The full complement of caffeine N-demethylation genes (ndmABCDE) was detected in (1) P. parafulva isolated from coffee berry borer frass from Beltsville; (2) Pseudomonas sp. S60 isolated from macerated sterilized heads from Hawai’i; (3) Pseudomonas sp. S31 isolated from whole non-sterilized coffee berry borer eggs from Mexico; and (4) Pseudomonas sp. S32, S37 isolated from whole sterilized eggs from Mexico (Table 1). In each case, the ndmABCDE genes were found in a gene cluster (Figure 2), with gene arrangement identical to the alkylxanthine (Alx) gene cluster encoding N-demethylase enzymes and which has been reported in Pseudomonas sp. CES (Summers et al., 2020) and Pseudomonas sp. NCIM 5235 (Retnadhas and Gummadi, 2018). A summary for the specific insect tissue from which the 21 bacterial species involved in caffeine N-demethylation were isolated is presented in Table 2. Nineteen bacterial species were isolated from only one source, i.e., heads, eggs, larvae, or frass (Table 2). Two bacterial species were isolated from two different sources: B. aryabhattai (eggs and frass), and P. aeruginosa (heads and eggs) (Table 2).

Some of the ndm genes were also detected in Acinetobacter sp. S40, S54, S55 (ndmCD), B. aryabhattai (ndmA), D. lacustris (ndmADE), Erwinia sp. S38, S43 (ndmAD), Erwinia sp. S63 (ndmD), K. oxytoca (ndmAD), Ochrobactrum sp. S45, S46 (ndmE), Pantoea sp. S61 (ndmE), P. aeruginosa (ndmDE), Pseudomonas sp. S30, S75 (ndmE), and Pseudomonas sp. S44 (ndmDE) (Table 1). A Venn diagram showing how the genes involved in the caffeine N-demethylation pathway are shared among the various bacterial species is shown in Figure 3.

Aside from ndmAD in Erwinia sp. S38, S43, which were found adjacent to each other (Figure 2), no other ndm gene homologs were in the same gene cluster. Rather, these genes were scattered throughout the genome (Figure 2). However, most of the homologs were surrounded by similar genes such as ABC family and other transporters, permeases, and GntR, LysR, and MarR transcriptional regulators (Figure 2; Supplementary Table 4).

While Pseudomonads are common caffeine degraders, two Acinetobacter sp. strains, a Klebsiella/Rhodococcus mixture, three Pantoea species, an Ochrobactrum sp., and S. maltophilia have been previously reported to degrade caffeine (Supplementary Table 1). Thus, the diversity observed here is not surprising but adds additional genera, such as Bacillus, Delftia, and Erwinia to the list of potential caffeine-degrading bacteria.

Of the 50 strains tested grown in M9 minimal media containing 0.25% caffeine as the sole carbon source to detect metabolites, only eight grew and produced sufficient metabolites to facilitate their detection by HPLC (Table 3). In each case, we detected theobromine and 7-methylxanthine in the growth medium, indicating that caffeine is metabolized by N-demethylation. The remaining 42 strains did not grow when caffeine was provided as the sole carbon and nitrogen source.

The presence of the full complement of ndm genes in several Pseudomonads was expected, given that N-demethylation is the most commonly observed caffeine degradation mechanism in bacteria (Supplementary Table 1) and that Pseudomonas is the most common caffeine-degrading genus worldwide (Supplementary Table 1). In contrast to these strains, only a subset of the genes was found in several bacterial strains. It is possible that the bacterial cells use other undiscovered genes and enzymes for caffeine metabolism or that the similarity of the homologous genes is lower than our cutoff threshold. To minimize false positives, E-values close to E 0.0 with high sequence similarity to known ndm genes were considered, since ndm genes have sequence similarity to proteins that might not be involved in caffeine breakdown. Similarly, the portion of the genome containing the relevant ndm and caffeine breakdown genes in these strains may not have been sequenced due to what Baptista and Kissinger (2019) refer to as “troublesome characteristics”, including extreme nucleotide bias, long homopolymeric runs, repetitive sequences, mobile elements, etc.

Another possibility for the low number of strains with a full complement of ndm genes (Figure 3) is that some strains may be able to carry out one step of the caffeine degradation pathway and rely on other strains to carry out additional steps. For example, the D. lacustris, Erwinia sp. S38, S43, and K. oxytoca strains contain ndmAD homologs, which may enable them to convert caffeine to the less-toxic theobromine. Acinetobacter sp. S40, S54, S55 contains ndmCD genes, which could potentially convert 7-methylxanthine to xanthine, which is formed during the biosynthesis of caffeine in coffee plants (Ashihara et al., 1996). Therefore, drivers for bacteria to carry out individual steps in the process may be to reduce toxicity of caffeine or to use metabolites already present, such as theobromine or 7-methylxanthine. Curiously, none of the strains with only a subset of genes contain ndmB homologs. However, given that as few as two mutations to the ndmA active site are enough to change the activity from N₁- to N₃-demethylation (Kim et al., 2019) some of the genes identified as ndmA may have a broader positional specificity than those which have already been characterized. Alternatively, there may be unculturable bacteria with ndmB homologs that we were not able to detect using the approach described here. NdmE is a glutathione-S-transferase homolog and is believed to play a structural role in the NdmCDE complex; there is no known catalytic role for NdmE (Summers et al., 2013). Thus, strains that only contain ndmE, such as Ochrobactrum sp. S45, S46 and Pantoea sp. S61, may not have any N-demethylation activity but may have a ndmE homolog that carries out a different function in the cell.

We note that the presence of ndm gene homologs does not indicate that the strains can completely degrade caffeine, or even that those genes are involved in caffeine metabolism. However, there are no other reports of N-demethylase enzymes capable of degrading caffeine and the NdmABCDE enzymes have only exhibited activity toward caffeine and related methylxanthines. These enzymes are part of the Rieske non-heme iron oxygenase family, which contains enzymes responsible for degradation of aromatic compounds toluene, naphthalene, biphenyl, and the herbicide dicamba (Gibson and Parales, 2000). Thus, there may be other aromatic compounds toward which these gene products are active. Regardless, the presence of ndm gene homologs does give insight into possible mechanisms and provides the basis for further studies to elucidate caffeine degradation by bacteria.

In each of the eight strains grown in M9 minimal media with 0.25% caffeine as the sole carbon and nitrogen source (Table 3) and which produced sufficient metabolites for HPLC detection, we detected the production of theobromine and 7-methylxanthine, indicating an active N-demethylation pathway. The P. parafulva and Pseudomonas sp. S32, S37, S60 containing the full set of ndm genes were found among this group. The other four strains that grew were Bacillus sp. S72, Pantoea sp. S62, and Stenotrophomonas sp. S39 among which no ndm genes were detected, and P. aeruginosa, in which only ndmD and ndmE were present. The ability of these strains to carry out N-demethylation of caffeine indicates that they may possess different genes encoding enzymes of similar function, the ndm genes were simply incompletely sequenced or not sequenced (Baptista and Kissinger, 2019), or that the ndm gene E-values did not meet our cutoff threshold. To date, no other genes/enzymes have been shown to carry out N-demethylation of caffeine or other methylxanthines in bacteria.

The 42 strains that did not grow when caffeine was provided as the sole carbon and nitrogen source may not have all of the genes necessary to completely degrade caffeine. Another explanation is that the genes were not induced under the growth conditions. It may be possible for strains with ndmAD gene combinations to convert caffeine to theobromine, but the single methyl group removed would likely not enable cells to grow. Additionally, very little is known about the regulation of the caffeine-degrading genes. Studies have indicated that these genes are regulated very tightly and only expressed under certain conditions. For example, P. putida CBB5 can express the ndm genes when grown with 0.25% caffeine and 0.4% soytone, but expression is greatly reduced when soytone is replaced with yeast nitrogen base (YNB) (Yu et al., 2009). In contrast, the caffeine dehydrogenase expression in Pseudomonas sp. CBB1 is not inhibited by YNB but is almost completely repressed by soytone (Yu et al., 2008). Thus, the absence of metabolites does not indicate a lack of genes, but observation of caffeine metabolites does provide evidence for specific enzymes. Indeed, we did not observe any caffeine degradation from any of the 50 strains when grown in nutrient broth supplemented with 0.25% caffeine, indicating that there are specific circumstances in which the ndm genes are expressed.

We also analyzed caffeine dehydrogenase genes (cdhA, cdhB, cdhC) involved in C-8 oxidation, the second pathway of caffeine break down (Figure 1; Supplementary Table 3). The C-8 oxidation pathway involves the oxidation of caffeine to 1,3,7-trimethyluric acid (TMU; Figure 1; Supplementary Table 3), which is further degraded in a pathway homologous to uric acid metabolism. The caffeine dehydrogenase enzyme, encoded by genes cdhABC, carries out the first step in this pathway (Yu et al., 2008). Further steps are catalyzed by trimethyluric acid-degrading genes (tmuM, tmuH, tmuD) (Mohanty et al., 2012). The peptide sequence of the cdhA gene, which forms one subunit of the heterotrimeric caffeine dehydrogenase is similar to xanthine dehydrogenases, aldehyde dehydrogenases, and carbon monoxide dehydrogenases (Yu et al., 2008). Similarly, the tmuM, tmuH, and tmuD genes have homologs involved in uric acid metabolism. It is possible that the genes listed in Supplementary Table 3 are instead involved in xanthine and uric acid metabolism. However, xanthine dehydrogenase is not known to metabolize caffeine, nor have the tmu genes shown activity toward trimethyluric acid. To date, we do not have experimental evidence of TMU formation from caffeine by any of these strains, indicating that either the homologs we detected do not metabolize caffeine or that the conditions for their expression have not yet been identified. Cloning and expression of the genes and characterization of the enzymes they encode would be necessary to determine if these genes are specific for caffeine.

For strains such as Ochrobactrum sp. S45, S46, Pantoea sp. S61, and Pseudomonas sp. S30, S75 that have ndmE and cdh homologs, it is more likely that they metabolize caffeine through the C-8 oxidative route rather than the N-demethylation route because NdmE has no known catalytic activity. Two B. aryabhattai strains contain the full set of C-8 oxidative genes, but only the ndmA gene from the N-demethylation pathway. Because this strain does not contain an ndmD homolog necessary for N-demethylation, it would most likely use the oxidative route, as well.

Possible involvement of the cdh and tmu genes (Supplementary Table 3) in the breakdown of caffeine and subsequent products needs to be experimentally determined to ascertain whether this role could be occurring in the coffee berry borer. It is interesting that ndm genes were not identified in K. cowanii, although it has CdhABC and TmuMD; Ceja-Navarro et al. (2015) identified K. cowanii as a caffeine-degrading bacterium in the coffee berry borer.

It is also important to note that there might be other unculturable bacteria that are capable of degrading caffeine, although that was not the focus of this paper.

The genus Pseudomonas comprises ca. 150 species (Pereira et al., 2018) and in the present study, Pseudomonas aeruginosa, P. parafulva, and Pseudomonas sp. S30, S31, S32, S37, S44, S60, S75 were found to have caffeine N-demethylation genes, with five of them having the full complement of caffeine demethylating genes (ndmABCDE; Tables 1, 2 and Figure 3): P. parafulva (frass), Pseudomonas sp. S60 (heads), and Pseudomonas sp. S31, S32, S37 (eggs). Pseudomonas aeruginosa had ndmDE (heads, eggs) and Pseudomonas sp. S30, S75 only had ndmE (heads) (Tables 1, 2 and Figure 3).

Mariño et al. (2018) reported 392 bacterial genera in the microbiome of eggs and whole adult female coffee berry borers, with a preponderance of Pseudomonas and Pantoea. Pseudomonas spp. represented 48.5% of the microbiota, with P. chloroaphis being common in laboratory-reared insects and P. putida being more common in field-collected insects. Two other Pseudomonas species identified in the microbiome were P. viridiflava and P. fulva.

This is the first report of potential caffeine N-demethylation genes in D. lacustris (ndmADE). Intriguingly, as for Pseudomonas sp. S30, S60, S75, the detection was made from sterilized, macerated heads. Delftia has been isolated from the coffee berry borer by Mariño et al. (2018).

Transmission of a symbiote from the mother to its offspring is known as vertical transmission (Jaenike, 2012; Ebert, 2013; Kucuk, 2020) and could be mediated internally or externally by the ovum and referred to as transovum transmission (Solter, 2006). One type of transovum transmission is transovarial transmission involving the internal invasion of the egg by the symbiote, as has been reported for many insects (Zchori-Fein et al., 2006; Skaljac et al., 2010). We detected eight bacteria with N-demethylation genes in sterile eggs: Acinetobacter sp. S40 (ndmCD), B. aryabhattai (ndmA), Erwinia sp. S38 (ndmAD), Ochrobactrum sp. S45, S46 (ndmE), P. aeruginosa (ndmDE), and Pseudomonas sp. S32, S37 (ndmABCDE) (Table 1). It is unlikely that Acinetobacter sp. S40, B. aryabhattai, Erwinia sp. S38, and Ochrobactrum sp. S45, S46 are transovarially transmitted and detection might have been a result of contamination or imperfect sterilization. Transovarial transmission would be rather surprising for these multiple free-living bacteria which readily grow on standard cultivation media and are otherwise not consistently associated with the host. Using fluorescent in situ hybridization, Sayaka Aoki (University of Hawai’i; personal communication) has detected Pseudomonas spp. with at least one caffeine degrading gene inside coffee berry borer eggs.

The other type of ovum mediated transmission involves inoculation of the egg shell via frass or secretions containing the symbiote, which is ingested upon egg hatching (Hosokawa et al., 2012; Salem et al., 2015b). As stated above, we have observed an egg grooming behavior in which female coffee berry borers appear to smear her eggs shortly after oviposition (Vega et al., 2017). A similar situation has been reported in Adomerus stinkbugs, which smear a secretion containing a symbiote over the eggs (Hosokawa et al., 2013). The detection of ndmABCDE in Pseudomonas sp. S60 isolated from macerated sterilized heads (Table 1) could be a mechanism for vertical transmission if the hatching embryos acquire the bacterial symbiote by eating the egg shell or becoming otherwise contaminated with the symbiote. This possible type of vertical transmission needs further study.

Coffee berry borer larvae could acquire Acinetobacter sp. S54, S55 (ndmCD) and Pseudomonas sp. S31 (ndmABCDE), which were isolated from non-sterile eggs (Table 1), from the egg shell as the larvae hatches. These bacteria could also contaminate the entire environment within the coffee berry. The coffee berry borer rectum ends with the anus, which is an opening independent of the genital orifice (gonopore) (Alba-Alejandre et al., 2019). Both orifices are very close to each other, with the anal opening being above the genital orifice. Due to the large size of coffee berry borer eggs, it is very likely that as they are being oviposited they will place pressure on the rectum, which could expel fecal matter that would then contaminate the egg shell.

In addition, the presence of ndmABCDE in P. parafulva and ndmA in B. aryabhattai isolated from coffee berry borer frass from the Beltsville colony (Table 1) suggests that potential caffeine demethylating bacteria could contaminate the egg shell or be acquired from the environment (i.e., horizontal transmission) via bacterial dissemination from frass through insect contact, probing, and subsequent spread in galleries. The acquisition of bacteria through coprophagy (Salem et al., 2015a; Onchuru et al., 2018) cannot be ruled out for the coffee berry borer. Erwinia sp. S43 (ndmAD) was detected in non-sterile larvae from Mexico (Table 1), suggesting they could be acquiring the bacterium inside the gallery. Horizontal transmission of caffeine-degrading bacteria also needs further study.

Caffeine-degrading functions might also be expressed by bacteria isolated from soil in coffee plantations, as has been shown for Pseudomonas sp. (Dash and Gummadi, 2006; Gokulakrishnan et al., 2007), P. alcaligenes (Sarath Babu et al., 2005), P. putida (Yamaoka-Yano and Mazzafera, 1998, 1999), and Serratia marcescens (Mazzafera et al., 1996; Supplementary Table 1). In addition, endophytic bacteria in coffee plants might also be capable of degrading caffeine. Vega et al. (2005) identified 19 bacterial genera as endophytes in coffee plants from Colombia, Hawai’i, and Mexico, with Bacillus, Burkholderia, Clavibacter, Curtobacterium, Escherichia, Micrococcus, Pantoea, Pseudomonas, Serratia, and Stenotrophomonas isolated from the seed. Pseudomonas sp., P. chloroaphis, and P. putida were identified as endophytes in various coffee plant tissues sampled (Vega et al., 2005) and P. monteilii has been isolated from coffee pulp waste (Arimurti et al., 2018). A coffee plant endophytic Pseudomonas sp. isolated in India was capable of breaking down 98.6% of the caffeine added to a nutrient broth, and six additional unidentified endophytic bacteria were also capable of catabolizing caffeine with a lower efficiency (Baker et al., 2012). Two P. putida strains endophytic in C. arabica and Coffea canephora leaves, stems, and roots in Brazil were capable of degrading caffeine (Nunes and de Melo, 2006).

For the coffee berry borer, the presence of bacteria on the surface of the fruit, or endophytically in the pulp or seed (Vega et al., 2005), might result in the acquisition of caffeine-degrading bacteria as the colonizing female infests fruit and seeds. A similar situation has been reported for 13 bark beetle species in the genus Dendroctonus, which acquire bacterial members of their microbiome from the environment, including their host trees (Hernández-García et al., 2017).