A Beetle in a Haystack: Are There Alternate Hosts of the Coffee Berry Borer (Hypothenemus hampei) in Puerto Rico?

Can the coffee berry borer (Hypothenemus hampei, or CBB) use host plants other than coffee for food and shelter? The use of fruits other than coffee has been reported. However, the validity of these reports depends on accurate identification of CBB, which is sometimes uncertain. In this study we sampled potential alternate hosts in coffee farms in Puerto Rico. Fruits with perforations were collected and examined for the presence of scolytid beetles (Coleoptera: Curculionidae: Scolytinae). Scolytids were identified by morphology and DNA barcoding of the COI gene. Association between the presence of Inga vera and Guarea guidonia trees and infestation rate of CBB in coffee fruits was evaluated. Food preference tests were performed in the laboratory. A total of 3563 beetles were found and 587 were identified as Hypothenemus spp.; of these, 85 identifications were confirmed by DNA barcoding. Twenty-seven of the beetles identified were H. hampei, mostly in I. vera fruits in periods between coffee crops. Most scolytids identified were H. obscurus. In preference tests, some CBB initially penetrated G. guidonia fruits, but eventually chose coffee. There was no evidence of feeding or reproduction in fruits of G. guidonia or Cajanus cajan. The results show that in Puerto Rico it is rare to find CBB in fruits of alternate hosts. The scarcity of coffee fruits in the off-season might cause some CBBs to take refuge in other fruits, but they did not feed or reproduce in them in laboratory tests. Understanding the refugia of CBB in the off-season may be useful for designing effective management strategies.


Introduction
How does the coffee berry borer Hypothenemus hampei (Ferrari) (Coleoptera: Curculionidae) survive between crops when there are few or no coffee fruits? This question is key for management of the coffee berry borer (or 'CBB') because this period is a population bottleneck; eliminating refugia would greatly reduce the population available to attack the next crop. This question is still unresolved and underlines the importance of identifying the potential alternative hosts for this species. The CBB is the most serious pest of coffee worldwide, so understanding its refugia is of great practical importance to coffee growers and the coffee industry.
(Hemileia vastatrix). None of the growers applied chemicals to control the CBB; the grower in Utuado did not use anything, and the other three growers occasionally used Mycotrol®(active ingredient: The fungus Beauveria bassiana) and artisanal traps to reduce CBB populations [4,10] (Table 1).  1 Used to control weeds, 2 Used to monitoring and control the coffee berry borer (CBB), 3 Used to control coffee leaf rust (Hemileia vastatrix).
Around and within these coffee plots, all plant species with fruits >1 cm in diameter were sampled, except grasses and herbs. Fruits were collected with a pole trimmer up to a height of 5 m. The irregular topography of the farms and distribution of trees precluded a precisely defined distance for sampling. Botanical samples were pressed and taken to the Herbarium of the University of Puerto Rico, Río Piedras (UPRRP) for identification. Up to 50 fruits per species with at least one perforation similar in size to those made by the CBB [19] were collected monthly to evaluate the presence of H. hampei in the fruits. The plant species sampled in each farm are listed in Table 2.

Identification and Phylogenetic Analysis
Fruits were examined under a dissecting microscope for adult beetles with morphological characteristics similar to H. hampei (i.e., of the family Curculionidae, subfamily Scolytinae). The beetles were sorted by host plant, farm and date, and were preserved dry at −20 • C. A subsample was selected for morphological identification. Identification was based on size and morphology of bristles, surface texture of the elytra and frons, and size and color of the insect [20]. Only Hypothenemus is considered here. The sample size for morphological identification was determined using Cochran's sample size formula for categorical data with an alpha level of 0.01 and an acceptable margin of error of 5% [21,22]. A subsample of the scolytids identified by morphology was selected for DNA barcoding and phylogenetic analysis. DNA barcoding was used to confirm morphological identification and determine how similar the CBB population in Puerto Rico was to previously studied populations in other areas. DNA barcoding used part of cytochrome oxidase c subunit 1 (COI) mitochondrial gene, the standard for insects (http://www.barcodinglife.org/index.php/IDS_OpenIdEngine). Tissue was crushed between two glass slides, and DNA was extracted with the MoBio Ultraclean Tissue & Cells DNA Extraction kit. The primers COIF and COIR [23] were used to amplify the COI region with an annealing temperature of 55 • C. The PCR products were cleaned with ExoSAP-IT (USB Corp.) and were sequenced in both directions at the Sequencing and Genomics Facility of the University of Puerto Rico (UPR-SGF). Sequences were assembled and examined for errors using Sequencher 4.8 (Gene Codes Corp.) and compared to the GenBank database. Only COI sequences with >98% similarity were accepted as confirmed identifications [24].
For phylogenetic analysis, GenBank reference sequences of H. hampei, H. obscurus (Fabricius), H. eruditus Westwood, H. seriatus (Eichhoff), and unidentified Hypothenemus spp. were included [23,25,26]. Although a phylogeny based on only one sequence has limited resolving power, no other sequences are available for all the reference taxa of interest. The sequences were aligned using Muscle in Mega [27]. Maximum-likelihood trees were constructed using RAxML (Randomized Axelerated Maximum Likelihood) version 8.0 [28] with GTRCAT model of evolution, 25 gamma categories, and the automatic Bootstrap MRE [29,30]. The analysis was conducted on a CIPRES Science Gateway Web server (on RAxML-HPC Black box tool version 8.2.10) [31]. Xylosandrus compactus (Eichhoff) (Curculionidae, Scolytinae) was used as an outgroup [25] and sequences of H. hampei from other areas were included for comparison.

Preference and Feeding Tests
Since G. guidonia was one of the most abundant tree species and produced the most fruits in all farms sampled, as well as produced fruits all year long (including when coffee fruits were not present), it was selected for preference tests in the laboratory. One lab-reared, adult CBB female was placed inside a jar with a mature fruit of G. guidonia and a mature fruit of C. arabica. Screw-top glass jars 70 mm in diameter and 90 mm high were used, large enough that the two fruits were not touching. CBBs were placed between the two fruits. The test was run twice with 50 replicates each. Behavior was observed daily for 10 days, registering which fruit was penetrated first and if the CBB remained within that fruit. After a month, the two fruits in each jar were dissected to search for evidence of reproduction and feeding.
Alternative food sources were also tested to determine whether H. hampei consumed fruits other than coffee. Fruits previously reported as food sources in laboratory tests with H. hampei were used: Leucaena leucocephala (Lam.) De Wit [6] and Cajanus cajan [32]. Individual, lab-reared adult CBB females were placed in transparent plastic vials (40 × 100 mm, plugged with bonded dense-weave cellulose acetate) with the fruit of one of the two species. Fruits were used within 24 h of harvest. Fifty individuals of H. hampei were used for each species. Each CBB was observed until it died or penetrated a fruit. The day after the fruit was drilled, it was opened to see if seed tissues had been consumed. To obtain a uniform cohort for these tests, we used young female CBBs raised in an incubator at 25 • C on the artificial diet Cenibroca [33]. The tests were performed in an incubator at 25 • C and 80-96% relative humidity.

Association between Common Trees and CBB
To determine if some shade trees are a refuge for CBB or otherwise associated with it, we evaluated whether the presence of I. vera (Fabaceae-Mimosoideae) and G. guidonia (Meliaceae) trees in coffee farms was associated with CBB infestation at each site. Inga vera and G. guidonia were chosen for this analysis because: (i) They were present in most of the farms visited, (ii) they are the most abundant and important species of the coffee agroecosystem in Puerto Rico according to Arango [34], and (iii) Agronomy 2020, 10, 228 5 of 12 legumes (e.g., I. vera) were reported as alternate hosts of CBB in Central Africa by Schedl [16,18]. This analysis included a larger dataset than that used above, from 110 sites in 61 farms planted with Coffea arabica, sampled from August to November 2014 [10]. (Sites were defined as a collection of coffee plants of any extent; they were selected to represent the geographical, topographical, and ecological diversity of coffee farms in Puerto Rico.) At each site, the total number of fruits and bored fruits were counted for three branches each on three coffee plants. Counting three branches for three plants is standard for sampling CBB; CBB spatial distribution tends to be irregular [10]. Infestation was defined as the proportion of fruits showing CBB perforation [10,35].
The effect of presence of each species on the CBB infestation was evaluated with a generalized linear mixed model (GLMM). This model is useful when the data are hierarchically structured [36]; in this case the sites were nested within farms, and farms within municipalities. This model was used in previous CBB infestation studies [35]. CBB infestation rate by site was the response variable; the presence of plant other than coffee was used as a fixed effect. The date of sampling, municipality, farm, and site were used as random effects. Binomial error distribution was used to evaluate the effect of presence compared to absence of plants on CBB infestation. There were fifteen different possible models using the presence of plants as a fixed effect and adding and/or removing the random effects (date of sampling, municipality, farm, and site) sequentially. To determinate the best model, the Akaike Information Criterion (AIC) was applied. We used the cumulative distribution function for the standard logistic distribution on the output coefficients to present the results of infestation in a probability scale [37]. This was done with the plogis function in R 3.4.3 [38].

Plants Evaluated in the Field
Of the eight plant species with perforated fruits, individuals of the subfamily Scolytinae were found only in C. cajan, G. guidonia, I. vera, and Schefflera actinophylla (Endl.) Harms (Table 3).  Both total Scolytinae and H. hampei were found in the highest numbers in April, at the beginning of the rainy season but before coffee fruits were available to infest (Figure 1). Patterns of total Scolytinae and H. hampei per month were generally similar except in September, when the number of total Scolytinae rose but the number of H. hampei did not. Reproduction of Scolytinae was only observed in three of the eight plant species examined: C. cajan, G. Guidonia, and I. vera. The frequency of life cycle stages (eggs, larvae, pupae, and juveniles) was less than <1% of the total number of fruits examined (Table 4). These stages were not identified because they are not included in taxonomic keys.
Scolytinae and H. hampei per month were generally similar except in September, when the number of total Scolytinae rose but the number of H. hampei did not. Reproduction of Scolytinae was only observed in three of the eight plant species examined: C. cajan, G. Guidonia, and I. vera. The frequency of life cycle stages (eggs, larvae, pupae, and juveniles) was less than <1% of the total number of fruits examined (Table 4). These stages were not identified because they are not included in taxonomic keys.

Identification and Phylogenetic Analysis
A total of 3563 individuals of the subfamily Scolytinae were recorded (Table 3). Of the 587 individuals identified, 27 (4.6%) were identified as H. hampei; they were found in G. guidonia (12), I. vera (10), C. cajan (4), and S. actinophylla (1). These 27 individuals mostly came from samples taken between January and June 2015. In most cases, DNA barcoding agreed with identifications based on morphology. All beetles that were identified by barcoding as H. hampei in this study had 99% COI sequence similarity with H. hampei reference sequences [25], and they formed a clade with 100% bootstrap support (Figure 2

Identification and Phylogenetic Analysis
A total of 3563 individuals of the subfamily Scolytinae were recorded (Table 3). Of the 587 individuals identified, 27 (4.6%) were identified as H. hampei; they were found in G. guidonia (12), I. vera (10), C. cajan (4), and S. actinophylla (1). These 27 individuals mostly came from samples taken between January and June 2015. In most cases, DNA barcoding agreed with identifications based on morphology. All beetles that were identified by barcoding as H. hampei in this study had 99% COI sequence similarity with H. hampei reference sequences [25], and they formed a clade with 100% bootstrap support ( Figure 2). The 560 remaining individuals were identified as other species of Hypothenemus: 90 as H. obscurus, 45 as H. eruditus, 11 as H. seriatus, and 414 as unidentified Hypothenemus spp. (Table 3).

Preference and Feeding Tests
Ninety of the 100 female CBBs initially selected G. guidonia fruits over coffee fruits. Seventy CBBs that selected G. guidonia fruits penetrated the fruits by the second day. The remaining 10 CBBs initially selected coffee fruits; all of these, and two that originally selected G. guidonia fruits but did not penetrate them, drilled coffee fruits by the second day of observation (reaching the endosperm). Eventually, 76 of the 90 CBBs that first selected G. guidonia fruits left them to perforate coffee fruits. Fourteen CBBs remained in G. guidonia fruits at the end of the experiment but were dead. Reproduction was observed in 36 CBBs in coffee fruits, but reproduction and feeding were never observed in fruits of G. guidonia.
Feeding on the seeds of L. leucocephala and C. cajan by CBBs was not observed. In one case, a CBB penetrated a C. cajan seed and left scrape marks on the cotyledon, but no frass were found (which would provide evidence of feeding) and the CBB died without reproducing. The average survival of CBBs without food was 11 days.

Preference and Feeding Tests
Ninety of the 100 female CBBs initially selected G. guidonia fruits over coffee fruits. Seventy CBBs that selected G. guidonia fruits penetrated the fruits by the second day. The remaining 10 CBBs initially selected coffee fruits; all of these, and two that originally selected G. guidonia fruits but did not penetrate them, drilled coffee fruits by the second day of observation (reaching the endosperm). Eventually, 76 of the 90 CBBs that first selected G. guidonia fruits left them to perforate coffee fruits. Fourteen CBBs remained in G. guidonia fruits at the end of the experiment but were dead. Reproduction was observed in 36 CBBs in coffee fruits, but reproduction and feeding were never observed in fruits of G. guidonia.
Feeding on the seeds of L. leucocephala and C. cajan by CBBs was not observed. In one case, a CBB penetrated a C. cajan seed and left scrape marks on the cotyledon, but no frass were found (which would provide evidence of feeding) and the CBB died without reproducing. The average survival of CBBs without food was 11 days.

Association between Common Trees and CBB
The presence of the two most common trees on coffee farms was not significantly associated with CBB infestation ( Table 5). The probability of infestation of coffee fruits was 9.5% in the presence of I. vera and 13.5% in its absence, but this difference was not significant. The best model (selected by AIC) included date, farm, municipality, and site as random effects. Similarly, the probability of infestation was 9% in the presence of G. guidonia and 14% in its absence, but the difference was not significant; the best model selected included date, farm, and site as random effects.  [25]; H. seriatus, KX818303-11 [25]; H. eruditus, KX818248-51 [25]; and Hypothenemus spp., KY800209, KY800236, KY800240 [26]. The tree was rooted to the outgroup Xylosandrus compactus, KX818319 [25]. Origin of samples: PR: Puerto Rico; DR: Dominican Republic.

Association between Common Trees and CBB
The presence of the two most common trees on coffee farms was not significantly associated with CBB infestation ( Table 5). The probability of infestation of coffee fruits was 9.5% in the presence of I. vera and 13.5% in its absence, but this difference was not significant. The best model (selected by AIC) included date, farm, municipality, and site as random effects. Similarly, the probability of infestation was 9% in the presence of G. guidonia and 14% in its absence, but the difference was not significant; the best model selected included date, farm, and site as random effects. Table 5. Generalized linear mixed model (GLMM) estimates of the association between infestation rate of coffee and presence of Inga vera and Guarea guidonia in coffee farms.

Species
Estimated Intercept Estimated Coefficient Z P

Discussion
Can plants other than coffee serve as alternate hosts for H. hampei? The use of alternate hosts in the coffee farms of Puerto Rico was infrequent compared to infestation rates in coffee [10]: Only 27 individuals of H. hampei were found in 3894 fruits examined. Scolytid eggs, larvae, pupae, and juveniles were infrequent in the fruits of potential alternate hosts. Observations of H. hampei in fruits of alternate hosts were mainly during the season in which no coffee fruits were present, or were too immature to be penetrated. This suggests that the lack of coffee fruits may trigger this behavior. However, the fact that one individual of H. hampei was found on an alternate host during the coffee Agronomy 2020, 10, 228 8 of 12 fruiting season indicates that there may be other factors that cause H. hampei to penetrate alternate hosts. This is also the case with H. obscurus, which occasionally attacks coffee fruits in Hawaii despite having a preference for macadamia [39]. In lab tests, most CBBs initially penetrated G. guidonia fruits instead of coffee fruits, a surprising finding given that its preferred host is coffee, but almost all later moved to coffee fruits. It would be interesting to explore volatile compounds produced by G. guidonia fruits and their potential use as lures for the CBB.
Inga vera and G. guidonia were the alternate hosts in which the greatest number of H. hampei individuals were found, and there was a lower probability of infestation of coffee fruits when these species were present, although the effect was not significant. It is possible that these species grow better under certain environmental conditions which are less conducive to CBB infestation, and there is no direct relationship with CBB. However, it is also possible that these trees attract some H. hampei individuals who bore them and die without ever attacking coffee fruits. Further work on this relationship is needed. It might confer an added value to these species, not only as shade trees, but also as a tool in pest management.
In preference tests, fruits of G. guidonia were initially drilled, but only occupied for a short time before H. hampei moved to coffee fruits. It can thus be inferred that these fruits only served as temporary shelter. This behavior could explain the reports of individuals of H. hampei on plants other than coffee in several coffee-producing countries [6,12,16,40].
Although a single CBB penetrated and gnawed a C. cajan seed, there was no evidence of feeding. These results contrast with experiments using C. cajan seeds as an alternative food to maintain colonies of H. hampei in Guatemala [32]. In those experiments, normal development of all stages was observed on this plant, supporting the conclusion that the seeds of C. cajan meet the nutritional requirements of H. hampei. Similarly, the seeds of L. leucocephala, which were not used by H. hampei in our laboratory tests, were reported as a food source in laboratory experiments in Philippines [6]. These authors reported that another 21 species of plants, distributed in nine families were also used by H. hampei as food in the laboratory. In contrast, our laboratory results appear to agree with previous claims that H. hampei is monophagous [41].
These reports of polyphagous behavior of H. hampei might reflect the phenotypic plasticity of many species of the genus Hypothenemus, which allows them to use varied sources of food including different species and even decomposing material [7]. Many species have developed specialized feeding behaviors (e.g., host specificity) to reduce competition for food [42][43][44]. However, these behaviors can complicate survival when the food source is scarce [43,45]. Phenotypic plasticity may allow the more flexible individuals to use alternative resources [46].
For example, H. obscurus mainly attacks macadamia in Hawaii, but occasionally damages coffee fruits and coffee branches [39]. In certain cases, specialized species can maintain genotypic characteristics that allow them to take advantage of ancestral hosts [47]. Moreover, it is possible that differences in monophagy compared to polyphagy might reflect variation among beetles identified as H. hampei, either because beetles were misidentified in some studies, because there may be cryptic species that differ in behavior, or because some populations may have retained more phenotypic plasticity than others.
In this study, we found 3563 Scolytinae beetles in potential alternate hosts. Likewise, in a previous study many Scolytinae were found in fruits of several of the most important plants associated with coffee farms in Puerto Rico [34]. This is why the task of finding H. hampei individuals in fruits of alternate hosts is similar to searching for a needle in a haystack, and suggests that some of the beetles identified as H. hampei in previous studies may have belonged to other species. Similarly, a recent survey of >18,000 herbarium specimens of potential alternate hosts from coffee-growing regions of Africa found H. hampei in fruits of several species of Coffea, but not in any other plants [19].
The most abundant species identified here was H. obscurus followed by H. eruditus, which are among the 16 species of Hypothenemus in Puerto Rico [20]. H. obscurus is the most economically important Hypothenemus after H. hampei, as it may affect the seeds of many other crops [20,39]. In addition, H. obscurus and H. eruditus have been reported feeding on the coffee mesocarp [7,39], which may imply additional damage to the fruit, since they can facilitate entry of pathogens [48,49].
DNA barcoding was used to support morphological identifications of H. hampei. The H. hampei clade based on COI sequences was well-defined and had 100% bootstrap support (Figure 2). Previous studies on possible alternate hosts of the H. hampei did not use DNA barcoding to confirm identifications, a limitation considering the large number of similar species in Hypothenemus. In some cases, barcoding data did not support morphological identifications because the COI sequences had no hits >95% in GenBank (Figure 2). These individuals could be new species or species not represented in GenBank. The unidentified Hypothenemus in this study were variable in the COI region sequenced, but none was closely related to H. hampei. The COI phylogenetic tree (Figure 2) differed from a previous study in the relationship among H. hampei, H. obscurus, H. seriatus, and H. eruditus [25], but neither had strong bootstrap support at all the relevant nodes. This discrepancy, together with unidentified sequences mentioned above, the economic importance and the ubiquity of Hypothenemus, shows that more phylogenetic studies on this group are needed.

Conclusions
This study shows that H. hampei in Puerto Rico can occasionally use fruits of plant species other than coffee as a shelter, but there is no evidence that alternate hosts are used as a source of food or place of reproduction. Although our results are in agreement with the previous general observation on the monophagy of H. hampei, there are reports of polyphagy of this species in other coffee-producing countries (e.g., Guatemala and the Philippines), which highlights the importance of the study of alternative hosts of H. hampei and their implications for the management of this pest in different locations along the coffee production belt.
This study shows that alternate hosts are most likely not a significant source of re-infestation of plots after the dry period. Consequently, and taking into account the life history of the insect, the presence of the CBB might be significantly reduced by the management of the remnant coffee fruits in the plants after harvest [10], as well as the elimination of abandoned coffee farms which serve as a reservoir of CBB between crops. Of course, it is possible that CBB have other refugia that have not yet been found or studied.