Arthropod facilitation mediated by abandoned dead domatia

In ant–plant mutualistic interactions, plants provide shelter (domatia) and/or food to ants and in exchange, and ants provide protection against herbivores. After plant tissue senescence, ants are expected to abandon dead domatia, leaving these empty spaces available for other arthropods. In this study, we tested for the role of the mutualistic interaction between Cordia alliodora and Azteca pittieri in promoting new habitats for arthropods through abandoned dead domatia. We predicted that species richness, abundance, and colonization frequency of secondary arthropods would be greater in dead branch domatia (dead domatia) compared with live branch domatia (live domatia). During March 2019, we selected 38 C. alliodora trees in a Mexican tropical dry forest. For each tree, we collected five live and five dead domatia, for 380 domatia in total. We found six morphospecies of secondary arthropods colonizing live domatia, while 42 were present in dead domatia. Ants were the most species-rich group (10 species) in abandoned dead domatia and utilized them as nesting sites (25 nests). Secondary arthropod species richness, abundance, and colonization frequency were greater in dead domatia compared with live domatia. We concluded that the Azteca–Cordia mutualistic interaction is an important habitat facilitator by promoting new habitats for arthropods through abandoned dead domatia.


INTRODUCTION
Several insects modify plant structures to complete their life cycle, such as leafroller caterpillars, gall-inducing insects, bark, and woodboring beetles. These new habitats might persist on plants and can be later used as shelter by other arthropods in a facilitative interaction process (Vieira and Romero 2013, Zuo et al. 2016, Novais et al. 2018. Particularly, ants can take advantage of the abandoned shelters made by other insects for nesting, such as wood-boring beetle abandoned cavities (Tschinkel 2002, Satoh et al. 2016, Novais et al. 2017, senescent galls (Fernandes et al. 1988, Mehltreter et al. 2003, Almeida et al. 2014, Santos et al. 2017, and empty cocoons (Raath et al. 2017).
In symbiotic ant-plant mutualisms, myrmecophytic plants already possess specialized structures for housing ants, known as domatia. These structures are present in different plant parts, such as in hollow stems (e.g., Cecropia, Cordia, Leonardoxa, Macaranga), thorns (Acacia), petioles (Piper), or leaf pouches (e.g., Hirtella, Maieta, Scaphopetalum, Tococa; Heil and Mckey 2003). In addition to nesting spaces, some myrmecophytic plants feed ants directly by producing extrafloral nectar (Leonardoxa, African Acacia species), cellular food bodies (Cecropia, Macaranga, Piper), or both (Central American Acacia species), or indirectly via hemipteran trophobionts tended by resident ants (Heil and Mckey 2003). In exchange for the nesting spaces and food resources, ants protect their myrmecophyte hosts against a broad range of herbivores, as already demonstrated by many experimental studies with different myrmecophytic plant species (Fonseca 1994, Gaume et al. 1997, Vasconcelos and Casimiro 1997, Letourneau 1998, Heil et al. 2001, Alvarez et al. 2001. In specialized ant-plant mutualistic interactions, senescence of plant tissues is expected to cause the abandonment of dead domatia by ants because of the loss of food resources provided by the plants (e.g., extrafloral nectar, cellular food bodies, hemipteran trophobionts). Under such a scenario, abandoned dead domatia become empty spaces and may facilitate potential refuges for other arthropod species. To understand this new type of ecological facilitation, we test for the role of the well-studied mutualistic interaction between Cordia alliodora (Ruiz and Pav on) Oken (Boraginaceae) and Azteca pittieri Forel (Dolichoderinae; Pringle et al. 2011) in promoting new habitats for other arthropods through abandoned dead domatia.
Cordia alliodora is a Neotropical myrmecophitic tree with domatia at stem nodes. In the Mexican tropical dry forest, C. alliodora is mostly inhabited by the specialized ant species Azteca pittieri, although tree domatia may also be inhabited in very lower frequency by generalized stem-nesting species from other genera such as Crematogaster (Pringle et al. 2011). To access the domatia, founding ant queens gnaw entrance holes into unoccupied domatia starting egg posture, and as the colony grows, worker ants gnaw openings in other domatia throughout the tree and occupy them (Pringle et al. 2011). C. alliodora offers no direct food rewards for ants, and Azteca ants host hemipteran insects of the superfamily Coccoidea, known as coccoids, inside the domatia and eat their honeydew (Tillberg 2004, Pringle et al. 2011. Although coccoids compete with their host plant for resources by removing sap, Pringle et al. (2011) demonstrated that C. alliodora trees benefit from ant defense against other herbivores.
When a C. alliodora branch dies, it usually remains attached to the plant or fall and remains suspended on vegetation under C. alliodora canopies (S. Novais, personal observation). We observed that Azteca ants leave domatia in dead branches, and we asked whether these abandoned structures could be colonized by other arthropods. We hypothesized that ants in specialized ant-plant mutualistic interactions will abandon dead domatia after plant tissue senescence, allowing the colonization of these structures by other arthropods, and consequently, enhancing arthropod diversity. We predicted that species richness, abundance, and colonization frequency of secondary arthropods (non-Aztecacoccoids arthropods) would be greater in dead branch domatia compared with live branch domatia. We also investigated the density of C. alliodora trees and estimated the amount of live domatia per hectare in order to indirectly demonstrate the importance of the interaction between A. pittieri and C. alliodora trees in promoting new habitats for arthropod communities through abandoned dead domatia.

Study area and sampling design
This study was carried out in the Chamela-Cuixmala Biosphere Reserve (19°30ʹ N, 105°03ʹ W) in Jalisco, Mexico. The vegetation within the 13,142 ha of the reserve consists primarily of tropical dry forests (TDFs) with a mean annual rainfall of 748 mm, and a dry season that extends from November to June (Garc ıa- Oliva et al. 2002).
During March 2019, we selected 38 C. alliodora trees at least 20 m apart from each other. For each C. alliodora tree, we collected five live v www.esajournals.org domatia and another five dead domatia, for 380 domatia in total. Dead domatia were localized on dry branches still attached on live C. alliodora trees or in felled branches that remain suspended on vegetation under C. alliodora tree canopy. Each domatium was placed in an individual fine-mesh bag and transported to the laboratory. These domatia had an average diameter of 1.47 cm (SD AE 0.4 cm). Domatia were frozen to incapacitate any inhabitants, and then carefully opened. Arthropods were transferred to vials with 70% ethanol for further morphotyping to the level of order (Garc ıa-Aldrete and Ayala 2004). All mites were classified as morphospecies of Acari. Particularly, ant species were identified at genus or species level (see Acknowledgments). Species richness (number of morphospecies), abundance (number of individuals), and colonization frequency (percentage of colonized C. alliodora trees) of secondary colonizing arthropods (non-Azteca-coccoids arthropods) per C. alliodora tree were determined. The sampled arthropods were deposited in the entomological collection of the Laboratory of Evolutionary Ecology and Conservation of Tropical Forests of the National Autonomous University of Mexico.

Density of C. alliodora trees and live domatia
In order to estimate the density of C. alliodora trees/ha, we utilized data on the abundance of woody tree species sampled in nine 3 9 50 m plots established at the study site (see Lopezaraiza-Mikel et al. 2013). The density of C. alliodora live domatia/ha was estimated through the allometric regression model between the number of domatia and C. alliodora tree diameter found by Pringle et al. (2012) in a study conducted in the same tropical dry forest.

Statistical analysis
Generalized linear mixed models were used to test whether species richness, abundance, and colonization frequency of secondary arthropods (response variables) are higher in abandoned dead domatia than live domatia of C. alliodora trees. Domatium condition (dead and live) was used a fixed explanatory variable, and the 38 individuals of C. alliodora as a random effect. Ant individuals were not considered for abundance analysis due to the social character of the group. The Akaike information criterion (AIC) was used to rank the models; since it represents the uncertainty of the model, a lower value of the AIC represents the more parsimonious model. All analyses were performed using R software (R Core Team 2020).
We estimated that the density of C. alliodora trees was~20 individuals/ha at the study site. Of a total of 144 plant species sampled in the nine 3 9 50 plots, C. alliodora was the sixteenth woody tree species with the highest density. The average diameter of all C. alliodora trees registered was 9.97 cm (SD AE 5.1 cm); a C. alliodora tree with this diameter is expected to have~132 live domatia. Therefore, since the density of C. alliodora trees was~20 individuals/ha, we estimated the presence of~2640 live domatia/ha.

DISCUSSION
Our results demonstrated for the first time the role of ant-plant mutualistic interactions in promoting new habitats for arthropods through v www.esajournals.org abandoned dead domatia. All living branches of C. alliodora had at least one domatium with the presence A. pittieri individuals, confirming that this species has a specialized association with the host plant C. alliodora. This result also suggests that the few unoccupied domatia in live branches, which had no entrance holes, would probably be occupied as the Azteca colonies expanded. In contrast, only three percent of live domatia were colonized by Pseudomyrmex cubaensis and Crematogaster crinosa, suggesting that these species opportunistically occupy empty domatia in this system. After C. alliodora branches with domatia senesce and dried out,  Azteca ants leave these dead tissues probably because they no longer serve as a food source for the hemipteran coccoids that, in turn, are not able to provide food resources for ants. In fact, no dead domatium was found occupied by A. pittieri or coccoids, but these structures are reused in great majority by a different arthropod community that is not present in live domatia. The abandoned dead domatia, therefore, represent a safe and durable refugia in which many species of secondary arthropods can use as breeding sites, shelter, or nesting sites (Figs. 1, 2).
A recent study conducted in the same tropical dry forest demonstrated the role of wood-boring beetles as important habitat facilitators for arthropods, which utilize the abandoned cavities made by beetle larvae during their development in the wood (Novais et al. 2018). Comparing the arthropod community that reused both abandoned cavities (branches vs domatia), Novais et al. (2018) found that spiders (115 individuals), centipedes (37), and crickets (30) were the most abundant nonsocial arthropods in a set of 240 Spondias purpurea L. (Anacardiaceae) branch samples. In the present study, we found only four spiders, seven centipedes, and no crickets within a total of 380 domatia samples. These differences are probably related to the variation of entrance hole sizes between both sorts abandoned cavities. S. purpurea branches are primary colonized by more than twenty species of woodboring beetles of various sizes (1.59-21.97 mm; Calder on-Cortes et al. 2011), which leave cavities with a great variety of entrance sizes (0.49-6.25 mm; Novais et al. 2017), allowing the colonization by different-sized arthropod species (S. Novais, personal observation). However, the entrance holes made by Azteca ants in domatia are little and homogeneous (%2 mm; see Fig. 2), allowing only small arthropods to have access.
Ants were an important group colonizing the abandoned dead domatia as nesting sites, as briefly described by Longino (1996) for dead domatia of C. alliodora trees in Costa Rica. This author reported that ant species that are generalist inhabitants of dead wood (i.e., stems, branches, and twigs) can take advantage of the abandoned dead domatia as nesting sites. Since most twig-dwelling ants, except for carpenter ants in the genus Camponotus, cannot excavate sound wood to build their nests (King et al. 2018), they depend on other organisms to create new cavities for them, nesting almost exclusively in abandoned tunnels of wood-boring beetles (Tschinkel 2002, Powell et al. 2011, Novais et al. 2017. Similarly here, our results suggest that the entrance holes made by ants (e.g., Azteca) able to excavate live tissues in domatia of myrmecophytic plants facilitate the access of twigdwelling ants into dead domatia after their abandonment.
The availability of nesting sites represents a limiting resource for arboreal ants (Philpott and Foster 2005, Mottl et al. 2019, Novais et al. 2020). Previous studies have demonstrated an increase in arboreal ant abundance and distribution when the availability and diversity of nesting sites is increased (Sagata et al. 2010, Powell et al. 2011, Jim enez-Soto and Philpott 2015. In a recent study, Novais et al. (2020) showed that dry fallen v www.esajournals.org petioles of Cecropia trees represent an important nesting resource for arboreal twig-nesting ants, and therefore, they suggested that Cecropia trees are key components of tropical forests that maintain arboreal ant diversity by promoting a substantial increase in the availability of nesting sites for arboreal twig-nesting ants through dry fallen petioles. In the present study, we suggest that dead domatia of C. alliodora trees also represent an important nesting resource for arboreal twignesting ants. Compared with Cecropia trees that shed a few leaves each day and hundreds of leaves can accumulate on the vegetation below their canopy over time (see Novais et al. 2020), domatia of C. alliodora trees do not seem to dry out as often. However, contrary to Cecropia petioles which their pith is the most voluminous portion and consists of soft, spongy parenchyma, C. alliodora stems consist of sound wood (Devall et al. 1988, Wightman et al. 2001, and therefore, the abandoned dry domatia are likely to remain available for colonization by twig-nesting ants for a much longer period of time until their degradation compared with Cecropia petioles. We found that almost 88% of dead domatia that were occupied by ants did not contain an ant queen. Similar results have also been reported by other abandoned structures made by insects, such senescent galls and wood-boring beetle abandoned cavities, which are mainly colonized by ant species that develop polydomic colonies (Almeida et al. 2014, Santos et al. 2017, Novais et al. 2017. The construction of polydomic nests is a strategy to increase the v www.esajournals.org distribution of the colony in order to avoid predation and reduce local intraspecific competition for the same resources by bringing workers closer to new food sources (Debout et al. 2007, Lanan et al. 2011, Robinson 2014. In an experimental study to investigate whether colonies of Crematogaster torosa form new polydomous nests to better exploit temporally stable food resources, Lanan et al. (2011) found that when food was present, most colonies formed polydomous nests nearby.

CONCLUSIONS
We conclude that the Azteca-Cordia mutualistic interaction is an important habitat facilitator by promoting new habitats for arthropods through abandoned dead domatia. After senescence of plant tissues, the abandoned dead domatia are occupied by a diverse arthropod community that is not present in live domatia colonized Azteca ants, increasing species richness and abundance of arthropods locally. Ants were the most Fig. 3. Boxplots of species richness (a) and abundance (b) of secondary arthropods per tree colonizing live and dead domatia of Cordia alliodora in a tropical dry forest, in Jalisco, Mexico. The horizontal thick gray band represents the median value, and the boxplot margins indicate first and third quartiles, the whiskers represent the maximum/minimum value within one and a half times the interquartile range, and each dot represents the arthropod species richness and abundance per tree (N = 38). Single asterisk represents significant differences (P < 0.05). species-rich group in the abandoned dead domatia, which take advantage of these structures left by Azteca ants to expand their polydomic colonies. At large scales, dead wood in the form of dying and dead trees, standing or fallen branches at different stages of decay are important resources (e.g., food, shelter, and breeding sites) for the conservation of a large number of species, such as saproxylic insects, lichens, bryophytes, birds, and mammals; Radu 2006). At a fine scale, we highlight the importance of dead plant tissues in the form dead branches attached on living trees and fallen branches hanged on vegetation as potential refugia that maintain arthropod diversity, which can be mediated by insects such as wood-boring beetles (Novais et al. 2018) or through abandoned dead domatia as demonstrated in the present study.

ACKNOWLEDGMENTS
We are grateful to Dora Luz Martinez Tlapa (INE-COL) for the identification of ant species. We are grateful to Gumersindo S anchez-Montoya for field help. We are grateful to anonymous reviewers for their comments on an earlier version of the manuscript. This study was supported by grants from Universidad Nacional Aut onoma de M exico (PAPIIT # IN212714-3, IV200418), SAGARPA-CONACYT 291333, CONA-CYT-UNAM-UAGro to LANASE (2015-LN250996, 2016-LN271449, 2017-LN280505, 2018-LN293701, and 2019-299033 and Programa Iberoamericana de Ciencia y Tecnolog ıa para el Desarrollo RED CYTED SEPODI (417RT0527). Finally, we thank the Chamela-Cuixmala Biosphere Reserve for all the physical structure and logistical support.