1Non-adaptive host-use specificity in tropical armored scale insects
2
Daniel A. Peterson1§, Nate B. Hardy2§*, Geoffrey E. Morse3, Takao Itioka4, Jiufeng Wei5, and
3
Benjamin B. Normark1
4
5
1. Department of Biology and Graduate Program in Organismic and Evolutionary Biology, University
6
of Massachusetts, 221 Morrill Science Center III, 611 North Pleasant Street, Amherst, MA 01003,
7
USA; 2. Department of Entomology and Plant Pathology, Auburn University, 301 Funchess Hall,
8
Auburn, AL 36849, USA; 3. Department of Biology, University of San Diego, 5998 Alcalá Park, San
9
Diego, CA 92110, USA; 4. Graduate School of Human and Environmental Studies, Kyoto University,
10
Yoshida-Nihonmatsu-cho, Sakyo-ku, Kyoto 606-8501, Japan; 5. College of Agriculture, Shanxi
11
Agricultural University, Taigu, Shanxi, 030801, China
12
13
§ DAP and NBH share first authorship.
14
* corresponding author; e-mail: n8@auburb.edu.
15
16
ORCIDs: Peterson, http://orcid.org/0000-0002-3024-3068; Hardy, http://orcid.org/0000-0001-6133-
17
7086; Morse, https://orcid.org/0000-0002-5763-8246; Wei, https://orcid.org/0000-0003-4705-2599;
18
Normark, http://orcid.org/0000-0002-6267-9552.
19
20
21
22
Running head: Non-adaptive specialization
23
1
24
Abstract
25
Most herbivorous insects are diet specialists in spite of the apparent advantages of being a generalist.
26
This conundrum might be explained by fitness trade-offs on alternative host plants, yet evidence of
27
such trade-offs has been elusive. Another hypothesis is that specialization is non-adaptive, evolving
28
through neutral population genetic processes and within the bounds of historical constraints. Here we
29
report on a striking lack of evidence for the adaptiveness of specificity in tropical canopy communities
30
of armored scale insects. We find evidence of pervasive diet specialization, and find that host-use is
31
phylogenetically conservative, but also find that more-specialized species occur on fewer of their
32
potential hosts than do less-specialized species, and are no more abundant where they do occur. Of
33
course local communities might not reflect regional diversity patterns. But based on our samples,
34
comprising hundreds of species of hosts and armored scale insects at two widely separated sites, more
35
specialized species do not appear to outperform more generalist species.
36
37
Keywords: Diaspididae, ecological specialization, herbivory, host range, niche breadth, polyphagy
38
39
Introduction
40
High species diversity has been attributed to the partitioning of available resources into narrow
41
ecological niches (Hutchinson 1959). Yet niche breadth varies greatly between species. Herbivorous
42
insects are classic subjects for the study of this variation (Futuyma and Moreno 1988; Hardy et al.
43
2020). Although diet specialists prevail, diet breadths vary continuously, and in some species are
44
extremely broad (Normark and Johnson 2011; Forister et al. 2015). How did this come to be?
45
46
For the most part, theorists have worked from the premise that diet specialization comes from genetic
47
trade-offs between adaptations to alternative resources, specifically antagonistic pleiotropy between
2
48
alleles at a few diet-determining loci (Futuyma and Moreno 1988; Ravigné et al. 2009). Although
49
empirical evidence for such genetic trade-offs is scarce (Futuyma 2008; Forister et al. 2012), they
50
might be difficult to detect, as they can be hidden by inter-individual fitness variation at linked loci
51
(Joshi and Thompson 1995), or arise from epistatic interactions between alleles (Remold 2012;
52
Rodriguez-Verdugo et al. 2014; Celorio-Mancera et al. 2016). In sum, the evidence is scarce for the
53
adaptive trade-off hypothesis, but it is difficult to falsify outright.
54
55
Alternatively, niche specialization could be driven by non-adaptive processes (Futuyma et al. 1995;
56
Gompert et al. 2015). In fact, theoretical spatial models have shown that adaptive trade-offs are not
57
necessary to produce niche-breadth distributions resembling those observed in natural communities
58
(Forister and Jenkins 2017). Evolving the ability to use a novel host almost certainly entails directional
59
selection. But alleles promoting fitness on other potential hosts can be lost through genetic draft during
60
strong directional selection on a novel host (Neher 2013), or genetic drift when insect and host
61
distributions cease to overlap (Gompert et al. 2015). One way or another, if host-use traits are easy to
62
lose but difficult to get back, neutral genetic processes could pull populations towards niche
63
specialization (Hardy et al. 2016).
64
65
Host-use trade-offs in herbivorous insects have traditionally been investigated by comparing
66
performance across multiple host plants of different insect genotypes within a population. But a
67
phylogenetically-informed comparison of host-use across multiple herbivore species offers a
68
complementary perspective that may be less obscured by short-term genetic contingencies (Funk et al.
69
1995; Futuyma 2010; Hardy and Otto 2014; Peterson et al. 2015, 2016). To wit, it could illuminate the
70
overall relationship between diet-breadth and ecological performance. If host-use specificity is
71
adaptive, we would expect that on any shared host specialists would tend to perform better than
3
72
closely-related generalists. Likewise, at the meta-population level, if host-use specificity is adaptive,
73
we might expect specialists to do a better job of colonizing specific host resources (Gryllenberg and
74
Metz 2001). Conversely, if specificity is non-adaptive, we would expect generalists to colonize more of
75
their potential hosts, and to perform just as well as specialists on shared hosts, or even perform better if
76
there is a population-genetic cost for specificity, for example reduced population size and more erosive
77
genetic drift.
78
79
We sought evidence of such performance differences in the relative abundances and patch occupancies
80
of 171 putative armored scale insect species (Hemiptera: Diaspididae) across 138 tree species in
81
tropical rainforest communities on two continents. As is the case for herbivorous insects in general,
82
most diaspidids are host-use specialists, but some can be extremely polyphagous (García Morales et al.
83
2016). Diaspidids are sessile and have a simple, pathogen-like life history in which new host trees are
84
colonized by wind-dispersed first-instar nymphs that cannot survive for long away from a host (Hardy
85
2018). A first-instar nymph landing on a host thus experiences something like a no-choice feeding trial,
86
in which it must either successfully develop on that host, or die. Once an individual starts to feed, it
87
loses its legs and never moves to another feeding site. Potential for host-choice is therefore limited and
88
occurrence of later-instar life stages on a plant is a clear indication that it is a suitable host for
89
development and reproduction (Hill and Holmes 2009). A female completes her development,
90
reproduces, and dies at the site where she settled as a first instar. A male regains motility as an adult —
91
but adult males are completely non-feeding and of course cannot establish new colonies on new host
92
plants. Because dispersal is mediated by wind, it presumably cannot be adaptively directed towards
93
favored hosts — rather, any host in the local environment is about equally likely to be colonized by any
94
given species of diaspidid. And the presence and abundance of a particular species of diaspidid on a
95
given host plant is a direct consequence of the ability or inability of members of that diaspidid species
4
96
to successfully develop on that host. In other words, in diaspidids, there is an unusually simple and
97
direct causal connection from ecological performance on a host to occupancy of that host and
98
abundance on it. Therefore, presence and abundance of diaspidids on alternative hosts are useful
99
indices of ecological performance on those hosts. Furthermore, with random, time-limited dispersal,
100
one might expect the greatest fitness for genotypes that perform best across most of the commonly
101
encountered host-plants. In fact, for diaspidids, we have previously shown that when host associations
102
are treated as a binary use-or-nonuse traits, the phylogenetic patterns of host use are incompatible with
103
strong adaptive trade-offs (Peterson et al. 2015). Nevertheless, we have not previously been able to
104
account for potentially important quantitative differences in performance across host plant groups.
105
106
Our approach was to (1) estimate allele genealogies among the sampled diaspidids for 3 loci, using
107
DNA sequence data; (2) estimate species boundaries using these genealogies and also using
108
morphology; (3) estimate the degree to which host use is phylogenetically conservative; (4) explicitly
109
test for diet specialization in each species; and (5) use abundance-based and patch-occupancy-based
110
indices of performance to test if more specialized species tend to do better than less specialized species
111
on shared hosts.
112
113
Methods
114
What to measure? Diet-breadth dependent performance trade-offs could result from any number of
115
mechanistic interactions between a herbivorous insect and a host plant. On a particular host plant, in
116
comparison to a more specialized species, a relative generalist might have (1) a reduced ability to
117
initiate feeding, (2) a lower feeding rate, (3) less efficient utilization of host nutrients, (4) greater
118
susceptibility to host defenses, or (5) more exposure to natural enemies. No matter the mechanism, any
119
trade-offs that drive the evolution of specialization would need to ultimately limit survival or fecundity.
5
120
If specialization is an adaptive response to trade-offs between performance on alternative hosts, more
121
specialized species should have higher survival or fecundity than less specialized species on shared
122
resources. In the studied tropical forest plots, we were not able to measure survival or fecundity
123
directly, but we were able to measure the abundance and patch occupancy of each diaspidid species on
124
each host-plant species. As mentioned earlier, because after the first-instar stage each diaspidid is stuck
125
for life on one host, an observation of a second-instar or adult individual on a host is evidence of
126
successful development on that host (Hill and Holmes 2009). Moreover, the relative abundance of
127
diaspidid species on each host-plant species is an integrative proxy for fitness – integrating across host-
128
dependent differences in diaspidid fecundity and survival. Although local patterns of abundance and
129
patch occupancy may be susceptible to stochastic variation and thus are only proxies of fitness,
130
regional-scale differences in a diaspidid species’ abundance on alternative hosts map directly to its
131
relative fitness on those hosts; if across its geographic distribution a diaspidid population has high
132
survival and fecundity on a host plant species, across its range it will be abundant on that host species.
133
Thus our inferences about the adaptiveness are contingent only on local samples reflecting broader
134
spatial patterns.
135
136
Sampling. We surveyed diaspidids at two wet lowland evergreen tropical rainforest sites: (1) San
137
Lorenzo National Park, Panama and (2) Lambir Hills National Park, Malaysia (on the island of
138
Borneo). The Panama site, with 3152mm of rainfall per year, is at 30m elevation near the Caribbean
139
coast within a large expanse of protected forest. The Malaysia site, with 2700mm of rainfall a year, is
140
at 160m elevation in a small protected remnant forest that is now mostly surrounded by recently-
141
cleared areas converted to oil-palm plantations. Both sites have high tree diversity; the Borneo site in
142
particular has exceptionally high diversity at the species level but is largely dominated by the family
143
Dipterocarpaceae (Basset et al. 2003). These sites were chosen because they provided access to the
6
144
forest canopy with a crane, had tagged and data-based each individual mature tree, and because the
145
diaspidid faunas in those canopies were diverse. We were not able to search each tree in each plot, so
146
we used the tree individual database at each site to divide tree specimens into sampling groups of one
147
randomly-selected individual per tree species. Only trees over 10 cm diameter at breast height were
148
considered. Otherwise, our samples were not biased by the age or size of tree specimens (or the age and
149
size of researchers). We did not sample any tree individual more than once, so tree species with only
150
one individual were present only in the first round of sampling, those with two individuals were present
151
in the first two rounds, and so on. This protocol allowed us to sample across the full diversity of host
152
taxa while also getting multiple samples from common host species.
153
154
In Panama we surveyed 90 trees over three rounds of sampling, representing 53 species, 48 genera, and
155
29 families (Table S1). In Malaysia, we surveyed 211 trees over 20 rounds of sampling, including 85
156
species, 48 genera, and 27 families (Table S2)
157
158
The canopy crane at each site was used to access canopy-foliage for focal tree. From a gondola
159
suspended from the canopy crane, at each focal tree we spent 20 person-minutes searching accessible
160
foliage. Any leaves and twigs that we saw were infested by scale insects, we cut from the tree and
161
collected. From each tree we also haphazardly took one 20 cm twig sample and one 20 cm2 bark
162
sample. Removed plant material was stored in plastic bags and transferred to the lab for processing
163
under magnification; live diaspidids were cut from the surrounding plant material and preserved in 95%
164
ethanol. Specimens were subsequently sorted to life stage and second-instars and adult females were
165
regarded as evidence of successful establishment.
166
7
167
Phylogenetics. DNA was extracted from all second-instar and adult female armored scale insects using
168
Qiagen DNeasy Blood & Tissue kits (Qiagen, Valencia, CA) following the procedure outlined in
169
Normark et al. (2019). We amplified three loci that have previously been used for diaspidid
170
phylogenetics: elongation factor 1-α (EF1α), part of the large ribosomal subunit rDNA gene (28S), and
171
a part of the mitochondrial genome spanning cytochrome c oxidase I and II (COI-II). PCR primers and
172
protocols followed Andersen et al. (2010) and Gwiazdowski et al. (2011). PCR products were
173
visualized using 1.5% agarose gels with SYBRsafe (Invitrogen, Carlsbad, CA, USA) and successful
174
reactions were purified with Exo SAP-IT enzymatic digestion (Affymetrix, Cleveland, OH, USA).
175
Sanger sequencing of the PCR products was completed by Macrogen (Cambridge, MA, USA) or Eton
176
Biosciences (San Diego, CA, USA). DNA sequences have been submitted to GenBank under accession
177
numbers MT641780 – MT642048 and MT676866 – MT677529; some sequences have been previously
178
published in connection with phylogenetic studies and these are given in Tables S1 and S2.
179
180
For each site, phylogenetic relationships among all sampled individuals were estimated from the DNA
181
sequence data. Sequences from each genetic locus were aligned using PASTA (Mirarab et al. 2014),
182
and alignments were trimmed to include only sites with non-gap sequence for at least 80% of
183
specimens (Capella-Gutiérrez et al. 2009). Genealogies were inferred using the GTR+CAT model in
184
RAxML (Stamatakis 2014). The three single-locus alignments were then combined as one supermatrix,
185
from which we also inferred a phylogeny with RAxML. For use in comparative analyses, we made a
186
version of the phylogeny with just one tip per species, and scaled branch lengths to time using an auto-
187
correlated model of among-lineage rate variation, fit with penalized likelihood as implemented in
188
treePL (Smith and O’Meara 2012), and constraining the armored scale root to be 50-75 million years
189
old (Vea and Grimaldi 2016).
190
8
191
Species Delimitation and Identification. In an attempt to make our inferences robust to errors in species
192
delimitation, we delimited species in two ways. First, we delimited putative species with a version of
193
the genealogical concordance method (as in Gwiazdowski et al. 2011). All clades shared by at least two
194
gene trees, and not contradicted by the third gene tree, were considered evolutionarily independent
195
lineages. Species were defined provisionally as the most inclusive independent lineages containing at
196
least three terminal branches and no more exclusive independent lineages. This method precludes
197
delimitation of species represented by fewer than three specimens. To work around this problem, we
198
calculated the minimum divergence between provisional species clades, and used that value as a
199
maximum threshold for within-species divergence. Any specimens separated by more than this distance
200
from all other specimens were also considered distinct species.
201
202
We also delimited and identified them according to standard morphological criteria to the extent that
203
this was possible. Because second instars and adults were both included in this study, whereas standard
204
keys and descriptions are based on adults only, direct morphological comparisons and identifications
205
were not always possible. The analyses below were repeated for DNA-based and morphology-based
206
species delimitations. We retained all specimens in both analyses, whether or not they were
207
morphologically identifiable; for a few specimens that were not morphologically identifiable, in the
208
morphology-based analysis we defaulted to the DNA-based species.
209
210
Statistical Analysis. We characterized host-use specialization by diaspidid species in two ways, each
211
applied at three levels of host plant taxonomy (species, genus, and family). This allowed us to assess
212
the sensitivity of our inferences to different units of host plant diversity, and to measure the degree to
213
which host-use constraints were hierarchical. First, we quantified diet specificity; we asked whether
214
diaspidids used less diverse hosts than expected by chance. Concretely, for each diaspidid species, we
9
215
quantified host-taxon diversity using Simpson’s Reciprocal Diversity Index (RDI), which is essentially
216
evenness-corrected host-taxon richness. We compared empirical RDIs to those expected under a null
217
model of random host use. We simulated 1000 null data sets by randomly permuting the associations
218
between diaspidid species and individual host trees; then for each permutation, we again calculated the
219
mean RDI for the hosts of each diaspidid. With this approach, a diaspdid species is specialized to the
220
extent that its host RDI is lower than expected under the null model.
221
222
In a second view of host-use specialization, we calculated the phylogenetic conservatism of host use
223
across diaspidid species. In other words, we asked if evolutionary history constrains host use. We used
224
the R package (R Core Team 2017) MCMCglmm (Hadfield and Nakagawa 2010) to measure the
225
phylogenetic signal of host use by estimating the proportion of variance in the binary use-or-non-use of
226
each host taxon that could be explained by the diaspidid phylogeny. Empirical values for phylogenetic
227
signal were then compared to those calculated under a null model. Null data sets were produced by
228
randomly swapping associations between diaspidid species and host taxa until the associations were
229
thoroughly shuffled (the number of random swaps was 10 times the overall number of associations).
230
This preserved the empirical distribution of diet breadths while randomizing specific associations. P-
231
values for the empirical phylogenetic signal values were calculated using a Z-test against each
232
parameter’s null data set values (which were approximately normally distributed). We corrected for
233
multiple comparisons by assigning statistical significance according to a false discovery rate (FDR;
234
Benjamini & Hochberg 1995) of 0.05. The FDR procedure was conducted separately for each host-
235
taxon level because these analyses were not independent, and must be interpreted as alternative
236
configurations of the same data.
237
10
238
We investigated the strength of performance trade-offs by calculating for each host tree taxon the
239
correlation between diaspidid diet breadth (count of host taxa) and mean abundance. If performance
240
trade-offs are strong, on any given host taxon, we expect more generalist (less specialized) species to
241
be less abundant than more specialized species. We also investigated the relationship between diet
242
breadth and the proportion of host trees of a taxon colonized at each site, as patch occupancy may be a
243
better indicator of fitness than local abundance in a metapopulation of discrete colonies (Gyllenberg
244
and Metz 2001). Using R, we fit generalized linear models. For local abundance, the response variable
245
was the number of diaspidid individuals identified per host tree, assuming a Poisson distribution. For
246
metapopulation colonization-rate, the response variable was the probability that an individual tree
247
within each host taxon would be colonized by a diaspidid species, assuming a binomial distribution and
248
excluding host taxa with fewer than 3 trees surveyed. Both models only incorporated data for host-
249
taxon-by-diaspidid associations with at least one record. To assess statistical significance, we compared
250
empirical coefficients to those estimated from 1000 null data sets, produced by randomly permuting the
251
empirical data.
252
253
The scripts used for the analysis will be made available at Dryad.
254
255
Results
256
DNA-based species delimitations. In Panama we found live diaspidids on 75 trees, yielding 380 female
257
specimens (adults and second instars). At least two loci were successfully amplified for 184 specimens,
258
belonging to 53 DNA-delimited species (Fig. A1.4; Table A1.1). Assignment to a morphologically
259
defined species was possible for 180 specimens, representing 32 described and 12 undescribed species.
260
Species assignments and trophic links are in Table S1. In Malaysia, we found live diaspidids on 102
261
trees, yielding 480 female specimens. At least two loci were successfully amplified for 266 specimens,
11
262
belonging to 123 DNA-delimited species (Fig. A1.5 Table A1.1). Assignment to a morphologically
263
defined species was possible for 259 specimens, representing 20 described and 58 undescribed species.
264
Species assignments and trophic links are in Table S2.
265
266
We found strong evidence for host-use specialization, both in terms of less-than-expected host plant
267
diversity, and more-than-expected phylogenetic conservatism of host use. The Simpson’s RDI of each
268
diaspidid species’ diet was significantly lower than expected at all host taxonomic levels and in both
269
locations, except at the host-species level in Panama (Table 1). Phylogenetic signal was significantly
270
stronger than its null expectation for 19 host taxa (Figure A1.6), although it was higher at the Malaysia
271
site (mean 0.61) than the Panama site (mean 0.45), with 18 Malaysian host taxa with significant
272
phylogenetic conservatism, compared to just one Panamanian host taxon with significant conservatism.
273
274
Despite the prevalence of diaspidid host-use specialization at our two sites, and of extensive
275
phylogenetic conservatism of host use in Malaysia, we found no evidence for performance trade-offs
276
on alternative hosts that would select against broad diets. More specialized species were no more
277
abundant than less specialized species on specific host trees; the number of live adult or second instar
278
female diaspidids found on each tree was not correlated with diet breadth (Fig. 1; Table 2). Moreover,
279
contra the metapopulation trade-off hypothesis, diaspidids with broader diets were observed on a higher
280
proportion of the trees in their host taxa, although this effect was not significant for Malaysian
281
diaspidids and their host plant species (Figure 2; Table 2).
282
283
Morphology-based delimitations. The results of the analyses using morphologically delimited species
284
were broadly consistent with those using DNA-delimited species. As for analyses using DNA-delimited
285
species, with morphology-delimited species we found that diaspidid species were more specialized than
12
286
expected by chance (Table A1.2), that more specialized species were no more abundant on their hosts
287
(Table A1.3), and tended to occupy a smaller proportion of their potential host plants (Table A1.3).
288
289
Discussion
290
Diaspidid species in tropical rainforest canopy habitats appear to use as hosts only a small proportion of
291
the plant taxa in their local environment; simply put, as is the case for herbivorous insects in general
292
(Forister 2015), diaspidids tend to be exhibit diet specialization (Figure A1.7). But across the hundreds
293
of trees that we surveyed, more specialized species were no more abundant on their hosts than more
294
generalist species, and occurred on a smaller proportion of their potential host plants. Is specialization
295
for these diaspidids non-adaptive?
296
297
Such a conclusion would hinge on the assumption that what we saw within the reach of canopy cranes
298
is what we would have seen elsewhere. But if abundance varies much over space, local differences in
299
abundance could be misleading. Some such heterogeneity in the spatial distribution of diaspdids is
300
expected. The quality of specific host-plant resources can vary due to spatial mosaics of natural enemy
301
pressure (Heard et al. 2006), as well as host-plant features such as genotype, induced defensive state,
302
and physical structure (Dixon 2005). Although we saw no abundant species among those that were the
303
most specialized, each could have been abundant somewhere else in the forest, where more suitable
304
resources occur. Nevertheless, extreme patchiness in the abundance of the most specialized species
305
would entail a meta-population fitness cost, as local catastrophes would be more likely to cause
306
extinction (Nurmi & Parvinen 2008). In sum, potential spatial variation in abundance keeps us from
307
making definite conclusions about the adaptiveness of specificity in diaspidids. But this potential is
308
diminished by the cost of meta-population patchiness, and the consistency of our observations across
309
species and communities.
13
310
311
Our inferences should be robust to temporal variation in the sampled diaspidid populations. Of course,
312
when we sampled each site, some diaspidid species could have been under-represented because of their
313
phenology or the vagaries of local population dynamics. But we sampled many diaspidid species at
314
each site, and see no basis to suspect that phenology would vary systematically with diet breadth, or
315
that lows in stochastic populations fluctuations would occur predominantly in the most specialized
316
species. In sum, species-specific temporal variation in population size and age structure added
317
statistical noise to our data, though which the signal of strong performance in generalists was strong
318
enough to be discerned.
319
320
We found that the use of many host taxa by diaspidids was phylogenetically conservative. Although
321
such conservatism of host use has been found for several other groups of herbivorous insects, such as
322
butterflies (Janz et al. 2001) and beetles (Kelley and Farrell 1998), it has a special significance for
323
diaspidids, as they colonize new hosts haphazardly via wind (Magsig-Castillo et al. 2010), and our
324
previous work (Peterson et al. 2015), along with the research presented here, suggest that constraints on
325
host-use may be non-adaptive. Consequently, the phylogenetic conservatism of host-use in diaspidids
326
may more likely denote historical constraints on contemporary niches than long-term niche
327
optimization. Given the evidence of a lack of performance trade-offs for diaspidids between alternative
328
hosts (Peterson et al. 2015), host-use constraints would seem to persist in the face of what may be
329
strong selection for broad diets.
330
331
Our results also shed light on the complexity of host-use traits in plant-feeding insects (Barrett and Heil
332
2012; Forister et al. 2012). We found that specialization in armored scale insects occurs at all three of
333
the host-taxonomic levels that we considered (species, genus, and family), suggesting that the genomic
14
334
architecture of host-use traits is both complex and hierarchical. Use of multiple hosts is often associated
335
with close phylogenetic relationships among those hosts (Gilbert and Webb 2007; Krasnov et al. 2012),
336
yet such results in flying insects may reflect host-preference or ease of host recognition more than host
337
performance (Bernays 2001). Because diaspidids have little opportunity to choose a host, phylogenetic
338
conservatism at multiple taxonomic levels implies that performance on a host likely depends on many
339
traits of various effect sizes. Although actual mechanisms are as yet unclear (but see Hogenhout and
340
Bos 2011; Ali and Agrawal 2012), the involvement of many genetic loci in plant-insect interactions is
341
consistent with both ecological (Singer and Stireman 2005) and genetic (Remold 2012) theory as well
342
as recent genome-wide association studies (e.g., Egan et al. 2015; Gompter et al. 2015).
343
344
Our DNA-based species delimitation allow us some insight into whether any species that have been
345
characterized as extremely polyphagous (Normark & Johnson 2011; Normark et al. 2014) are in fact
346
clusters of cryptic species that are more specialized. The answer is mixed. On the one hand, in Panama,
347
the single most polyphagous species in the sample, Selenaspidus articulatus (Morgan) shows no hint of
348
cryptic species diversity — not surprisingly, as it is native to Africa and invasive in Panama (Normark et
349
al. 2019). On the other hand, several other reportedly highly polyphagous species do appear to
350
represent cryptic species clusters. In Panama, only a single morphologically-delimited species shows
351
evidence of cryptic diversity: samples of Diaspis boisduvalii (Signoret) were apportioned across five
352
DNA-delimited species. In contrast, at the Malaysian site, cryptic diversity appears rampant, especially
353
among the most polyphagous species: Chrysomphalus dictyospermi (Morgan), purported to use 80 host
354
families worldwide, was recovered as two cryptic species; Chrysomphalus pinnulifer (Maskell), with
355
40 host families worldwide, was also recovered as two cryptic species; Morganella longispina
356
(Morgan), 22 host families worldwide, three cryptic species; Aonidiella inornata McKenzie, 24 host
357
families worldwide, three cryptic species. We also found cryptic diversity in less polyphagous
15
358
Southeast Asian species: Silvestraspis uberifera (Lindinger), three cryptic species, and Aulacaspis
359
calcarata (Takagi), eight cryptic species, as well as in several undescribed species. Most strikingly, one
360
undescribed species provisionally designated Sishanaspis ud4977 appears to comprise a complex of 10
361
cryptic species. The upshot is that in Malaysia traditional morphology-based species delimitation seems
362
to miss much of the true diversity. But our inferences about the extent and consequence of diet
363
specificity in diaspidids appear robust to how species are delimitated.
364
365
Although it falls outside of the main theme of this study, one other insight afforded by the
366
morphological species identifications, which may help explain difference in diet breadth and host
367
occupancy observed between the two sites, is the incidence of invasive species. At the Malaysian site
368
we found no genera native to regions other than Southeast Asia, whereas in Panama nearly half of
369
morphologically identifiable individuals (77/180 = 43%) belong to invasive species (Normark et al.
370
2019). In addition to Selenaspidus articulatus (sampled on 18 host species), these include several
371
genetically uniform populations that we sampled on multiple host species, including Lepidosaphes
372
rubrovittata (Cockerell) (6 hosts), Chrysomphalus dictyospermi (4 hosts), Aspidiotus excisus Green (3
373
hosts), and Lepidosaphes punicae Laing (3 hosts). Thus the narrower diets and higher host occupancy
374
in Panama could have something to do with the relatively recent arrival of much of the diaspidid fauna,
375
although such an effect of geographic range expansion on diet breadth would be the opposite of what
376
has been found for some other herbivorous insects (Lancaster 2020).
377
378
In conclusion, evolutionary fitness is difficult to measure and we can not draw straight lines connecting
379
it to differences in local abundance and patch occupancy. It could be that for diaspidids the quality of
380
host resources is extremely uneven across tropical canopies, and that for each of the relatively
381
specialized species we sampled there was an unsampled population booming somewhere else in the
16
382
forest. Or it could simply be that host specialization is not adaptive for wind-dispersed plant pathogens
383
in diverse host-plant communities. If host-use specialization is adaptive and high quality hosts are
384
patchy and rare, then the question becomes this: why are the most specialized species so much less
385
abundant than expected? What are the conditions that must be met for a specialist to make good on
386
their specialty? But as it stands, the patchiness and rarity of specialist-supporting host plants are ad hoc
387
hypotheses, that is, extraneous additions to the theory of adaptive host-use specialization to prevent its
388
falsification.
389
390
Acknowledgements
391
We thank the Smithsonian Tropical Research Institute for facilitating our fieldwork in Panama. Our
392
field survey at Lambir Hills National Park was conducted in accordance with the Memorandums of
393
Understanding signed between the Sarawak Forest Department (SFD, Kuching, Malaysia) and the
394
Japan Research Consortium for Tropical Forests in Sarawak (JRCTS, Sendai, Japan) in December
395
2012. Thanks to Mohd. Shahbudin Sabki, Engkamat Lading, and Mohamad bin Kohdi, Paulus Meleng
396
of SFD for help in obtaining research permission at the Lambir Hills National Park. We thank Tohru
397
Nakashizuka (Tohoku University, Sendai, Japan) and Tomoaki Ichie (Kochi University, Kochi, Japan)
398
for their support. Diaspidid sample processing was by Katelyn Mullen, Ryan McCarthy, Mitchel
399
Logan, Michael Fizdale, Kyara Romeu, Shannon Trujillo, Hannah Shapiro, and Anna Whitfield. DAP’s
400
dissertation committee (Michael Hood, Adam Porter and Laura A. Katz) and anonymous reviewers
401
provided helpful suggestions that have improve the manuscript. JW’s participation was supported by
402
the China Scholarship Council. This work was supported by the National Science Foundation (DEB-
403
1258001; DEB-1744552) and by Grants-in-Aid (no. 21255004 to TI) from the Japanese Ministry of
404
Education, Science and Culture.
17
405
Figures
406
Figure 1. We found no relationship between the observed host range of diaspidid species and their
407
abundance on each individual host. Here we plot every tree colonized by each diaspidid species
408
independently, and dot area is proportional to the number of data points at that coordinate. Results are
409
divided by location and host taxonomic level: a) Panama, species; b) Panama, genus; c) Panama,
410
family; d) Malaysia, species; e) Malaysia, genus; f) Malaysia, family. None of these relationships (as
411
fitted by a linear model, dashed line) was statistically different from expectations under a null model
412
(all P > 0.9).
18
413
414
Figure 2. Diaspidid species with larger host ranges were present on a higher proportion of the
415
individual trees in their host range. Here each observed host-taxon-by-diaspidid-species interaction is
416
plotted independently, although host taxa with fewer than three tree individuals surveyed were
417
excluded from this analysis. Circle area is proportional to the number of data points at that coordinate.
418
Results are divided by location and host taxonomic level: a) Panama, species; b) Panama, genus; c)
419
Panama, family; d) Malaysia, species; e) Malaysia, genus; f) Malaysia, family. All fitted slopes
19
420
(dashed lines) were positive and all were statistically significant (P < 0.05), except in Malaysia by host
421
genus (P = 0.062).
20
422
Literature Cited
423
Ali, J. G., and A. A. Agrawal. 2012. Specialist versus generalist insect herbivores and plant defense.
424
Trends Plant Sci. 17:293–302.
425
Andersen, J. C., J. Wu, M. E. Gruwell, R. Gwiazdowski, S. E. Santana, N. M. Feliciano, G. E. Morse,
426
and B. B. Normark. 2010. A phylogenetic analysis of armored scale insects (Hemiptera: Diaspididae),
427
based upon nuclear, mitochondrial, and endosymbiont gene sequences. Mol. Phylogenet. Evol. 57:992–
428
1003.
429
Barrett, L. G., and M. Heil. 2012. Unifying concepts and mechanisms in the specificity of plant-enemy
430
interactions. Trends Plant Sci. 17:282–292.
431
Benjamini, Y., and Y. Hochberg. 1995. Controlling the false discovery rate: a practical and powerful
432
approach to multiple testing. J. R. Stat. Soc. Ser. B 57:289–300.
433
Bernays, E. A. 2001. Neural limitations in phytophagous insects: implications for diet breadth and
434
evolution of host affiliation. Annu. Rev. Entomol. 46:703–727.
435
Capella-Gutiérrez, S., J. M. Silla-Martínez, and T. Gabaldón. 2009. trimAl: a tool for automated
436
alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25:1972–1973.
437
Celorio-Mancera, M. de la P., C. W. Wheat, M. Huss, F. Vezzi, R. Neethiraj, J. Reimegård, S. Nylin,
438
and N. Janz. 2016. Evolutionary history of host use, rather than plant phylogeny, determines gene
439
expression in a generalist butterfly. BMC Evol. Biol. 16:59. BMC Evolutionary Biology.
440
Dixon, A. F. G. 2005. Insect herbivore-host dynamics: tree-dwelling aphids. Cambridge University
441
Press, Cambridge, 199pp.
21
442
Egan SP, Ragland GJ, Assour L, Powell TH, Hood GR, Emrich S, Nosil P, Feder JL. 2015.
443
Experimental evidence of genome wide impact of ecological selection during early stages of speciation
444
with gene flow. Eco. Let. 18(8): 817–825
445
Forister, M. L., L. A. Dyer, M. S. Singer, J. O. Stireman, and J. T. Lill. 2012. Revisiting the evolution of
446
ecological specialization, with emphasis on insect-plant interactions. Ecology 93:981–991.
447
Forister, M. L., and S. H. Jenkins. 2017. A neutral model for the evolution of diet breadth. Am. Nat.
448
190:E40–E54.
449
Forister, M. L., V. Novotny, A. K. Panorska, L. Baje, Y. Basset, P. T. Butterill, L. Cizek, P. D. Coley, F.
450
Dem, I. R. Diniz, P. Drozd, M. Fox, A. E. Glassmire, R. Hazen, J. Hrcek, J. P. Jahner, O. Kaman, T. J.
451
Kozubowski, T. A. Kursar, O. T. Lewis, J. Lill, R. J. Marquis, S. E. Miller, H. C. Morais, M. Murakami,
452
H. Nickel, N. A. Pardikes, R. E. Ricklefs, M. S. Singer, A. M. Smilanich, J. O. Stireman, S. Villamarín-
453
Cortez, S. Vodka, M. Volf, D. L. Wagner, T. Walla, G. D. Weiblen, and L. A. Dyer. 2015. The global
454
distribution of diet breadth in insect herbivores. Proc. Natl. Acad. Sci. U. S. A. 112:442–7.
455
Funk, D. J., D. J. Futuyma, G. Orti, and A. Meyer. 1995. A history of host associations and
456
evolutionary diversification for Ophraella (Coleoptera: Chrysomelidae): new evidence from
457
mitochondrial DNA. Evolution 49:1008–1017.
458
Futuyma, D. J. 2010. Evolutionary constraint and ecological consequences. Evolution 64:1865–84.
459
Futuyma, D. J. 2008. Sympatric speciation: norm or exception? Pp. 136–147 in Specialization,
460
Speciation, and Radiation: The Evolutionary Biology of Herbivorous Insects.
461
Futuyma, D. J., M. Keese, and D. J. Funk. 1995. Genetic constraints on macroevolution: the evolution
462
of host affiliation in the leaf beetle genus Ophraella. Evolution 49:797–809.
22
463
Futuyma, D. J., and G. Moreno. 1988. The evolution of ecological specialization. Annu. Rev. Ecol.
464
Syst. 19:207–233.
465
García Morales, M., B. D. Denno, D. R. Miller, G. L. Miller, Y. Ben-Dov, N. B. Hardy. 2016. ScaleNet:
466
a literature-based model of scale insect biology and systematics. Database, 2016.
467
Gilbert, G. S., and C. O. Webb. 2007. Phylogenetic signal in plant pathogen-host range. Proc. Natl.
468
Acad. Sci. U. S. A. 104:4979–83.
469
Gompert, Z., J. P. Jahner, C. F. Scholl, J. S. Wilson, L. K. Lucas, V. Soria-Carrasco, J. A. Fordyce, C. C.
470
Nice, C. A. Buerkle, and M. L. Forister. 2015. The evolution of novel host use is unlikely to be
471
constrained by trade-offs or a lack of genetic variation. Mol. Ecol. 24:2777–2793.
472
Gwiazdowski, R. A., I. M. Vea, J. C. Andersen, and B. B. Normark. 2011. Discovery of cryptic species
473
among North American pine-feeding Chionaspis scale insects (Hemiptera: Diaspididae). Biol. J. Linn.
474
Soc. 104:47–62.
475
Gyllenberg, M., and J. Metz. 2001. On fitness in structured metapopulations. J. Math. Biol. 43:545–60.
476
Hadfield, J. D., and S. Nakagawa. 2010. General quantitative genetic methods for comparative biology:
477
phylogenies, taxonomies and multi-trait models for continuous and categorical characters. J. Evol. Biol.
478
23:494–508.
479
Hardy, N. B. 2018. The Biodiversity of Sternorrhyncha: Scale insects, aphids, psyllids, and whiteflies.
480
Insect Biodiversity: Science and Society, pp.591–625.
481
Hardy, N. B., C. Kaczvinsky, G. Bird, B. B. Normark. 2020. What we don’t know about diet-breadth
482
evolution in herbivorous insects. Annual Review of Ecology, Evolution and Systematics 51:103–122.
483
https://doi.org/10.1146/annurev-ecolsys-011720-023322
23
484
Hardy, N. B., and S. P. Otto. 2014. Specialization and generalization in the diversification of
485
phytophagous insects: tests of the musical chairs and oscillation hypotheses. Proceedings of the Royal
486
Society B -- Biological Sciences 281: 20132960 http://dx.doi.org/10.1098/rspb.2013.2960.
487
Hardy, N. B., D. A. Peterson, and B. B. Normark. 2016. Nonadaptive radiation: pervasive diet
488
specialization by drift in scale insects? Evolution 70:2421–2428.
489
Heard, S. B., J. O. Stireman, J. D. Nason, G. H. Cox, C. R. Kolacz, and J. M. Brown. 2006. On the
490
elusiveness of enemy-free space: spatial, temporal, and host-plant-related variation in parasitoid attack
491
rates on three gallmakers of goldenrods. Oecologia, 150:421–434.
492
Hill, M. G., and T. Holmes. 2009. An analysis of latania scale (Hemiberlesia lataniae) crawler
493
settlement behaviour on kiwifruit leaves and bark. New Zeal. Plant Prot. 62:56–62.
494
Hogenhout, S. A., and J. I. B. Bos. 2011. Effector proteins that modulate plant-insect interactions. Curr.
495
Opin. Plant Biol. 14:422–428.
496
Hutchinson, G. E. 1959. Homage to Santa Rosalia or why are there so many kinds of animals? Am.
497
Nat. 93:145–159.
498
Janz, N., K. Nyblom, and S. Nylin. 2001. Evolutionary dynamics of host‐plant specialization: a case
499
study of the tribe Nymphalini. Evolution 55:783–796.
500
Joshi, A., and J. N. Thompson. 1995. Trade-offs and the evolution of host specialization. Evol. Ecol.
501
9:82–92.
502
Kelley, S. T., and B. D. Farrell. 1998. Is specialization a dead end? The phylogeny of host use in
503
Dendroctonus bark beetles (Scolytidae). Evolution 52:1731–1743.
24
504
Krasnov, B. R., M. A. Fortuna, D. Mouillot, I. S. Khokhlova, G. I. Shenbrot, and R. Poulin. 2012.
505
Phylogenetic signal in module composition and species connectivity in compartmentalized host-
506
parasite networks. Am. Nat. 179:501–511.
507
Lancaster, L. T, 2020. Host use diversification during range shifts shapes global variation in
508
Lepidopteran dietary breadth. Nature Ecology & Evolution 4:963–969.
509
Magsig-Castillo, J., J. G. Morse, G. P. Walker, J. L. Bi, P. F. Rugman-Jones, and R. Stouthamer. 2010.
510
Phoretic dispersal of armored scale crawlers (Hemiptera: Diaspididae). J. Econ. Entomol. 103:1172–
511
1179.
512
Mirarab, S., N. Nguyen, and T. Warnow. 2014. PASTA: ultra-large multiple sequence alignment. Res.
513
Comput. Mol. Biol. 177–191.
514
Neher, R. A. 2013. Genetic draft, selective interference, and population genetics of rapid adaptation.
515
Ann. Rev. Eco. Evo. Sys. 44:195–215.
516
Normark, B. B., G. E. Morse, A. Krewinski, and A. Okusu. 2014. Armored scale insects (Hemiptera:
517
Diaspididae) of San Lorenzo National Park, Panama, with descriptions of two new species. Ann.
518
Entomol. Soc. Am. 107:37–49.
519
Normark, B. B., and N. A. Johnson. 2011. Niche explosion. Genetica 139:551–564.
520
Nurmi, T. and K. Parvinen, 2008. On the evolution of specialization with a mechanistic underpinning in
521
structured metapopulations. Theor. Pop. Bio. 73:22–3243.
522
Peterson, D. A., N. B. Hardy, G. E. Morse, I. C. Stocks, A. Okusu, and B. B. Normark. 2015.
523
Phylogenetic analysis reveals positive correlations between adaptations to diverse hosts in a group of
524
pathogen-like herbivores. Evolution 69:2785–2792.
25
525
Peterson, D. A., N. B. Hardy, and B. B. Normark. 2016. Micro- and macroevolutionary trade-offs in
526
plant-feeding insects. Am. Nat. 188:640–650.
527
R Core Team. 2017. R: A language and environment for statistical computing. R Foundation for
528
Statistical Computing, Vienna, Austria.
529
Ravigné, V., U. Dieckmann, and I. Olivieri. 2009. Live where you thrive: joint evolution of habitat
530
choice and local adaptation facilitates specialization and promotes diversity. Am. Nat. 174:E141–E169.
531
Remold, S. 2012. Understanding specialism when the jack of all trades can be the master of all. Proc.
532
R. Soc. B Biol. Sci. 279:4861–4869.
533
Rodriguez-Verdugo, A., D. Carrillo-Cisneros, A. Gonzalez-Gonzalez, B. S. Gaut, and A. F. Bennett.
534
2014. Different tradeoffs result from alternate genetic adaptations to a common environment. Proc.
535
Natl. Acad. Sci. 111:12121–12126.
536
Singer, M. S., and J. O. Stireman. 2005. The tri-trophic niche concept and adaptive radiation of
537
phytophagous insects. Ecol. Lett. 8:1247–1255.
538
Smith, S. A., and B. C. O’Meara. 2012. treePL: divergence time estimation using penalized likelihood
539
for large phylogenies. Bioinformatics 28:2689–2690.
540
Stamatakis, A. 2014. RAxML version 8: a tool for phylogenetic analysis and post-analysis of large
541
phylogenies. Bioinformatics 30:1312–3.
542
Vea, I. M., and D. A. Grimaldi. 2016. Putting scales into evolutionary time: the divergence of major
543
scale insect lineages (Hemiptera) predates the radiation of modern angiosperm hosts. Sci. Rep.
544
6:23487. Nature Publishing Group.
545
26
546
27
547
Data Accessibility
548
DNA sequences will be deposited in GenBank, and trophic link data and analysis scripts will be
549
uploaded to Dryad upon acceptance.
550
551
Competing Interests
552
The authors are unaware of any conflicts of interest.
553
554
Author Contributions
555
BBN, GEM, and DAP designed the study. All authors except NBH obtained the data. DAP and NBH
556
analyzed the data and wrote the paper, with input from BBN, GEM, JW, and TI.
557
558
28
559
Table 1: Mean Simpson’s reciprocal diversity index (1/D) of individual host trees colonized by each
560
DNA-delimited diaspidid species for both sampling locations and all three host taxonomic levels.
Location
Taxon Level Empirical 1/D
Null 1/D
Z
P
Panama
Species
3.162
3.321
-1.449
0.147
Panama
Genus
3.008
3.295
-2.721
0.007
Panama
Family
2.671
2.983
-2.887
0.004
Malaysia
Species
1.643
2.087
-9.902
< 0.001
Malaysia
Malaysia
Genus
Family
1.461
1.461
1.955
1.785
-6.400
-3.472
< 0.001
< 0.001
561
Table 2: Statistical results from the models relating abundance per host and the proportion of host
562
taxon occupancy to the local diet breadth of each DNA-delimited diaspidid species.
Location
Taxon Level
Abundance
Occupancy proportion
Slope
Z
P
Slope
Z
P
Panama
Species
0.000
0.059
0.953
0.030
2.616
0.009
Panama
Genus
0.000
0.070
0.944
0.036
3.125
0.002
Panama
Family
0.001
0.074
0.941
0.052
2.908
0.004
Malaysia
Species
0.007
0.098
0.922
0.361
2.077
0.038
Malaysia
Malaysia
Genus
Family
0.006
0.006
0.115
0.111
0.909
0.912
0.455
0.765
1.867
5.381
0.062
< 0.001
29
Order an Essay Now & Get These Features For Free:
Turnitin Report
Formatting
Title Page
Citation
Outline
