Supplementary material for Salazar, Jaramillo and Marquis 2015.
Table S1: Generalized linear mixed effects models used for model selection.
Model A.1 / d.f. / AIC / p-value / ∆ AICTotal herbivory = chemical diversity + non-Piper diversity + phylogenetic diversity + (light) + (Piper Diversity) / 7 / 86.78 / 0.46 / 1.45
Total herbivory = chemical diversity + phylogenetic diversity + (light) + (Piper diversity) / 6 / 85.33 / 0.4 / 1.3
*Total herbivory = chemical diversity + (light) + (Piper diversity) / 5 / 84.03 / - / 0
Model A.2
*Generalist herbivory = chemical diversity + non-Piper diversity + phylogenetic diversity + (light) + (Piper diversity) / 7 / 45.34 / - / 0
Generalist herbivory = non-Piper diversity + phylogenetic diversity + (light) + (Piper diversity) / 6 / 45.84 / 0.03 / 0.5
Generalist herbivory = chemical diversity + non-Piper diversity + (light) + (Piper diversity) / 6 / 46.42 / 0.08 / 1.08
Model A.3
Specialist herbivory = chemical diversity + non-Piper diversity + phylogenetic diversity + (light) + (Piper diversity) / 7 / 79.9 / 0.9 / 3.75
Specialist herbivory = chemical diversity + non-Piper diversity + (light) + (Piper diversity) / 6 / 77.91 / 0.62 / 1.76
*Specialist herbivory = chemical diversity + (light) + (Piper diversity) / 5 / 76.15 / - / 0
Model B.1
Specialist herbivory = high-volatility chem. diversity + low-volatility chem. diversity + non-Piper diversity + phylogenetic diversity + (light) + (Piper diversity) / 8 / 85.3 / 0.84 / 1.97
Specialist herbivory = high-volatility chem. + non-Piper diversity + phylogenetic diversity + (light) + (Piper diversity) / 7 / 83.33 / 0.39 / 1.27
*Specialist herbivory = high-volatility chem. + phylogenetic diversity + (light) + (Piper diversity) / 6 / 82.06 / - / 0
Model B.2
Generalist herbivory = high-volatility chem. + low-volatility chem.+ non-Piper diversity + phylogenetic diversity + (light) + (Piper diversity) / 8 / 68.28 / 0.42 / 1.35
Generalist herbivory =high-volatility chem. + low-volatility chem. + phylogenetic diversity + (light) + (Piper diversity) / 7 / 66.93 / 0.44 / 1.41
Generalist herbivory = high-volatility chem. + low-volatility chem. + (light) + (Piper diversity) / 6 / 65.52 / 0.96 / 1.99
*Generalist herbivory = low-volatility chem.+ (light) + (Piper diversity) / 5 / 63.53 / - / 0
All models were selected using the optimization method described in the Appendix (see below). Piper diversity and non-Piper diversity were calculated using the Gini-Simpson index. Light was measured as the percentage of canopy openness. Herbivory, diversity, and light were logit function transformed to achieve normality. Δ AIC is the difference between a model and the “optimal model”. Models in bold are the “optimal” models.
Model C.1 / Estimate / df / F / p-valueTotal herbivory = Piper diversity + non-Piper diversity +Piper chemical diversity + (light)
Piper diversity* / -7.76 / 1 / -1.00 / 0.31
Non-Piper diversity
Piperchemical diversity / -1.07
-0.08 / 1
1 / -1.37
-2.27 / 0.17
0.02
Model C.2
Generalist herbivory = Piper diversity +non-Piper diversity + Piper chemical diversity + (light)
Piper diversity*
Non-Piper diversity
Piperchemical diversity / -2.08
-0.06
-0.07 / 1
1
1 / -0.36
-0.12
-2.90 / 0. 71
0.90
0.004
Model C.3
Specialist herbivory = Piperdiversity +non-Piper diversity +Piper chemical diversity + (light)
Piper diversity*
Non-Piper diversity
Piperchemical diversity / -5.51
-0.97
-0.1 / 1
1
1 / -0.94
-1.65
-0.41 / 0.34
0.10
0.67
Table S2: Additional generalized linear mixed effects models for Piper species diversity.
Table S2: Results from the generalized linear mixed model for models including Piper diversity as a fixed variable. Random variables are shown in parentheses.
Model Selection
We first built a “beyond optimal” model in which the fixed effects included the maximum number of independent explanatory variables based on biological reasonable hypotheses (hereafter, the full model). This full model was first tested for multicollinearity using the variable inflation factor (VIF). All VIF values were less than 2.4. Then we used a “top-down” approach to improve the model by systematically omitting one fixed effect variable at a time. Omitted variables were selected using the ANOVA function (the variable with the least explanatory power based on the data was dropped). The new model was then compared with the previous model (model1 vs. model1 - dropped variable) using AIC with maximum likelihood test; the model that had the smaller AIC but still P0.05 was selected. The optimal model was reached when all fixed effects variables achieved P < 0.05 on the ANOVA and AICs did not improve afterfurther removal of any fixed effects variables. All models used maximum likelihood estimation method. It is important to note that the combination of AIC and Maximum Likelihood is known to yield conservative models with a smaller number of fixed effects terms in the final model. Final models were evaluated via: (1) visually inspecting the fitted vs. residual plots for signs of bias and heteroscedasticity, (2) fitting a simple linear regression on the quantile-quantile plots, and (3) testing the normality of model residuals using the Shapiro-Wilk test. All models showed a good fit with normally distributed residuals. For all models we used a Gaussian distribution and the identity link function.Models were analyzed using R 2.15.2 (R Core Team 2012) and the nlme package (Pinheiro et al. 2013). See Table S1 for a complete list of the models and Table 2 for results of the mixed models.
Table S3: List of Piper species found in the study and their total abundances.
Piper species / N / 0.2 haP. asymmetricum / 28
P. augustum / 3
P. auritifolium / 77
P. biolleyi / 1
P. biseriatum / 4
P. cenocladum / 227
P. colonense / 93
P. concepcionis / 1
P. darienense / 1
P. decurrens / 11
P. dolichotrichum / 13
P. dryadum / 36
P. euryphyllum / 20
P. friedrichsthalii / 1
P. garagaranum / 91
P. glabrescens / 87
P. hispidum / 14
P. holdridgeanum / 108
P. imperiale / 56
P. melanocladum / 100
P. multiplinervium / 294
P. nudifolium / 33
P. peracuminatum / 10
P. pseudobumbratum / 49
P. reticulatum / 44
P. sancti-felicis / 14
P. schiedeanum / 21
P. silvivagum / 4
P. tonduzii / 1
P. trigonum / 346
P. urophyllum / 10
P. urostachyum / 240
Table S4: Non-exhaustive list of some putative compounds found in the chemical extracts of Piper species in this study. Identification is based on the primary literature and available mass spectra databases. This list should not be considered as definitive or exhaustive.
Compound(-)-Spathulenol
(E)-β-Farnesene
1,3-Benzodioxole, 5-(1-propenyl)-
1,3-Cyclohexadiene, 1-methyl-4-(1-methylethyl)-
1,3-Dimethyl-5-(propen-1-yl)adamantane (Guaiene)
1,5-Cyclodecadiene, 1,5-dimethyl-8-(1-methylethenyl
1,6-Cyclodecadiene, 1-methyl-5-methylene-8-(1-methylethyl)
12-Oxabicyclo[9.1.0]dodeca-3,7-diene, 1,5,5,8-tetramethyl
1H-3a,7-Methanoazulene, 2,3,6,7,8,8a-hexahydro-1,4,9,9-tetramethyl
1H-Cyclopenta[1,3]cyclopropa[1,2]benzene, octahydro-7-methyl-3-methylene-4-(1-methylethyl)
3-Carene, 4-isopropenyl-
3,5-Dihydroxy-4',7-dimethoxyflavone
3,7,11,15-Tetramethyl-2-hexadecen-1-ol
4-nerolidylcatechol
9-Methyltetracyclo[7.3.1.0(2.7).1(7.11)]tetradecane
Andrographolide
Apiole
Aromadendrene
Aspidinol
Benzene, 1,2-dimethoxy-4-(2-propenyl)-
Benzenecarboxylic acid
Bicyclo[2.2.1]heptane, 2-cyclopropylidene-1,7,7-trimethyl-
Cadinol
Calacorene
Carotene, 5,6-dihydro-5,6-dihydroxy-
Carotol
Caryophyllene
Caryophyllene oxide
Cenocladamide
Cerulignol
Chrysin
Cinnamamide, N-(p-hydroxyphenethyl)-
Cinnamic acid
Conanine
Copaene-8-ol
Cubebene
Cycloisolongifolene
Cyclopentanol, 1,2-dimethyl-3-(1-methylethenyl)
Desaspidinol
Dillapiole
Dimethoxyflavanone
Eicosane
Epi-bicyclosesquiphellandrene
Eucalyptol
Eugenol
Farnesene
Flemi chapparin
Galangin
Isoasarone
Isohomogenol
Isoledene
Isovanillin
Linalol
Myristicine
Naphthalene, 1,2,3,5,6,8a-hexahydro-4,7-dimethyl-1-(1-methylethyl)
Naringenin
Phellandrene
Phloroglucinol
Phytol
Pinocembrin
Piplartine
Pyrrolidine, 1-[5-(1,3-benzodioxol-5-yl)-1-oxo-2,4-pentadienyl]
Sabinene
Safrole
Selinene
Sitosterol
Sitosterol acetate
Solavetivone
Spathulenol
Spiro[4.4]non-1-ene (Thujene)
Squalene
Stigmasterol
Thymol
Tocopherol
trans-‡-Bergamotene
trans-Cinnamic acid
Vanillin
Veridiflorol
Vitamin E
Zingiberene
α-Humulene
α-Phellandrene
α-Pinene
α-Thujene
Supplementary figures:
Figure S1: Analysis of 3545 records of natural products found in Piper. Compounds with stars can be detected with GCMS. Records from NAPRALERT 2013 (Loub et al. 1985).
Figure S2: Local phylogeny for the 42 most common La Selva Piper species based on the ITS and the chloroplast intron psbJ-petA markers. Phylogenetic analysis was performed on MEGA using Maximum Likelihood (GTR-GAMMA model). Right-side labels indicate Piper subgenera (sections) sensu Jaramillo et al (2008).
Figure S3: Dendrograms for the high-volatility and low-volatility chemical similarity for the 27 Piper species found within the plots. Lines were added to link the same species in each dendrogram. The figure shows low congruence between the two chemical classes.
Figure S4: Total chemical similarity dendrogram for the 27 species of Piper found within the 81 plots. A non-exhaustive heat map has been added to help visualize the differences in chemical composition among the species. Colors show the secondary compound richness (number of compounds found) for different chemical groups. Darker colors signify higher number of compounds. Data are normalized across columns (chemical groups). Note that the dendrogram on the left is based on chemical similarity, and was calculated using the complete chemical data from the chemical analysis, not from the values of the heat map. The heat map in this figure is added as a visualization tool. For more details on the plant species secondary chemical composition see Supplementary material.
General results from the herbivory and chemical surveys
Piper diversity and herbivore damage
A total of 2035 individuals from 27 species of Piper were found across the 81 plots sampled for this study (see Table S3). The mean number of Piperindividuals present in a plot was 25.2 (SE; min-max = 1.1; 4-51), and the mean number of Piper species was 5.2 (SE; min-max= 1.4; 3-11). Levels of herbivory were relatively high (mean total herbivory: 20.13%; mean generalist herbivory: 9.97%, mean specialist herbivory: 11.15%). For non-Piper taxa, the average species richness within the plots was 17.9 (SE; min-max = 0.87; 9-28) and the average number of individuals was 54.6 (SE; min-max = 4.41; 29-128). Non-Piperdiversity within the plots was variable with an average of 0.88, a minimum of 0.86, and a maximum of 0.97 (values of the Gini-Simpson index range between 0 for low diversity to 1 for high diversity).
Piper chemical diversity
The GC-MS analysis yielded more than 1100 chromatographic features, of which approximately 40% were present in all Piper species (e.g., phytol, stigmasterol, sitosterol, tocopherol). Because these shared features were non-informative in the context of a chemical similarity analysis, and most likely related to plant primary chemistry, they were not used for the analysis. Among the remaining features we found a great diversity of terpenoids, phenylpropanoids, alkaloids, flavonoids, benzenoids, and some lignans (Table S4). The total, high-volatility, and low-volatility chemistry hierarchical clustering showed strong variation in chemical composition among species (Fig. S3 and S4). We found no congruence between the patterns of high-volatility and low-volatility chemical similarity among the sampled species (Fig. S3).
Proof of concept for the analysis of chemical similarity and variability
across Piper species via GC-MS
To assess the performance of our chromatographic technique and analytical approach for characterizing chemical diversity among Piper species, we performed a small “proof of concept” study(Salazar 2013). This study had two main goals. The first was to assess levels of intra- and interspecific variation among Piper species. If Piper species had less chemical variability within species than across species, this would suggest that our analytical approach is adequate to assess chemical similarity across species as well as at the community level.
Our second goal was to evaluate the ability of our analysis to assess chemical similarities between species without precise identification of each secondary compound. Specifically, we wished to test the efficacy of using mass spectral fragmentation patterns, molecular weights, and retention timesin combination to detect different chemical species across samples and plant species. This would allow us to assess chemical similarity between Piper species without the need to identify all compounds.
To achieve these goals we compared intraspecific chemical variation in the dominant leaf secondary metabolites via GC-MS for five sympatric Piper species that vary greatly in secondary chemistry and represent a gradient along the Piper chemical diversity spectrum. Additionally, three of these species are closely related and chemically similar. Furthermore, Piper species were compared among and between two distinctive geographical sites.
Material and methods:
Target species: For this study we worked with four pioneer Piper species: P. umbellatum (L.), P. peltatum (L.), P. auritum (Kunth), and P. aduncum (L.). Even though Piper species are mostly low light adapted, understory shrubs, these four species are common and abundant in secondary forests, forest gaps, and forest edges. Combined they exemplify the diverse secondary chemistry of the genus. Piper auritum is the species that shows the simplest secondary chemistry of the group in terms of the number of abundant secondary compounds. This species has a very low chemical diversity and a very dominant secondary metabolite: safrole(McBurnett et al. 2007; Monzote et al. 2010). Piper peltatum is slightly more chemically diverse thanP. auritum due to the presence of two abundant secondary metabolites. One of these metabolites is 4-nerolidylcatechol, a prenylated catechol that can be found in P. peltatum plants as well as in other members of the Pothomorphe clade (Kijjoa et al. 1980; Rezende & Barros 2004; Pinto et al. 2010; Lopes et al. 2013; Mendanha da Cunha et al. 2013). The other very abundant metabolite in P. peltatum is the sesquiterpene: germacrene D(Parmar et al. 1997; Cicció & Segnini 1998).Piper umbellatum (the sister species of P. peltatum) is highly diverse in sesquiterpenes. The major component of P. umbellatum chemical profiles is caryophyllene, a very common compound in other Piper species but that is found in high abundance in this taxon (Martins et al. 1998; Núñez et al. 2005; Tabopda et al. 2008; Cruz et al. 2012; Tabopda et al. 2012). In addition to caryophyllene, high abundances of humulen, spathunelol, and copaene are also found (Martins et al. 1998; Núñez et al. 2005; Tabopda et al. 2008; Cruz et al. 2012; Tabopda et al. 2012). The Piper with the highest chemical diversity of these pioneer species is P. aduncum. The most abundant secondary metabolite of this species is dillapiol, a propenylphenol that has been found in all sampled plants of P. aduncum(Cicció & Ballestero 1997; Vila et al. 2005; De Oliveira et al. 2006; De Almeida et al. 2009; Parise-Filho et al. 2011). Additionally, P. aduncum can have an array of abundant secondary metabolites that vary in identity from one population to the other (Cicció & Ballestero 1997; Parmar et al. 1997; Vila et al. 2005; De Oliveira et al. 2006; Bernardo et al. 2008; De Almeida et al. 2009). For our populations in Costa Rica the most abundant metabolites are -asarone, -cubebene, eudesmol, and an additional unknown compound (mw. 264) (Cicció & Ballestero 1997). Finally, we also sampled P. aequale, a species that had not been profiledchemically.
Collection Sites: Samples were collected in two sites within the Atlantic slope of Costa Rica. The first group of samples (lowland site) was collected in the vicinity of the northern limits of the Gandoca-Manzanillo National Wildlife Refuge, located at 73 km north of Limon (9°37’45” N, 82°40’06” W; Talamaca, Limon). This site is a lowland tropical rain forest with an average elevation of 25 m. The second group of samples (low elevation montane site) was collected in the vicinity of the eastern limits of Braulio Carrillo National Park (10°09’51” N, 83°53’45” W; Vasquez de Coronado, San José). This site is located in the transition between the lowland rain forest and cloud forest and has an average elevation of 550 m. Finally, to further test this approach, we added one sample of P. auritum from Chiapas, Mexico to assess the performance of your analytical technique with samples that are expected to be significant different than the ones collected in our main field locations.
Sample collection: For all five Piper species we collected 10 leaves from eachof 10 different individuals. Half of the samples were collected in the lowland site and half in the mountain base site. All leaves were young, but fully expanded. Additionally, all samples were selected to have similar herbivore damage (between 5 and 10%, damage was assessed visually). Samples were dried with silica gel and transported to the University of Missouri-St Louis for chemical analysis.
Sample Extraction: For each sample 0.4g of material was pulverized under liquid nitrogen. Samples were extracted using 1:1 methanol-chloroform solution. As an internalstandard, 0.1 mg of piperine was added to all samples. Samples were finally filtered and stored at -80°C until analysis.
Chemical analysis: Analysis of secondary metabolites was performed using GC-MS (HP 5890 coupled with a quadrapole Model 5988A mass detector) with helium as a carrier gas and a HP-5 capillary column (30 m, 0.32mm ID, 0.25 μm). The mass spectra of the different compounds in the samples were compared with NIST and MassBank Databases as well as primary literature. Chromatograms of each sample were integrated and the area of the peaks of the all detected metabolites in each species was calculated. Chromatograms were then aligned using COTW (Correlation Optimized Time Warping).
Statistical analysis: For each species of Piper we constructed a library of all chemical compound mass spectra. This library was used to identify similar compounds present across all studied Piper samples. All library work was done using AMDIS. Using the results of the former analysis we constructed a similarity matrix and performed a two-way hierarchical clustering (Ward algorithm).
Results: Our analysis showed that the intraspecific chemical variability is much smaller than the interspecific chemical variability in the sampled species. Furthermore, all samples from each species formed distinctive and discrete clusters suggesting that our analytical approach is adequate to assess chemical similarity given Piper’s intra- and interspecific variation. Furthermore, at least for these five species, the pairs of most closely related species showed a higher degree of chemical similarity than pairs of more distantly related species (Fig. POC 1). The two “low chemical diversity” species (P. auritum and P. peltatum) showed the smallest cluster distance. Also, all species from the Pothomorphe clade (P. auritum, P. peltatum, and P. umbellatum) grouped in a single cluster. Finally, P. aduncum, the most chemically diverse of all sampled species, showed the most distantfrom the “low-chemical diversity” P. auritum. All of these results were consistent even in the presence of samples from distinctive geographical locations. It is important to note that in this proof of concept we used quantitative (peak area) as well as qualitative (chemical identity) data and that almost all of the intraspecific variation observed in the clustering analysis is due to variation in the concentration on secondary metabolites rather than the composition of the species chemical profiles.
Figure POC 1: Result of the hierarchical clustering analysis (Ward’s algorithm) based on the chemical similarity between fivePiper species across two geographical locations. Metabolites across species and samples were linked using a combination of mass spectra fragmentation pattern, molecular weight, and retention time under identical chromatographic conditions. All samples were collected in Costa Rica with the exception of au_10 (in blue) that was collected in Mexico (Los Tuxtlas).