Phytogeographic and genetic variation in Sorbus, a traditional antidiabetic medicine—adaptation in action in both a plant and a discipline

Field collection

Samples were collected in August 2008 in northern Quebec (Canada) within the James Bay territory, which is independently administered by the Cree Nation Government and the local Band councils of each community. Twenty populations were sampled in the areas surrounding the villages of Waskaganish (51°29^′N, 78°45^′W), Chisasibi (53°47^′N, 78°53^′W) and Whapmagoostui (55°16^′N, 77°45^′W) on the coast of James Bay and Hudson Bay, and Nemaska (51°41^′N, 75°15^′W), Mistissini (50°25^′N, 73°51^′W) located inland. All the villages, except Whapmagoostui, are in the boreal forest ecozone, with spruce and other coniferous trees dominating. Whapmagoostui is located in the hemiarctic, where the tundra ecozone begins. The field permit was covered within a Research Agreement among universities, the Cree Board of Health and Social Services of James Bay, and the participating communities (see http://www.taam-emaad.umontreal.ca/about%20us/CIHR-TAAM_Final_Research_Agreement_signed_091030.pdf).

For each of twenty populations, bark and leaves samples were collected for phytochemical extraction from two, and up to twenty, trees respectively. These were dried either in a plant press or in paper bags (approximately 10 g of dry material). Effort was taken to collect both S. americana and S. decora individuals in equal number, but S. americana was found to be less common or absent from Whapmagoostui. At each site, GPS location was recorded and several pictures of the surroundings were taken. All samples were collected between 10 AM and 2 PM. At each site, voucher specimens of both S. decora and S. americana were collected in duplicate for depositing at the Marie-Victorin Herbarium (MT) at the Montreal Botanical Garden and at the McGill Herbarium (MTMG) on the Macdonald Campus; identities were confirmed by S Bailleul, S Hay and A Cuerrier at the Montreal Botanical Garden.

Samples were also collected for RNA extraction from both leaf and bark tissues. Approximately 200 mg of bark and leaf samples were immediately immersed in 1 ml of RNAlater^® (Qiagen, Venlo, Netherlands). This approach had been previously tested using samples collected from the Montreal Botanical Garden, and found to successfully maintain RNA integrity. Once collected, samples were placed in a −20 °C freezer for transfer between villages and back to the laboratory.

RNA extraction, selection of candidate genes and primers

Two enzymes, flavonol synthase (FLS) and squalene synthase (SquaS), were selected given that both are involved in the biosynthesis of precursors of important secondary metabolites in Sorbus (Fischer et al., 2007) and possess well known antioxidants properties. Flavonol synthase is a key enzyme in the antioxidant flavonoid pathway (Hukkanen et al., 2006; Fischer et al., 2007) and the gene expression of FLS has been demonstrated to control enzyme activity in many plants species (Xu et al., 2012). For example, in Vitis vinifera, gene expression of FLS increased during ripening of the fruit, coincident with the increase in flavonols per berry (Downey, Harvey & Robinson, 2003). Squalene synthase (SquaS) is an enzyme expressed early in the biosynthetic chain for terpenes, a large group of volatile unsaturated hydrocarbons found in the essential oils of plants and well known for their antioxidant capacities (Gonzalez-Burgos & Gomez-Serranillos, 2012). In addition, the positive correlation between SquaS gene expression, protein levels and enzymatic activity has been clearly demonstrated in Nicotiana tabacum(Devarenne, Ghosh & Chappell, 2002).

As no published sequences for FLS and SquaS from Sorbus were available, primers were designed based on an Expressed Sequence Tag (EST) sequence from Malus (GenBank Accession #AB331947 and #DQ849001, see Table S1). Actin gene expression (GenBank Accession served as a control for real time quantitative PCR, as commonly performed (Heid et al., 1996). Primers were designed based on Sorbus aucuparia Linnaeus sequences. Primers for all genes were designed in Primer3 (Rozen & Skaletsky, 2000) and synthesized by Alpha DNA (Montreal, Quebec). Primers were first tested by amplifying genomic DNA samples, which were then used to create purified PCR standards for all three genes (PCR Purification Kit, Qiagen). PCR products for the two candidate genes were sequenced by Genome Quebec and consensus sequences compared (BLASTn) against Expressed Sequences Tags (ESTs) database in GenBank (http://blast.ncbi.nlm.nih.gov/Blast.cgi, performed in July 2008). Both genes had hits against Malus ESTs (FLS S. decora, GenBank accession number GQ423752, FLS S. americana GenBank accession number GQ423751, SquaS S. decora GenBank accession number GQ423753, SquaS S. americana GenBank accession number GQ423754).

RNA was extracted following the protocol outlined in the Qiagen RNEasy Mini Plant Kit manual. The samples were extracted following the standard RNEasy protocol, with a standard (200 µl) amount of ethanol added to each sample, instead of approximately half the volume as recommended. At the end of the RNA extraction protocol, concentration and quality were verified using a NanoDrop spectrophotometer (Thermo Scientific, Waltham, Ma, USA). Ten samples out of the 78 collected had a 260/280 ratio below 1.6, or a low yield of total RNA, were not processed any further. Using the Wipeout Qiagen Quantitect reverse transcriptase protocol, cDNA was synthesized, with a standard amount of 40 ng of RNA as starting material.

Quantitative RT-PCR

RT-PCR was performed on a Stratagene (Agilent Technologies, Santa Clara, CA, USA) MX3000P real-time PCR system with a Brilliant SYBR Green rtPCR Master Mix kit. A PCR performed against actin provided a base value, which was then compared to the purified genomic PCR products of each gene, and then diluted to a concentration between 10⁻⁹ and 10⁻¹² of the original product (QiaQuick, Qiagen). Products were run in strip tubes of eight, with standards in every reaction, and then samples were randomized in rows of gene of interest, followed by the control gene (actin). It took three sets of samples to complete one repeat of each gene, and all sets were run three times. Amplification conditions were as follows for FLS: denaturing step (10 min at 95 °C), followed by 45 cycles of 95 °C denaturation (30 s), 55 °C annealing (1 min), 72 °C extension (30 s), completed by a final cycle of 1 min at 95 °C, 30 s at 55 °C and 30 s at 95 °C. For squalene synthase, amplification conditions were as follows: denaturing step (10 min at 95 °C), followed by 50 cycles of 95 °C denaturation (30 s), 53 °C annealing (1 min), 72 °C extension (30 s), completed by a final cycle of 1 min at 95 °C, 30 s at 53 °C and 30 s at 95 °C.

The slopes of product accumulation for the genes of interest against the standards were checked to be between −3.5 and −4.5, along with the efficiency of the reaction being between 90 and 110%, and a singular dissociation curve. Final copy value numbers, based on the standard curve of each reaction, the C(t) of each sample and the efficiency of the reaction were exported to a spreadsheet. Copy values for each sample were normalized for each gene of interest against actin. Each sample was analysed three times, and outliers were identified using the Grubbs test (Grubbs, 1950) and then removed using Prism GraphPad (San Diego, CA, USA). Gene expression data are available as Table S2.

We analysed gene expression differences separately for flavonol and squalene synthase using linear mixed effect models in R (R Core Team, 2013). First, expression ratios (FLS/actin or SquaS/actin expression) were log transformed. Several linear mixed effect models (package lme4, Bates et al., 2015) with a combination of fixed effects (species, latitude, longitude, tissue) and random effects (individuals, replicates) were tested. The model with the lowest Akaike information criterion (AIC, Akaike, 1974) was retained (see Table S3 for list of models tested; Burnham & Anderson, 2004). F statistics and p-values based on Satterthwaite’s approximations were calculated with LmerTest (Kuznetsova, Christensen & Brockhoff, 2013) in R. Post hoc Tukey t-tests were also performed in R.

Extract preparation

A total of 24 bark samples based on sample availability, from all five communities and four latitudes, were extracted and analysed for their antioxidant activity and phenolic content. For extraction, approximately 5 g of dried material was ground using a blender (40 mesh), placed in an Erlenmeyer flask with 100 ml of 80% ethanol, sealed and swirled for 24 h at approximately 150 rpm. The filtrate was separated from residue and the pellet was extracted twice. Pooled filtrate (200 ml) was dried using a rotary evaporator (water bath temperature of 40 °C), freeze dried in Supermoduylo freeze drier (Thermo Fisher Inc.) and stored at –20 °C.

Antioxidant analysis

Stock concentrations made from the crude bark extracts were analyzed for antioxidant activity using both oxygen radical absorbance capacity (ORAC) and 2,2-Diphenyl-1-Picrylhydrazyl (DPPH) assays.

ORAC assays were based on a protocol designed by Cao & Prior (1999) and adapted by Ou, Hampsch-Woodill & Prior (2001). It uses fluorescein reacting with AAPH, a peroxyl radical, to measure an extract’s ability to protect against oxidation compared to Trolox, a known antioxidant analogous to vitamin E and serving as positive control. Several modifications were made to the Ou, Hampsch-Woodill & Prior (2001) protocol. Each extract (0.005 g) was dissolved in 25 ml of 50% methanol and potassium phosphate buffer (PBS, pH 7.0), while Trolox, AAPH and fluorescein, prepared fresh for each reaction, were diluted in 100% PBS. Trolox was diluted to four concentrations, 0.78 µM, 0.39 µM, 0.195 µM and 0.0975 µM. Extracts were diluted to 6.5 µg/ml, 3.25 µg/ml, 1.6125 µg/ml and 0.806125 µg/ ml. Each reaction took place at room temperature and included 150 µl of fluorescein and 25 µl of each concentration of Trolox or sample; 25 µl of AAPH was added to each well at the start of the reaction. A fluorescence reading was taken as soon as AAPH was added to the 96-wells plate; readings were then taken every 10 min for 90 min Each concentration of sample or Trolox was conducted in duplicate for every reaction, with the reaction being repeated three times. Areas under the curve for each concentration were calculated in a spreadsheet with the blank subtracted from each sample area under the curve (AUC). Final values were determined using a regression equation based on Trolox curves, expressed as micromoles Trolox equivalents per gram.

DPPH assays were conducted based on a protocol designed by Blois (1958) with modifications outlined in McCune & Johns (2002), Owen & Johns (2002) and Fraser et al. (2007). The free radical scavenging activities of extracts when exposed to DPPH (1,1-diphenyl-2-picrylhydrazyl; Sigma-Aldrich, Oshawa, Ontario, CAN) for 10 min were measured using a Wallac Victor² 1420 Multilabel counter (PerkinElmer, Waltham, Ma, USA), with an absorbance of 490 nm. An IC₅₀ (inhibition concentration based on concentration of ascorbic acid required to quench 50% of radicals in the reaction) was calculated for each sample based on the linear portion of the dose–response curve for samples. The amount quenched by the standards for comparison to samples was calculated using both the linear and plateau portions of the line, using the equation: ${\displaystyle \hat{y}-y_0=\frac{\sum \left\{\left(x-x_0\right)\left(y-y_0\right)\right\}}{\sum {\left\{x-y_0\right\}}^2}}$

A smaller value thus denotes a higher free radical scavenging ability. Reactions were performed three times, with every sample in duplicate within each reaction. Catechin, epicatechin and quercetin served as controls along with ascorbic acid. Statistical analysis (ANOVAs and post hoc Tukey t-tests) for both ORAC and DPPH data sets was performed in SPSS (v 16.0).

Finally, total soluble phenolic content of each bark sample (n = 72) was determined using the Folin-Ciocalteu method as described by Fraser et al. (2007) and significance tested with non-parametric Kruskal–Wallis test and post hoc Tukey t-tests.

RNA analysis

Of the 78 collected samples, 59 were successfully extracted and transcribed into high quality cDNA, suitable for gene expression analyses. It is likely that the high phenolic content of Sorbus made it difficult to use large amounts of RNA suitable for reverse transcription (Kiefer, Heller & Ernst, 2000). We observed that flavonol synthase gene expression was consistently higher than squalene synthase (t-test, t = 11.1, p-value = 3.14e−24, Fig. 1).

For both flavonol and squalene synthase, a model containing latitude as a fixed factor was chosen (lowest AIC value) and expression differed significantly according to latitude (F statistic = 3.48, p-value = 0.029 and F statistic = 4.02, p-value = 0.015, respectively). In both cases, an increase in expression according to latitude was observed (Figs. 2A and 2B) and post hoc Tukey tests indicated that at latitude 55, expression was significantly higher. For both squalene and flavonol synthase, we also observed that gene expression in coastal samples was higher than those from inland locales in the full mixed effects model that contained all variables including longitude, but this effect was non-significant (Table S3, squalene synthase, F statistic = 0.34, p-value = 0.56; flavonol synthase, F statistic = 0.36, p-value = 0.55). Similarly, we observed non-significant differences in the full model between gene expression in leaf and bark samples for both squalene synthase (Table S3, F statistic = 0.28, p-value = 0.60) and flavonol synthase (Table S3, F statistic = 1.07, p-value = 0.30).

Antioxidant analysis

Both ORAC and DPPH analyses showed similar trends in antioxidant capacities of samples pooled according to location (r² = 0.92, p-value = 0.04), although ANOVAs did not reveal significant differences (p-value > 0.05; Figs. 3A and 3B). Both analyses showed increasing antioxidant capacity with latitude (note that for DPPH, lower values imply greater antioxidant activity, Fig. 3A) similar to the trend observed with levels of gene expression for FLS (r²_{(ORAC−FLS expression)} = 0.90, p-value = 0.05) and SquaS (r²_{(ORAC−SquaS expression)} = 0.84, p-value = 0.08). Sorbus decora samples tended to show higher capacity than S. americana (Figs. 4A and 4B). It was also observed that samples collected from coastal locations showed a higher capacity compared to inland populations, non-significantly in the ORAC assay but significantly in DPPH analysis (two-tailed t-test, p-value = 0.03, Figs. 5A and 5B).

Using the Folin-Ciocalteu method of phenolics measurement, we observed that soluble chemical content increased significantly with latitude (Kruskal–Wallis test, p-value ≤ 0.05, Fig. 6), along with significantly greater phenolic content in coastal samples (costal = 142.2, inland = 109.2, p-value = 0.02) and S. decora compared to S. americana, but not significantly (132.4 vs. 124.5 respectively, two-tailed t-test, p-value = 0.6).