Climate change alters leaf anatomy, but has no effects on volatile emissions from arctic plants

Biogenic volatile organic compound (BVOC) emissions are expected to change substantially because of the rapid advancement of climate change in the Arctic. BVOC emission changes can feed back both positively and negatively on climate warming.We investigated the effects of elevated temperature and shading on BVOC emissions from arctic plant species Empetrum hermaphroditum , Cassiope tetragona , Betula nana and Salix arctica . Measurements were per-formed in situ in long-term ﬁeld experiments in subarctic and high Arctic using a dynamic enclosure system and collection of BVOCs into adsorbent cartridges analysed by gas chromatography-mass spectrometry. In order to assess whether the treatments had resulted in anatomical adapta-tions, we additionally examined leaf anatomy using light microscopy and scanning electron microscopy. Against expectations based on the known temperature and light-dependency of BVOC emissions, the emissions were barely affected by the treatments. In contrast, leaf anatomy of the studied plants was signiﬁcantly altered in response to the treatments, and these responses appear to differ from species found at lower latitudes. We suggest that leaf anatomical acclimation may partially explain the lacking treatment effects on BVOC emissions at plant shoot-level. However, more studies are needed to unravel why BVOC emission responses in arctic plants differ from temperate species.


INTRODUCTION
Over the last three decades, the Arctic has had a warming trend of about 1°C per decade, which is significantly greater than the global mean (IPCC 2013). Also in the future, the Arctic region is projected to experience pronounced effects of the rapidly advancing climate warming (IPCC 2013), which impose substantial changes in the environment. As emissions of biogenic volatile organic compounds (BVOCs) from plants are highly temperature sensitive (Laothawornkitkul et al. 2009;Grote et al. 2013), drastically increased emissions from arctic plants are expected under the projected arctic warming of 3-11°C by 2100 (IPCC 2013).
Alterations in BVOC emissions are of great interest as they contribute to complex climate feedback mechanisms via both positive and negative indirect effects on climate warming (Arneth et al. 2010;Kulmala et al. 2013). They prolong the atmospheric lifetime of methane (CH4) and function as precursors of secondary organic aerosols (SOAs) possibly affecting cloud cover (Laothawornkitkul et al. 2009;Kulmala et al. 2013). Locally, BVOCs released by vegetation are involved in plant defence, growth, development and communication (Laothawornkitkul et al. 2009). Furthermore, BVOCs, in particular isoprene, are known for their protection against abiotic stresses (Vickers et al. 2009;Loreto & Schnitzler 2010;Possell & Loreto 2013). It is urgent to investigate how BVOC emissions from arctic plants respond to the predicted changes in climate.
Even high arctic heaths, dominated by dwarf shrubs or graminoids, have shown to be a source of various BVOCs and to have emissions that cannot be ignored (Schollert et al. 2014). However, relative to the global BVOC emissions, the emissions are likely to be low due to the scarce emitting plant biomass and short growing seasons . Besides being subjected to warming, which is likely to considerably increase the emissions (Tiiva et al. 2008;Faubert et al. 2010), the Arctic will also experience changes in precipitation, evaporation patterns and cloud cover (IPCC 2013). Increased cloudiness at high latitudes would potentially reduce the solar radiation reaching the Earth's surface general positive short-term temperature response is due to the accelerated chemical reaction rates, higher volatility of stored compounds and increased cellular diffusion rates (Laothawornkitkul et al. 2009).
BVOC emission rates are largely driven by the temperature and light levels prevailing during the measurements, although the temperature of a preceding time period might also influence the emission rates as observed for isoprene (Potosnak et al. 2013). Methods indicative of BVOC synthesis rates, for example enzyme availability, gene expression or substrate availability could not be used in the remote locations because of logistical constraints. Instead, we assessed long-term acclimation to elevated temperature and shading by analysis of leaf anatomy by light and scanning electron microscopy, because leaf function and structure are interconnected.
Warming has been shown to induce anatomical changes in both deciduous and conifer trees, such as thinning of leaves/ needles and leaf tissues, and larger leaf size (Higuchi et al. 1999;Luomala et al. 2005;Hartikainen et al. 2009Hartikainen et al. , 2014. Hartikainen et al. (2009) suggested that warming-induced thinner leaves, with shorter diffusion pathways for BVOCs, could be linked to increased BVOC emissions from Populus tremula under warming. Recently, Rasulov et al. (2014) showed that leaf structural modifications were the primary reasons for temperature acclimation of photosynthesis and isoprene emission of hybrid aspen (P. tremula × Populus tremuloides). Shade-adapted leaves are also usually thinner than sun-adapted leaves (Larcher 2003) and have a lower net photosynthesis and thus a potentially reduced synthesis of BVOCs (Loreto & Schnitzler 2010;Li & Sharkey 2013). Both warming (Hartikainen et al. 2014) and shading (Marques et al. 1999) have been shown to reduce the density of trichomes, which are specialized storage structures for BVOCs (Loreto & Schnitzler 2010;Fineschi et al. 2013). Climate change factors can also alter stomatal density (Beerling & Chaloner 1993;Luomala et al. 2005), which can exert at least short-term control over BVOC emissions (see Niinemets & Reichstein 2003). It is largely unknown how the leaf anatomy of arctic plants responds to climate change.
We examined how elevated temperature and shading, applied in long-term field experiments in a subarctic and a high Arctic location, affect BVOC emission and leaf anatomy of the most dominant vascular plant species. The present study was conducted after 23 or 8 years of warming and shading treatments in the subarctic and the high Arctic location, respectively. We hypothesized that experimental warming would clearly increase the emissions from the individual plant shoots in a similar manner as observed at ecosystem level (Tiiva et al. 2008;Faubert et al. 2010), and that the leaves would get thinner in response to warming as commonly observed for trees (Higuchi et al. 1999;Luomala et al. 2005;Hartikainen et al. 2009Hartikainen et al. , 2014. Shading was expected to decrease the emissions of the light-dependent BVOCs (Laothawornkitkul et al. 2009;Monson & Baldocchi 2014a) and lead to thinner leaves and reduced density of trichomes and stomata (Marques et al. 1999;Larcher 2003).

Study site and experimental design
The study was conducted in the growing season of 2012 using three long-term field experimental set-ups with warming and shading treatments. One experiment was located in a heath in Abisko, Northern Sweden (68°21′ N, 18°49′ E), where the growing season usually extends from June to September. The climate is subarctic with a mean annual temperature and precipitation of 0.07°C and 308 mm , respectively.The other two experiments were situated in a dry heath in Zackenberg, Northeast Greenland (71°30′ N, 20°30′ W). The climate is high Arctic with mean annual temperature of −7. 9°C (2003-2005), an average annual precipitation of 261 mm (1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005), and a growing season length of 2.0-3.5 months depending on year (Ellebjerg et al. 2008;Hansen et al. 2008). Air temperature and incoming photosynthetic photon flux density (PPFD over the waveband 400-700 nm) data for June to August 2012 at the two locations are presented in Fig. 1.
In all experiments, warming was achieved using domeshaped open-top polyethylene film tents increasing air temperature by approximately 1°C in Zackenberg (Ellebjerg et al. 2008) and 3-4°C in Abisko (Havström et al. 1993). The shading was provided by dome-shaped hessian tents reducing the incoming PPFD by 50-60% (Ellebjerg et al. 2008; see  Table 1 for temperature and PPFD during the present study). Tents were set up each year for the snow-free period. Control plots were not covered with tents (Supporting Information Fig. S1).
In Abisko, the experiment has been maintained since 1989 (Havström et al. 1993). The treatments were arranged in six replicate blocks with control, warming and shading making up a total of 18 plots. In each of the two experiments in Zackenberg, established in 2004, the same treatments were replicated in five blocks (15 plots in each experiment). These experiments in Zackenberg only differed regarding vegetation cover: one was located in Cassiope tetragona (L.) D. Don-dominated heath and the other in Salix arctica Pall.dominated heath.
Individual shoots of the dominant plant species were selected for the study. Plant shoots of each of the evergreen Empetrum nigrum ssp. hermaphroditum Hagerup and C. tetragona and the deciduous Betula nana L. were chosen in the Abisko experiment, whereas shoots of C. tetragona and the deciduous S. arctica were selected in the Zackenberg experiments (Supporting Information Fig. S2). Two measurement campaigns were conducted on July 3-4 and August 4-7 in Abisko. In Zackenberg, three measurement campaigns were done on July 17, July 31 and August 14 for C. tetragona and four were done on July 16, July 30, August 13 and August 29 for S. arctica.

Sampling and analysis of BVOC emissions
BVOC emissions from shoots of individual plant species were measured in situ by a dynamic enclosure system using polyethylene terephthalate (PET) bags (Rul-let Quality, Abena, Aabenraa, Denmark) as enclosures (Stewart-Jones & Poppy 2006). Flexible, disposable PET bags were convenient for collecting BVOCs from a large number of miniaturesized plants in remote field sites. The PET bags were precleaned (120°C, 60 min). Air was circulated through the PET bag of 1 L for 30 min by battery-operated pumps (12 V; Rietschle Thomas, Puchheim, Germany). The opening of the PET bag was tied around the stem together with the air inlet, while the air outlet was led through a hole in the bag. The incoming air with a flow of 500 mL min −1 was purified by a charcoal filter to remove particles and volatile contaminants present in the ambient air and by a MnO2 scrubber to remove ozone (Ortega & Helmig 2008). BVOCs from the 6 L air sample, drawn through by 200 mL min −1 , were adsorbed in stainless steel cartridges (Markes International, Ltd., Llantrisant, UK) containing Tenax TA (150 mg) and Carbopack 1 TD (200 mg). The flow rates were calibrated using mini BUCK Calibrator M-5 and regulated using mass flow sensors (D6F-P0010A1, Omron, Kyoto, Japan). The cartridges were sealed with Teflon-coated brass caps until analysis. During the sampling, temperature and relative humidity inside the enclosure (DS1923, iButton Hygrochron temperature/humidity logger, Maxim integrated, San Jose, CA, USA) and PPFD (photosynthetic light smart sensor S-LIA-M003, connected to HOBO micro station data logger H21-002, Onset Computer Corporation, Boston, MA, USA) were recorded every minute and every tenth second, respectively. The enclosures led to a temperature increase of 1.7 ± 1.7°C (P = 0.3, t-test) averaged across the campaigns in Abisko. After all BVOC measurement campaigns, the plant shoots were harvested, air dried, and with the exception of C. tetragona, sorted into leaves and stems to measure dry mass and determine leaf area.
BVOCs were analysed by thermal desorber (UNITY2, Markes International, Ltd.) coupled to an ULTRA autosampler and gas chromatograph-mass spectrometer (7890A Series GC coupled with a 5975C inert MSD/DS Perfomance Turbo EI System, Agilent Technologies, Inc., Santa Clara, CA, USA). BVOCs were identified according to 26 standard compounds and by using the mass spectra in the NIST library. Subsequently, pure standards (Fluka, Buchs, Switzerland) in methanol were used for quantification in a following manner: isoprene for isoprene, α-pinene for monoterpenes (MTs), 1.8-cineole for oxygenated MTs (oMTs), α-humulene for sesquiterpenes (SQTs) and (E)-2hexenal for other compounds. The standard solutions were injected into a stream of helium, which was directed through the cartridges and analysed as samples. The BVOCs not belonging to MTs, oMTs or SQTs were classified into other reactive volatile organic compounds (ORVOCs, having atmospheric lifetimes of <24 h, Guenther et al. 1995) or other volatile organic compounds (OVOCs).
Chromatograms were analysed using Enhanced ChemStation software (MSD ChemStation, E.02.01.1177, Agilent Technologies, Inc.) and information was extracted and sorted using an in-house function. Compounds that had an identification quality of the match with the NIST library above 90% were included in the dataset. Compounds originating from measurement or the analysis system were excluded. The excluded compounds were identified based on collection of volatile samples from empty PET bags.
The emission rate (E) was calculated according to: E = [(Cout − Cin) × F]/mdry, where Cout and Cin are the concentrations of BVOCs in the outlet and inlet air, F is the flow rate into the enclosure and mdry is the leaf dry mass. Emissions per leaf area were additionally calculated for B. nana and S. arctica, for which, it was possible to accurately determine the leaf area. The concentration in the filtered inlet air was assumed to be zero (see Ortega & Helmig 2008;Niinemets et al. 2011).
Emissions of isoprene, MTs, oMTs and SQTs were standardized to 30°C  and for isoprene to a PPFD of 1000 μmol m −2 s −1 using the algorithms presented by Guenther et al. (1993). This was done in order to estimate the emission potentials by excluding the momentary temperature effect during sampling and diurnal and daily variation. A β coefficient of 0.09°C −1 was used for monoterpenes (Guenther et al. 1993) and 0.18°C −1 for SQTs (Rinnan et al. 2011).

Leaf anatomy
Leaves were sampled when they were fully developed, but before the start of senescence. This was on July 6 in Abisko and on August 15 for C. tetragona and August 30 for S. arctica in Zackenberg. For the evergreen E. hermaphroditum, leaves from the previous growing season were used as the current year's leaves were not fully developed at the sampling time.
All sampled leaves had developed during the long-term manipulations. The leaves of E. hermaphroditum are needlelike with margins curled downwards creating a cavity in the middle, where the stomata are found (Fig. 2a,c). On C. tetragona leaves, the margins are curled upwards creating an open cavity in the middle of the leaves making the leaves look deeply grooved (Fig. 2b,d). S. arctica and B. nana are both broadleaved with less complex leaves with adaxial (upper) palisade parenchyma layer and abaxial (lower) spongy parenchyma layer (Fig. 3).
Two leaves per species per plot were collected for light microscopy (LM) and immediately placed in 2.5% (v/v) glutaraldehyde fixative (in 0.1 M sodium cacodylate buffer). While in glutaraldehyde fixative, sections of about 1.5 × 1.5 mm were cut along the central midrib of B. nana and S. arctica leaves whereas 1.5 mm segments were cut from the middle part of C. tetragona and E. hermaphroditum leaves. Samples were stored in cold fixative until further processing.
Samples were rinsed in sodium cacodylate buffer (0.1 M), post-fixated with 1% osmium tetroxide (OsO4) (in 0.1 M sodium cacodylate buffer), dehydrated with an ethanol series followed by a propylene oxide treatment, embedded in epon [Electron Microscopy Sciences, Hatfield, Pennsylvania, USA, Embed 812 (Zackenberg samples) and Ladd Research Industries, Williston, Vermont, USA, LX-112 (Abisko samples)]. 1.5 μm sections for LM were cut from the embedded samples using an ultramicrotome [Reichert-Jung Ultracut E, Wien, Austria, Diatomen histo-knife (Hi 4967)] and stained with toluidine blue and p-phenylene diamine (Kivimäenpää et al. 2010). LM sections were studied with a light microscope (Zeiss Primo Star light microscope, Jena, Germany and Olympus BX61 Automated upright microscope, Olympus Corporation, Tokyo, Japan) and photographed (Zeiss Axiocam ERc 5s camera, Jena, Germany and Olympus DP70 digital microscope camera, Tokyo, Japan). The objective magnification varied for the species. Digital images were analysed for thicknesses of the leaf, epidermis (including cuticle), and palisade and spongy parenchyma (Figs 2 & 3; Supporting Information Tables S 1 & S2). A shape change index was measured for E. hermaphroditum by dividing leaf cross section width by thickness (Fig. 2c). For C. tetragona and E. hermaphroditum the LM images were also used to calculate the numbers of trichomes and stomata per inner leaf surface length ( Fig. 2c-f). In order to calculate the proportion of palisade and spongy parenchyma cells and intercellular space of the parenchyma, a grid was added to each image and the type of cell or intercellular space was recorded in each grid point.
Leaves for scanning electron microscopy (SEM) analysis were air dried, cut to pieces that fitted to be mounted on aluminium stubs and coated with gold-platinum (∼50 nm) under vacuum (Automatic Sputter Coater B7341, Agar Scientific, Ltd., Stansted, UK). The samples were examined using SEM (Philips XL30 ESEM-TMP, FEI Company, Eindhoven, the Netherlands) and digitally photographed at ×300-700 magnification depending on species and surface structure to be examined. Digital images of surface structures were analysed for trichome and stomatal density. All analyses of Arctic climate change, leaf anatomy and BVOCs 2051 digital images were done using ImageJ (1.47 v, Wayne Rasband, National Institutes of Health, Bethesda, Maryland, USA).
Leaf dry mass per area (MA) was measured for the broadleaved B. nana and. S. arctica.

Data analysis
Statistical analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). Treatment effects on BVOC emissions across the study period were tested by repeated Palisade (P) and spongy (S) parenchyma are marked. Glandular trichomes also occurred on the inner surface of C. tetragona and on the surface facing the stem, particularly for C. tetragona collected in Abisko. For E. hermaphroditum, the shape change index was measured as the ratio between length 1 (L1) and 2 (L2) and leaf thickness (T) was measured at the thickest point where the leaf curls. For C. tetragona, T was estimated as an average of three points. Thicknesses of other tissues were measured at the same point as T for both species. measures linear mixed model. Treatment and time (and their interaction) were included as fixed factors, whereas block was used as a random factor. However, time and the interaction term were excluded from the model when P ≥ 0.2. Furthermore, the block factor was excluded from the model when background variation could not be assigned to the differences between blocks, that is when the covariance parameter estimate was ≤0. Repeated factor was set to a number of campaign and subject to plot. Selection of covariance structure was based on the smallest Akaike's information criteria (AIC) and was set to either compound symmetry (CS), Autoregressive [Ar(1)] or Toeplitz (TOEP).
Analysis of variance was used to assess the treatment effects on BVOC emissions within each campaign with treatment and block as factors. Block was excluded when P ≥ 0.2. Dunnett's test was used to compare the warming and shading treatment to the control. Similar models were also used to assess treatment effects on anatomy variables.As comparison of species was not the purpose of this study, each species was subjected to separate analyses. Transformations were made when the variances were heterogenic.
Partial least squares regression (PLS) was used to test whether the common leaf anatomy data for the broadleaved species B. nana and S. arctica correlates with the BVOC emission data using SIMCA (Umetrics, version 13.0.3.0, Umeå, Sweden). The emission data used for the PLS originated from the campaign with highest BVOC emissions. The data from July 3 to 4 was therefore chosen for B. nana and those from July 30 for S. arctica. PLS components were extracted following centring and unit variance scaling of the variables, and the PLS model was evaluated by segmented cross validation using six segments based on the maximum number of blocks in the experiments. The scores for the two first PLS components were analysed by similar models as described earlier.

Treatment effects on BVOC emissions
Abisko For E. hermaphroditum, no significant treatment effects were found for emission or emission potential of any of the compound groups (Supporting Information Tables S3a & S5a). ORVOCs, with 3-hexenol and 3-hexenal as the major compounds, constituted 47% of the total emissions averaged across campaigns (Fig. 4a). Isoprene constituted 26% of the total emissions averaged across campaigns, and MT and SQT minor part of the emissions with 1 and 12%, respectively.
There were no statistically significant treatment effects on emissions or emission potentials of C. tetragona (Supporting  Information Tables S3b & S5a). However, shading tended to increase the oMT and MT emission potentials within the August 4-7 campaign (P < 0.1, Dunnett's test) (Supporting Information Table S5a). MTs and oMTs constituted 11 and 4% of the total measured emissions averaged across campaigns with limonene and cymene being the most emitted MTs, and eucalyptol and α-terpineol the most emitted oMTs. ORVOCs, with butyl benzoate and 2,2,4-trimethyl-1,2-dihydroquinoline as the dominant compounds, represented the largest proportion (31%) of the averaged total BVOC emission (Fig. 4b).
C. tetragona emitted a notable amount of isoprene especially in the August 4-7 campaign, where the emission averaged for the treatments was 1.5 μg g −1 h −1 . The total SQT emission constituted 16% of the total mean BVOC emission.
For B. nana, treatment effects were only found in the July 3-4 campaign, where warming led to a significantly decreased emission of isoprene, OVOCs and total BVOCs (P < 0.05, Dunnett's test) (Supporting Information Table S3c). Shading significantly decreased the isoprene and OVOC emission, while increasing the SQT emission on July 3-4. Emission potentials showed similar effects for isoprene and SQTs (Supporting Information Table S5a). ORVOCs, which constituted 79% of the total emission averaged across campaigns, were not affected by the treatments (Fig. 4c). The dominant compounds in this group were 3-hexenol and 3-hexenyl butyrate. The dry mass-and leaf area-based emissions showed consistent results (data not shown).

Zackenberg
S. arctica only showed a significant treatment effect on July 30 when isoprene emission and emission potential were significantly reduced by the shading (Supporting Information  Tables S4a & S5b). Additionally, shading led to a tendency towards a decrease in the measured OVOC emission on July 16. The S. arctica emission was dominated by ORVOCs (Fig. 5a), which represented 60% of the emissions and had 2-hexenyl acetate and 2-hexenal as dominant compounds. The most emitted single compound, isoprene, constituted 38% of the total mean BVOC emission. Isoprene emission was particularly high on July 30 where the emission rate averaged for the treatments was 8.53 μg g −1 h −1 . Also for S. arctica, the dry mass-and leaf area-based emissions showed consistent results (data not shown).
No significant treatment effects were found for emissions or emission potentials of C. tetragona (Supporting Information Tables S4b & S5b). A tendency towards an increased SQTs emission was found on July 17 because of warming (Supporting Information Table S4b). ORVOCs constituted 52% of the total BVOC emission averaged across the  warming (W) treatments in Zackenberg during four and three measurement campaigns, respectively, in the growing season. Contributions of isoprene, oxygated monoterpenes (oMTs) non-oxygenated monoterpenes (MTs), sesquiterpenes (SQTs), other reactive volatile organic compounds (ORVOCs) and other volatile organic compounds (OVOCs) to the total BVOC emission are shown. Significant treatment effects were only found for S. arctica (see Supporting Information Table S4a and b for statistics). Note the different y-axis scales. campaigns (Fig. 5b), and had benzaldehyde, hexane and methylcyclopentane as the major compounds. Isoprene was detected with noteworthy emissions on August 14, when a mean emission for all treatments was 0.5 μg g −1 h −1 . Total MT emission represented 10% of the mean total BVOC emission, whereas SQTs constituted 2%.
Lists of individual compounds detected for each species are shown in Supporting Information Table S6 and S7.

Treatment effects on leaf anatomy
Abisko While warming had no significant effects on the leaf anatomy of E. hermaphroditum, shading altered it by making the needle-like curled leaves broader, but flatter (significantly higher shape change index in Table 2). The leaf thickness was also significantly increased by shading. Furthermore, shading significantly decreased the stomatal density and tended to decrease the spongy parenchyma thickness (Table 2).
For C. tetragona, warming significantly increased the thickness of the inner epidermis (towards the cavity) and had a tendency to decrease palisade : spongy parenchyma ratio (Table 2). Warming also had a tendency to decrease the density of glandular trichomes on the inner surface, while leading to a significant increase in the stomatal density (Table 2). Shading significantly increased the stomatal density, decreased the thickness of the outer epidermis (facing outwards), and tended to cause fewer trichomes on the surface facing the stem.
For B. nana, warming increased the leaf thickness by 13%, and this increase was nearly significant (Table 2). Warming also caused a significantly thicker spongy parenchyma and a decreased palisade : spongy parenchyma ratio (Table 2). Shading significantly decreased this ratio, and the thickness of the palisade parenchyma.

Zackenberg
Warming significantly increased the thickness of the adaxial epidermis and decreased the palisade : spongy parenchyma ratio for S. arctica (Table 2).
Neither warming nor shading had statistically significant effects on the leaf anatomy of C. tetragona from Zackenberg. However, a not statistically significant increase in the thickness of the inner epidermis and a decrease in the palisade : spongy parenchyma ratio under warming was similar to the significant alterations found for C. tetragona from Abisko (Table 2 & Supporting Information Table S2b).
Only variables with significant effects of warming or shading are shown (Dunnett's test, *P < 0.1; **P < 0.05). Values are means (±SE; n = 5-6). Arrows indicate a decrease or increase compared with the control treatment.
(y variables) in the broadleaved S. arctica and B. nana. The PLS component 1 showed a clear separation between the species (P < 0.01), and explained 36% of the variation in the x variables and 14% of the variation in the y variables (Fig. 6a). For B. nana, the shading treatment was clearly different from the control (P < 0.05, Dunnett's test). The PLS component 2 explained 20% of the variation in the x variables and 7% of the variation in the y variables. The regression coefficients revealed a positive correlation between isoprene emission, and the leaf anatomy variables palisade : spongy parenchyma ratio, leaf thickness and palisade parenchyma thickness (Figs 6b & 7). Furthermore, isoprene emission had a significant negative correlation with adaxial epidermis thickness (Figs 6b & 7). Isoprene emission did not correlate with abaxial epidermis, spongy parenchyma thickness, stomatal density or leaf dry mass per area. No significant correlations were found between leaf anatomy variables and oMTs, MTs or ORVOCs, respectively (Supporting Information Fig. S3a,b,d). SQT emission showed a significant negative correlation with palisade : spongy parenchyma ratio (Supporting Information Fig. S3c). The emission of OVOCs correlated negatively with the palisade : spongy parenchyma ratio and the palisade parenchyma thickness (Supporting Information Fig. S3e).

DISCUSSION
We expected to find increased BVOC emissions because of warming as observed at the ecosystem level in the Subarctic (Tiiva et al. 2008;Faubert et al. 2010) and in the majority of studies applying experimental warming (Peñuelas & Staudt 2010). Shading was expected to reduce BVOC emissions due to the light-dependency of many BVOCs (Laothawornkitkul et al. 2009;Monson & Baldocchi 2014a). However, BVOC emissions from the studied arctic plant species were barely   Figure 7. Regression coefficients of a partial least squares (PLS) model for the various leaf anatomy variables affecting isoprene emission. Error bars indicate ± 2 standard deviations of the regression coefficients. When the error bar does not include zero, the variable has a significant contribution in explaining the isoprene emission rate.
affected by warming and shading. This was an unexpected result, which is in accordance with previous findings from the arctic region (Rinnan et al. 2011). Whereas Rinnan et al. (2011) could offer no mechanistic explanation for the lack of effects, we suggest that it may be partly explained by long-term acclimation of the plants to the mimicked climate change, as shown by the anatomy responses. Furthermore, low emission rates combined with high natural variation may have reduced the potential of detecting differences.
Our results suggest that anatomy responses of arctic plants to warming differ from species from lower latitudes, whereas the effects of shading were more comparable with earlier studies (Larcher 2003). Most boreal forest species have responded to long-term warming in field exposures by modifying leaf structure towards for example thinner leaves and epidermis, and lower density of stomata. In an open-field experiment, Hartikainen et al. (2009Hartikainen et al. ( , 2014 found thinner leaves and epidermis of P. tremula and thinner epidermis of Betula pendula when exposed to infrared heaters, while Luomala et al. (2005) observed thinner needles in Pinus sylvestris in response to warming by closed-top enclosures of individual trees. Luomala et al. (2005) also reported decreased stomatal density in P. sylvestris. In contrast to these boreal species, we found warming to result in thicker leaves of B. nana and thicker epidermis for both C. tetragona and S. arctica indicating prevention of water loss (Larcher 2003) and increased stomatal density in C. tetragona in Abisko. In the studies of Hartikainen et al. (2009Hartikainen et al. ( , 2012, P. tremula and B. pendula were additionally found to emit significantly more BVOCs under warming, which is also opposite from our results. In accordance with Hartikainen et al. (2009), we found that leaf anatomy variables for the studied broadleaved species explained some variation in BVOC emissions. By a PLS regression we showed that isoprene emission increased with increasing thickness of the leaf and palisade parenchyma, increasing palisade : spongy parenchyma ratio and decreasing epidermis thickness. Isoprene is formed in the chloroplasts of the parenchyma and the emission is strongly bound to photosynthetic carbon metabolism (Loreto & Schnitzler 2010;Li & Sharkey 2013). As the tightly packed palisade parenchyma beneath the top epidermis has a high potential for photosynthesis (Monson & Baldocchi 2014b), a leaf with thinner palisade parenchyma might offer less sites for photosynthesis and BVOC synthesis (Hartikainen et al. 2009) leading to lower isoprene emission. The observed lower palisade : spongy parenchyma ratio could thus partly explain why warming did not increase isoprene emission in S. arctica. The production of other isoprenoids synthesized in the palisade cells (Loreto & Schnitzler 2010;Li & Sharkey 2013) could likewise be affected by a decrease in the palisade : spongy parenchyma ratio, also observed for B. nana and C. tetragona. Also, acclimation of BVOC synthesis because of decrease in gene transcription and enzyme activities could be a potential explanation for the unaltered emissions. However, molecular biological analyses were not feasible in this field-based study.
Thinner leaves might also have less BVOC synthetizing tissue per unit leaf area, and isoprene emission did positively correlate with leaf thickness. However, as opposed to other studies (e.g. Larcher 2003;Luomala et al. 2005;Hartikainen et al. 2009;Rasulov et al. 2014), we found that warming increased leaf thickness in B. nana and shading in E. hermaphroditum. In their study, Hartikainen et al. (2009) reasoned that warming-induced thinner leaves promoted BVOC emissions from P. tremula due to shorter diffusion pathways for BVOCs to exit the leaves. A greater thickness would on the other hand result in longer diffusion pathways (cf. Noe et al. 2008;Hartikainen et al. 2009), increased residence time of BVOCs inside the leaf and thereby enhanced probability of within-leaf metabolism (Niinemets et al. 2014) and reduced net emissions. New evidence suggests a reactive diffusion pathway, where a large proportion of isoprenoids can be oxidized along the pathway from site of synthesis to the atmosphere, whereby the emission does not equal the production (especially under oxidative stress Jardine et al. 2013;Niinemets et al. 2014).
The epidermis thickness has been suggested to influence the diffusion of lipophilic BVOCs across the epidermis (Hartikainen et al. 2009). Our observation of a negative correlation between the adaxial epidermis thickness and isoprene emission supports this hypothesis. A thicker adaxial/ inner epidermis in S. arctica and C. tetragona in response to warming may be a structural adaptation to avoid increased transpiration under warming and at the same time constrain BVOC emissions. Except for stored BVOCs in for example trichomes, BVOCs however, almost entirely escape through stomata when these are open (Harley 2013;Niinemets et al. 2014). Stomatal conductance can exert short-term control on the BVOC emission until a steady state is resumed (Niinemets et al. 2004;Harley 2013). However, our PLS results showed no correlation between BVOC emissions and stomatal density, and thus support the idea that stomata cannot control emissions over long term (Niinemets & Reichstein 2003). The changes we observed in stomatal density -a decrease under shading in E. hermaphroditum and an increase due to both warming and shading in C. tetragona in Abisko -are thus unlikely to affect BVOC emissions.
BVOCs stored in surface structures, for example many MTs, escape directly to the atmosphere depending on the temperature-dependent volatilization from storage pools (Laothawornkitkul et al. 2009). Warming decreased the density of glandular trichomes for C. tetragona in Abisko, whereas MT emissions were not affected. Less trichomes could reduce emission of stored BVOCs from this strong MT emitter (Rinnan et al. 2011;Schollert et al. 2014) and thus limit the expected increased emission under warming.
E. hermaphroditum was the only studied species, which showed neither leaf anatomy nor emission responses to elevated temperature. This lack of responses may be explained by the wide circumpolar and -boreal distribution of the Empetrum spp. in the northern hemisphere as far south as 40°N (Popp et al. 2011). The subspecies E. hermaphroditum might thus be able to thrive at higher Arctic climate change, leaf anatomy and BVOCs 2057 temperatures than simulated by this experiment and have no need for structural acclimation. The anatomy responses of E. hermaphroditum to shading, on the contrary, indicate a greater sensitivity towards decreased amount of light. In general, the responses to shading in all the species studied here were similar to typical acclimations to shady habitats (Larcher 2003) with increased leaf thickness in E. hermaphroditum and increased stomatal density of C. tetragona in Abisko being exceptions.
C. tetragona showed no significant emission responses to the treatments at either of the two locations. The anatomy responses were, however, similar: the inner epidermis thickness increased and palisade : spongy parenchyma ratio decreased because of warming at both locations, but statistically significantly only in Abisko. There are three potential explanations to stronger anatomical alterations in Abisko: (1) C. tetragona may be more sensitive to temperature changes near its southern limit, Abisko (Weijers et al. 2012), while the species might benefit from warming in the high arctic Zackenberg; (2) The species had a longer time to respond in Abisko as the experiment had been running 15 more years than in Zackenberg and as the growing season is longer; and (3) The temperature impact of open-top chambers (OTCs) differs widely across sites (Bokhorst et al. 2013) and here the greater temperature increase in the warming treatment in Abisko might be partly responsible for the greater anatomy responses.
In Abisko, the open-top tents used in the warming treatment reduced the PPFD level during the measurements by 51%. Even though summer warming by OTCs is strongly coupled with the solar irradiance (Bokhorst et al. 2013), a >4°C temperature increase still occurred.The reduced PPFD level could have limited the warming effect because of the light-dependency of many BVOCs (Laothawornkitkul et al. 2009;Monson & Baldocchi 2014a). Nevertheless, the ecosystem-level BVOC emissions have drastically increased under warming by similar tents in another experiment in Abisko (Tiiva et al. 2008;Faubert et al. 2010), and significant changes in vegetation and soil indicating long-term warming have been reported (Michelsen et al. 2012). In contrast to Abisko, the OTCs in Zackenberg caused no or only a slight temperature increase and at most a 22% decrease in the PPFD level. As earlier measurements report a 1°C increase in the warming treatment (Ellebjerg et al. 2008;Campioli et al. 2013), we expect this to be the general pattern. Most studies using OTCs and greenhouses acknowledge the potentially confounding effects on the enclosed microclimate, which can differ in intensity and direction for different sites (Bokhorst et al. 2013). Confounding effects such as decreased soil moisture, PPFD level and wind speeds are well-known (Havström et al. 1993;Bret-Harte et al. 2001;Bokhorst et al. 2013). The confounding temperature decrease in the shading treatment in this study might have affected the BVOC emissions negatively especially in Zackenberg, where the decrease was most profound. It should, thus, be kept in mind that the measured emissions only give a snapshot at a point in time, whereas the observed alterations in leaf anatomy reflect long-term effects of the treatments.