The study area is located in a native Aleppo pine forest growing on coastal sand dunes in the Biogenetic Natural Reserve Stornara in Apulia region, Southern Italy (40° 27′ 43.46″ N, 16° 55′ 00.59″ E) (Figure 1). The soil is a loose siliceous-calcareous sand (Dumontet et al., 1996). Pine is associated with evergreen sclerophyllous (Pistacia lentiscus, Phillyrea angustifolia, Myrtus communis, Rhamnus alaternus, Rosmarinus officinalis), and malacophyllous (Cistus salvifolius) shrub species. The climate is semiarid Mediterranean, with a mean annual temperature of 15.8°C, and total precipitation of 535 mm.

Three patches with different fire histories were selected within the same forest, in order to minimize environmental variability and isolate the role of fire histories. Selection was done by using the fire register of the National Forest Service (since 1960) and fire maps drawn up by means of an aerial photogrammetric chronosequence (1947, 1954, 1974, 1990, 2002). In this forest, two stand-replacing crown fires occurred, partially overlapping and forming three patches with different fire history, burning 0, 1, and 2 times (FC0, FC1, and FC2, respectively, Table 1; Figure 1). In FC0, in addition to the analysis of aerial photos, the absence of recent fires was confirmed by the absence of charcoal fragments in the first 40 cm soil profile. In January 2008, 50–100 cones older than 2 years were collected from the whole perimeter of the middle and upper crown of 10 randomly selected plants from each fire cohort. Tree total height and diameter at breast height were measured, using a Tandem clinometer (Suunto, Finland) and a meter tape, respectively, to be used as a proxy of ontogenetic age, for all sampled trees (Table 1). In order to accurately represent the seed pool accessible during a fire event, both brown and serotinous cones were collected.

Cones from each plant were separately dried over 1 week in a well-ventilated and shaded environment, put in jute bags and moved to an oven ventilated at 50°C to facilitate scale shrinkage. The seeds extracted were stored in vacuum-type PVC bags at 4°C. From each of the different fire cohorts at least 1200 seeds for the germination tests and 125–150 for morphometric analysis were used. Seeds from the various plants of each fire cohort were pooled together in the germination experiment, in order to investigate the overall germination responses of each cohort. The seeds were carefully selected from those with intact coats. A preliminary cutting test allowed the determination of 11 mg as the threshold mass for fully developed seeds to be used in the experiments. Unwinged seeds were weighed by means of a semi-micro balance to the nearest 10⁻² mg.

Morphometric analyses were performed on 125–150 seeds per fire cohort (125 from FC0, 125 from FC1, and 150 from FC2), for a total of 400 seeds. Each seed was photographed in its coronal/abaxial plane (Figure 2A; see Kolotelo, 1997 and Terskikh et al., 2005) at 2.5× magnification using an optical microscope (Dialux 20, Leitz, Germany) equipped with a CoolSnap K4 camera (Photometrics, USA). The coronal shape of each seed was then numerically derived from the digital images of the seeds through an ad hoc computer script (unpublished data) developed for the R programming environment (R Core Team, 2014), based on the codes reported in Claude (2008). The script (i) imported the binary images of the seeds (threshold performed using the ImageJ 1.43u software package), (ii) derived the Cartesian coordinates of each pixel on the contour of the seed, (iii) calculated the centroid of the contour points, (iv) rotated the point set around the centroid, and (v) compressed the information by storing the length of 36 vectors, from the centroid to the contour points, spread by 10° from each other. The shape of the seeds was then defined by 36 variables, each one corresponding to the length of a radius.

The seeds were carefully uncoated from the micropylar end using hand cutters and weighed again. Uncoated seeds were free-hand cross-sectioned with a razor blade (Figure 2B), obtaining three to five thin sections, which were mounted on a glass slide and photographed with the aforementioned apparatus. The digital images were then analyzed with the free software package ImageJ 1.43u (http://rsb.info.nih.gov/ij/index.html), fitting their perimeters with polylines. The section with the largest surface was selected and used for the measurement of the section area of the embryo with and without the megagametophyte. The area of the megagametophyte and the mass of the coat were calculated as the difference between the previous measurements.

Germination tests involved experiments under different osmotic potentials (Ψ_π) and pH. Seeds were germinated in 9 cm disposable Petri dishes over four layers of filter paper wetted with 10 ml of test solution. In order to minimize Ψ_π and pH variations over time, seeds were transferred to freshly prepared dishes every 3 days. Petri dishes were placed in the dark in a germination chamber set on a 13 h-15°C:11 h-25°C cycle. This experimental setting aimed to reproduce the environmental conditions experienced by seeds in the field during germination: seeds buried in the dark under 2–3 cm of black post-fire substrate (A. Saracino, unpublished results), at the night:day soil temperatures previously recorded during seedling emergence from October to November (Saracino et al., 1993).

In order to simulate the variable conditions of the post-fire field substrates (Bodí et al., 2014), the germination tests involved six Ψ_π levels between 0 and −1.2 MPa, and six pH levels between 6 and 11. Mannitol solutions were employed for the Ψ_π germination tests to avoid the toxic effect of polyethylene glycols on the germination of P. halepensis seeds (Leshem, 1966), whereas different buffered solutions were employed in the pH germination tests (pH 6.0: MES 50 mM; pH 7.0: MOPS 50 mM; pH 8.0 or pH 9.0: TRIS 50 mM; pH 10.0 or pH 11.0: CAPS 50 mM). Each treatment was replicated five times on batches of 20 seeds. Germinated seeds were counted daily and removed as soon as radicles had protruded until the 28th day after seeding. Ungerminated seeds were then tested for viability by means of tetrazolium tests to refer the germination percentage to the total of viable seeds.

The overall differences among the fire cohorts in the morphological traits of the seeds were inferred through a multivariate analysis of variance (MANOVA). The mass of the uncoated seeds (Mu), the mass of the coat (Mc), the area of the megagametophyte (Am), the area of the embryo (Ae), and the area of the coronal plane (Ac), were included in the model as dependent variables. A generalized canonical discriminant analysis (CVA) was used to evaluate the contribution of the biometrical traits in differentiating the seeds among the three cohorts. Moreover, a MANOVA and a CVA were performed to analyse the variations in seed shape in relation to the number of fires, using the 36 radii defining the contour of the seeds as dependent variables and the fire cohort as the fixed factor. The relationship between each morphometric parameter and fire history was assessed through linear mixed models using the number of fires and the tree size (both height and diameter at breast height as descriptors of plant ontogenetic stage) as fixed factors and the mother plant as a random factor. The significance of fire history and tree size was then evaluated using a Kenward-Roger adjusted F-test.

Cumulative germination percentages (G_f) and mean germination time (MGT), defined as the time when half of the viable seeds had germinated, were selected as descriptors of seed germination performance. G_f values were logit-transformed, according to Warton and Hui (2011) using the equation: Gf=(Gf+ε)/(1−Gf+ε) where ε = 0.01 to obtain a normal distribution.

The relation between G_f or MGT with fire history and tree age was evaluated through ridge regression using the number of fires, tree age and Ψ_π or pH as predictors. Tree age, rather than tree size, has been used in analysing germination data, since in the experiments, seeds from various plants were pooled together.

The ridge regression technique was chosen in relation to its L2 regularization, avoiding the shrinkage to zero of the weakest predictors, and to its usefulness in dealing with non-orthogonal problems (Hastie et al., 2009). Moreover, the predictors were scaled to mean zero and unit variance to allow their importance ranking in predicting G_f and MGT. The penalty parameter (λ) for model selection was chosen using the method proposed by Cule and De Iorio (2012).

The relationship between G_f or MGT and Ψ_π or pH was then modeled separately for the three fire cohorts using linear models and the differences in the slopes were tested with z-tests. All analyses and graphics were performed with the stats, lme4, lmtest, nortest, candisc, pbkrtest, ridge, and vioplot packages in R 3.1.1 (R Core Team, 2014).

Differences in seed morphology among the three cohorts were highlighted in a multivariate context, with coronal area, the mass of uncoated seed and embryo size as the morphometric traits that contributed more to separate the three fire cohorts. The differentiation among cohorts is primarily related to the coronal area and the mass of the uncoated seeds, indicating an effect of the number of fires on these traits. Although variations in seed coronal area in relation to age/ontogenetic stage of the mother plants could not be definitively excluded, our result strongly suggests that the increase in coronal area, and to a lesser extent in other morphometric traits, is related to the fire history rather than to tree age. Indeed, young pine trees should produce fewer and smaller seeds than older trees in the same environment (Thanos and Daskalakou, 2000), whereas the opposite was observed in our study, supporting the hypothesis that seed size and shape are traits positively affected by fire. Our results could not confirm the influence of fire passages on the size of the megagametophyte and seed coat, which are considered important in the fire ecology of this species (Salvatore et al., 2010). Conversely, embryo size contributed to separate the three fire cohorts, and slightly increased from FC0 to FC2. Such an occurrence can be explained considering the importance of seed germination kinetics on seedling survival, which is a function of the embryo rather than of the megagametophyte. Indeed, larger embryos can germinate earlier (Vivrette, 1995) and overcome intraspecific competition for space, rather than nutrients, which are usually not limiting in post-fire substrates. Improved competitive advantage and seedling survival in larger Aleppo pine seeds, attributed to the higher number of cotyledons, were also reported by Daskalakou and Thanos (2010). Crown ontogeny may instead play a role in determining the higher variability in seed morphological traits exhibited by the younger fire cohort, since we collected cones on both stems and branches.

The variation in seed traits is often involved in plant evolution: in obligate seeders, seed-related traits like shape and mass (Weiher et al., 1999) usually have a major role in both post-fire seedling distribution and establishment (Hammill et al., 1998; Gómez-González et al., 2011, 2016; Shryock et al., 2014; Ruprecht et al., 2015; Jiménez-Alfaro et al., 2016). In our case the observed anisotropic increase in seed size, which leads to a change in seed shape, might affect a number of processes like dispersion, penetration and persistence in post-fire substrates, in which also seed wing traits may be involved. However, we cannot definitively exclude a non-adaptive age dependent plasticity in these traits. Therefore, this topic deserves further investigation.

In our study, the patterns of seed germination varied in relation to fire history, with seeds from FC2 exhibiting a greater ability to cope with the constraints imposed by the ash bed in the elective post-fire germination microsite of the crown projection area. Specifically, the number of fires improved seed tolerance to alkaline pH more than to the osmotic potential, both in terms of germination percentage and mean germination time. By contrast, the age of mother plants seems to have a greater effect on the germination performances under osmotic constraints. Cumulative germination showed the clearest responses, gradually shifting from marked inhibition at alkaline pH in FC0 seeds to a neutral response in FC2 seeds. Wood ashes of coniferous species typically reach pH values up to 11 (Raison, 1979; Thomas and Wein, 1990; Henig-Sever et al., 1996), which can inhibit germination. Indeed, Henig-Sever et al. (1996) considered the high soil pH induced by post-fire ash accumulation as the main abiotic factor controlling Aleppo pine forest regeneration, despite the promoting effect of nitrate and ammonium availability on seed germination (Henig-Sever et al., 2000). In addition, leaching by rain is usually not sufficient to produce a favorable seedbed before the occurrence of seed fall, as demonstrated in P. banksiana stands (Thomas and Wein, 1994). The neutral response to pH of FC2 seeds suggessts that, at least in the short term, fire recurrence does not constrain the reproduction ability. Actually, it seems that after two repeated stand-replacing crown fires, Aleppo pine has the same, or even a better, capability to overcome the main abiotic constraint for the germination of seeds in post-fire substrates. Total germination under the osmotic constraint is above 80% in all fire cohorts, and age is the most important trait in predicting this behavior. This evidence suggests that fire frequency affects reproductive performance to a lesser extent than osmotic stress and is consistent with previous work which considers this environmental constraint more related to drought than to fire regimes (Falusi et al., 1983; Thanos and Skordilis, 1987). As such, in our case the high resistance to osmotic stress could be considered an adaptation to the semi-arid climate of the area. However, the potential interaction between drought and osmotic potential deserves further investigation, since Aleppo pine seed germination occurs in autumn-winter, when soil is rewetted by precipitation.

The mean germination time pattern in relation to the osmotic potential was non-linear in relation to fire history, with FC1 seeds consistently germinating earlier than the others. This pattern may be related to the length of fire intervals (about 50 y for FC1 vs. 24 y for FC2). Indeed, mother plants of FC1 seeds were recruited on ash beds deeper and richer in soluble cations (Alifragis et al., 2001) than those of FC2 seeds, due to the higher litter and biomass accumulation during the FC1 life span. These features increase both fire intensity and the amount of ash deposited (Cerdà and Doerr, 2008; Bodí et al., 2014), suggesting an involvement of fire also in determining seed resistance to osmotic stresses.

Experimental results on the effects of the osmotic potential on Aleppo pine seed germination vary in relation to seed provenance (Thanos and Daskalakou, 2000). Henig-Sever et al. (1996) pointed out that osmotic potential only marginally affects the germination performance, and Thanos and Skordilis (1987) reported that Aleppo pine can show remarkable germination rates under osmotic stress attaining Ψ_π > −1.46 MPa. Conversely, Falusi et al. (1983) showed that Ψ_π ranging from −0.2 to −0.6 MPa can reduce germination rates and root growth in some provenances. However, literature results do not resolve the concomitant fire history and seed provenance effects, which limit the reach of the comparisons with our findings. It is well documented that repeated stand-replacing fires reduce the vegetative and reproductive fitness of Aleppo pine by delaying the onset of cone production, reducing the cone crop per tree and the number of reproductive pines (Eugenio et al., 2006; Espelta et al., 2008). In addition, future scenarios in the Mediterranean Basin predict an increase in fire frequency (Mouillot et al., 2002; Pausas and Keeley, 2009) due to a reduction in rainfall (IPCC, 2013). According to our findings, it seems that, at least in the germination phase, fire recurrence does not constrain the reproduction ability of P. halepensis due to production of seeds with increased tolerance to the constraints imposed by the ash bed. Instead, the neutral response in relation to pH entails the absence of ash pH constraint on the reproductive niche of the Aleppo pine cohort experiencing repeated fires, ensuring maintenance of fitness and population persistence in fire-prone areas. These changes should occur under several mechanisms implying genetic selection, at the germination stage, of seeds with better germination performance on ash beds. Under the genetic selection hypothesis, the ash layer selective pressure on seed germination would drive population adaptation to repeated fires, with future plants differentially recruited in relation to morphological, physiological or biochemical seed traits, and the eventual fixation of fire-related traits. As known, the environment can also provide adaptive transgenerational plasticity. This is often related to changes to seed provisioning and biochemistry as well as epigenetic mechanisms (reviewed by Herman and Sultan, 2011). They are environmentally sensitive and heritable, and thus play a key role in regulating transgenerational effects following plant stress conditions (Jablonka and Raz, 2009). In this perspective, an alternative explanation of the differences observed among the fire cohorts could be related to short-term adaptation due to epigenetic modifications, e.g., post-translational changes of histone proteins, cytosine DNA methylation, action of small RNA (Danchin et al., 2011). These modifications are heritable, can all cause changes in gene expression and, as recently reported by Yakovlev et al. (2012), may represent an efficient mechanism to survive under changing environments. The age-dependent responses highlighted in this work can be considered positive traits to cope with frequent fire regime. Further studies are required to assess if they are phenotypic traits varying with growth and development, or if they are the result of long-term adaptation of this species to a fire prone environment.

We are confident that this work may open many research fronts, with further research involved in clarifying whether these observed short term responses are able to drive the local adaptation of new ecotypes, eventually allowing the maintenance of the reproduction ability of P. halepensis populations under recurrent fire regime.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.