The field study was conducted at a xeric site at 750 m asl in a dry inner Alpine valley of the Eastern Alps in Austria (47° 13′ 53″ N, 10° 50′ 51″ E). Based on >100 years of climate records at Ötz (812 m asl, 5 km from the study area) mean annual air temperature and total precipitation amount to 7.3°C and 724 mm, respectively (long-term means during 1911–2017). According to the FAO classification system (FAO, 2006) soils of the protorendzina type, i.e., rendzic and lithic leptosols, are primarily developed. As a result of low soil depth and the coarse-textured structure, soils have a low water holding capacity. In this drought-prone environment, Scots pine (Pinus sylvestris L.) dominates and forms poorly growing open stands (Erico-Pinetum typicum, Ellenberg and Leuschner, 2010). We selected a south-west facing steep slope (c. 30°), where P. sylvestris rejuvenates naturally under open canopy. In order to minimize the impact on the protected forest area, we used young trees for the girdling experiment instead of mature trees. Stem height and diameter of selected trees (n = 7) amounted to 1.5 m and 2.9 cm, respectively. Age of trees at 30 cm stem height was 35 ± 5 years (Table 1).

Girdling was applied to block phloem C transport from aboveground to belowground sinks. Trees (n = 3) were girdled after cessation of shoot and RG in mid-July 2019 (day of the year (doy 199) at a stem height of 25 cm above the soil surface by carefully detaching a 2–3-cm-wide band of bark including periderm, living phloem, and cambium. Exposed wood was treated with a fat-containing cream to prevent dehydration. At the time of girdling, cambial cell division of trees has already stopped for about 6 weeks (cf. Figure 2).

Temperature compensated electronic diameter dendrometers with resolutions of <3 μm (DD-S, Ecomatik, Munich, Germany) were installed on trees (controls: n = 4; girdled trees: n = 3) at 5–7 cm above girdling to record intra-annual dynamics of RG. The temperature coefficient of the sensor was <0.2 μm/K (unverified information of the producer). The dead outermost loose layer of the bark (periderm) was slightly removed to allow proper mounting of the dendrometer and to ensure close contact with the stem. Data were recorded every 30 min with analog data loggers (HOBO UX120-006M, ONSET, Bourne, MA, United States).

Daily stem diameter variations (SDVs) were calculated by averaging all daily measurements (48 values per day), i.e., one value per day was extracted from the time series. The daily mean approach yields time series of daily SDVs, which consist of both water- and growth-induced diameter changes (Deslauriers et al., 2007). It has to be considered that irreversible growth-related diameter increments recorded by dendrometers include formation of xylem, phloem, and outer bark (Plomion et al., 2001). Time series of daily SDVs were set to zero on March 1, and we modeled short-term variations in intra-annual RG with the Gompertz function, which is commonly used to describe RG dynamics in trees (Zeide, 1993; Rossi et al., 2003; De Micco et al., 2019), by applying the Origin software package (OriginLab Corporation, Northampton, MA, United States).

At the study site, environmental conditions [air temperature, relative air humidity (RH), and daily precipitation] were continuously monitored during the study period (March through October 2019) at 2 m aboveground (ONSET, Pocasset, MA, United States). In addition, three soil moisture sensors (ThetaProbes Type ML2x, Delta-T, Cambridge, England) were installed at 5–10 cm soil depth to record changes in volumetric soil water content (SWC). In data loggers, measuring intervals for all sensors were programmed to 30 min. All measurements per day (i.e., 48 values) were used to calculate mean daily air temperature and relative air humidity (RH). The equation presented in Prenger and Ling (2000) was used to compute vapor pressure deficit of the air (VPD). Climate records and SWC during the study period are depicted in Figure 1. Student’s t test was applied to detect significant differences among climate and environmental variables during the regular (i.e., pre-girdling) and induced (i.e., post-girdling) growing period (cf. Figure 2).

Changes in wood anatomy (radial CLD; CA; and radial cell wall thickness, CWT) in response to C manipulation were analyzed at the end of the growing season. Stem cross-sections were collected from all trees (controls and girdled trees) at the position of dendrometers. In girdled trees additional cross-sections were sampled 5 cm above and below the girdling zone. The chosen distance to the girdling zone ensured that influence of wound responses on wood formation was avoided. CWT, CA, and CLD were measured in earlywood and latewood on stem transverse sections of ∼20-μm thickness, which were cut using a sliding microtome. To distinguish between earlywood and latewood tracheids, the Mork’s index, i.e., the ratio between the double CWT and CLD (both measured in radial direction), was computed (Denne, 1988). In non-girdled controls (n = 4 trees) and in girdled trees (n = 3 trees), anatomical features of tracheids were determined in earlywood and latewood pre- and post-girdling. Additionally, continuous records of wood anatomical parameters were determined above and below girdling and in non-girdled controls along five radial cell rows per tree, and mean values were calculated. In these time series time of girdling was deduced from intra-annual wood formation dynamics determined in P. sylvestris within the study area (Gruber et al., 2010) and abrupt alterations in anatomical traits. Because separation of individual cell walls could not be unequivocally detected in all samples, CWT of adjacent cells was recorded (double radial CWT), and this value was then halved yielding radial CWT. Cell anatomical parameters were measured by applying the image analysis software PROGRES Gryphax (Version 2.0.0.68, Jenoptik Optical Systems GmbH, Jena, Germany) and were recorded throughout five earlywood and latewood cells along five cell rows (i.e., n = 25 cells per sample), and mean values and standard deviations were calculated. The proportion of cell wall material was calculated as the ratio between CLD and CWT. A decrease/increase in CLD:CWT indicates an increase/decrease in wood density. Student’s dependent sample t test was used to determine significant differences among cell anatomical traits pre- and post-girdling and non-girdled controls.

In contrast to tree species from the Mediterranean regions (e.g., Camarero et al., 2010; Battipaglia et al., 2016; Pacheco et al., 2018), which use adequate water availability in spring and autumn for tree growth, but show strongly reduced or no growth during the dry and hot summer period, a comparable bimodal growth pattern does not occur in P. sylvestris in dry inner Alpine environments. Several authors (Gruber et al., 2010; Oberhuber et al., 2014; Swidrak et al., 2014) reported a unimodal growth pattern characterized by early peak of RG in mid-May to early June in coniferous species (P. sylvestris, P. abies, and Larix decidua) within the study area, although extended dry periods frequently occur in spring and higher precipitation during summer would provide more favorable environmental conditions for tree growth. It was suggested that extreme environmental conditions, i.e., drought stress and nutrient deficiency of the predominantly dolomite bedrock cause an early shift in C allocation from aboveground stem growth to belowground sinks to ensure adequate resource acquisition (Oberhuber and Gruber, 2010; Swidrak et al., 2013). Dendrometer records of root RG of mature P. sylvestris trees having a diameter of approximately 10 mm also revealed unimodal growth lasting from June through early August (Supplementary Figure 1).

Based on a previous experimental set-up using potted P. abies saplings (Oberhuber et al., 2017), we expected that girdling also triggers cambial reactivation of P. sylvestris under field conditions. Results of this study confirmed our hypothesis and revealed reactivation of cambial activity by phloem blockage 2 weeks after girdling at a drought-prone site. Girdling occurred in mid-July (doy 199), i.e., 6 weeks after cessation of the regular RG period excluding wall-thickening of latewood tracheids, which at xeric sites lasts until September (Gruber et al., 2010). Phloem blockage triggered cambial reactivation causing a bimodal RG pattern, i.e., a spring peak (doy 104) was followed by a secondary peak in summer (doy 228). While in Mediterranean areas bimodal growth is initiated by increase in water availability in autumn after intense summer drought, we could show that reactivation of cambial activity can be induced by an endogenous trigger, i.e., an increase in tree C status. In accordance with other studies (e.g., Daudet et al., 2005; De Schepper and Steppe, 2011; Oberhuber et al., 2017) results strongly suggest that interruption of C transport in the phloem to belowground sinks increased stem C availability above girdling inducing reactivation of cambial activity. This reasoning is corroborated by significant decrease in fine root mass of potted P. abies saplings in response to blockage of phloem C transport (Rainer-Lethaus and Oberhuber, 2018). Below girdling phloem blockage caused cessation of cell differentiation (evident in the form of reduced CWT of last latewood tracheids) and did not lead to a resumption of cambial activity. Hence, C availability in the stem, i.e., an endogenous factor is important for RG to occur under extreme environmental conditions. Based on findings of Smith and Stitt (2007) and Lastdrager et al. (2014), we suggest that sugar signaling—induced by interruption of phloem C transport—is involved in reactivation of cambial activity and triggering of bimodal RG in P. sylvestris trees.

Plant hormones direct growth and development and also responses to environmental stimuli (Davies, 2010). Accordingly, tree-ring formation, i.e., cambial activity and xylem cell differentiation, is induced and controlled by hormones (for a review see Buttò et al., 2020). Specifically, cambial activity is known to be highly responsive to the growth-promoting hormone auxin [indole-3-acetic acid (IAA); Uggla et al., 1998; Dünser and Kleine-Vehn, 2015; Bhalerao and Fischer, 2016] and Uggla et al. (1996) found a steep radial concentration gradient of IAA in Scots pine peaking in the cambium meristem. IAA is transported in the mature phloem in a basipetal flux from source tissues (leaf primordia and young leaves) to proximal regions (Muday and DeLong, 2001; Hacke et al., 2017). Several findings argue against the influence of accumulation of IAA above girdling on cambial reactivation: (i) heating of the stem can trigger cambial activity during dormancy (Oribe et al., 2003; Gričar et al., 2006; Begum et al., 2010), (ii) high levels of IAA are found in cambial tissues in the dormant period (Sundberg et al., 1990; Eklund et al., 1998; Funada et al., 2002), (iii) reduced responsiveness of cambial tissues to IAA during activity-dormancy transition (Baba et al., 2011), and (iv) sugar signaling can stimulate cambial cell division (e.g., Uggla et al., 2001; Lastdrager et al., 2014). The lag of about 2 weeks between girdling and onset of the induced RG period above girdling might be due to an extensive drought period at the time of girdling or as Oribe et al. (2003) suggested, that newly formed phloem cells are necessary for xylem formation to resume. The latter assumption is supported by results of Swidrak et al. (2014), who reported that phloem formation in P. sylvestris precedes xylem formation within the study area by about 3 weeks. In general, girdling induced cambial reactivation after regular spring growth was made possible because at the time of girdling the cambial meristem was not yet in endodormancy state (Lang et al., 1987; Chang et al., 2021), which can only be released by external triggers, i.e., chilling, photoperiod, and temperature (e.g., Badeck et al., 2004; Pallardy, 2008; Swidrak et al., 2011; Basler and Körner, 2014; Begum et al., 2018).

The increase of C availability above girdling strongly affected RG rate and intensity (increase by 154 and 136%, respectively, compared with regular RG), while RG duration remained quite constant amounting to about 2 months (excluding duration of cell wall thickening). Hence, more than a doubling of increment in the induced compared with regular RG period resulted from striking increase in RG rate. This finding is in accordance with Rathgeber et al. (2011), who reported that the extent of RG in silver fir (Abies alba) is more related to the rate of cell production than to its duration. Hence, phloem blockage of C transport to belowground organs provided a continuous supply of carbohydrates to sustain increased rate of cambial cell division and wood formation, processes that are considerable energy sinks (Oribe et al., 2003; Koch, 2004; Muller et al., 2011). Several authors (e.g., Simard et al., 2013; Delpierre et al., 2015; Guillemot et al., 2017) also reported that tree growth is limited by C competition between sinks rather than by C resources. During the induced RG period, which extended from August through September, C allocation was primarily restricted to cambial activity and wood formation, because within the study area shoot and needle growth of P. sylvestris ceases in June and July, respectively (Swidrak et al., 2013), and apical meristems remained in dormancy state after girdling. Similarly, internal shifts of C allocation were found to affect the second growth peak in Pinus pinaster, a species showing a bimodal RG pattern (Garcia-Forner et al., 2019). It has to be considered that water availability is the primary growth-limiting environmental factor within the study area (Pichler and Oberhuber, 2007; Oberhuber et al., 2011; Gruber et al., 2012) and therefore increase in precipitation by c. 30% during the induced compared with regular RG period might also be involved in RG increase at the xeric study site. However, air temperature was found to be inversely related to RG and stem water deficit of P. sylvestris within the study area (Oberhuber et al., 1998, 2015; Oberhuber, 2017). Therefore, strikingly higher temperatures during the induced compared with regular RG period (+6.7°C) increased evapotranspiration, which probably compensated the stimulating effect of higher precipitation on RG for the most part. Lack of increase in SWC despite higher precipitation is explained by increase in evapotranspiration together with surface run-off on the steep slope during high precipitation events in summer.

Although maximum daily RG in conifers from cold environments is related to photoperiodic growth constraint to allow xylem differentiation to be completed before early frosts occur (e.g., Rossi et al., 2006; Gruber et al., 2009), this does not apply for P. sylvestris on drought-prone sites (Swidrak et al., 2011). Because both RG periods are characterized by (i) the same duration (about 2 months), and (ii) the same length of time until the maximum daily RG rate was reached (about 2 weeks), an endogenous control of cambial activity and tracheid differentiation as an adaptation to extreme environmental conditions (drought stress, nutrient deficiency) can be assumed.

Wood formation is a highly C demanding process (Koch, 2004; Simard et al., 2013; Cuny et al., 2015) and low sink priority of the cambium for C was frequently reported (Cannell and Dewar, 1994; Ericsson et al., 1996; Muller et al., 2011; Heinrich et al., 2015). Therefore, we expected a clear role of C accumulation above girdling on wood anatomical traits after interruption of basipetal C transport by girdling. This hypothesis was confirmed by missing significant differences (p > 0.05) of wood anatomical parameters (i.e., CLD, CA, CWT, and CLD:CWT) between earlywood and latewood tracheids after girdling, i.e., the characteristic wood anatomical pattern in conifers (thin-walled earlywood cells with large lumina vs. thick-walled latewood cells having narrow lumina) was not sustained as a result of increased C supply. Significant decrease in CLD and CA in earlywood after girdling can be explained by (i) decrease in xylem sap flow frequently reported to be a consequence of physically blocking phloem transport (e.g., Zwieniecki et al., 2004; Lopez et al., 2015; Oberhuber et al., 2017) and leading to decrease in turgor pressure and cell enlargement (Cabon et al., 2020), and/or (ii) high C availability resulting in faster cell wall deposition, which reduces cell enlargement time (Cartenì et al., 2018). That cell enlargement time primarily explains CLD in conifers was reported by Cuny et al. (2014). Increase in osmotically active C compounds after phloem blockage are required to produce adequate wall-yielding turgor pressure for cell expansion (Pantin et al., 2013; Steppe et al., 2015) and explain significant increase in CLD and CA and concomitant decrease in wood density (increase of CLD:CWT ratio) of latewood tracheids formed after girdling. Furthermore, significant increase in CWT and wood density (decrease in CLD:CWT ratio) in earlywood tracheids after girdling can also be regarded as an outcome of C accumulation above the girdling zone, because C availability is a constraint of cell wall formation (Cuny et al., 2015; Deslauriers et al., 2016; Winkler and Oberhuber, 2017; Cartenì et al., 2018).

Below the girdling zone cell differentiation ceased after interruption of C transport by phloem blockage causing a striking decrease in CWT in last few latewood tracheids (resulting in increase of CLD). Although a decrease in CWT in last latewood tracheids frequently occurs (Arzac et al., 2018; Pacheco et al., 2018; Castagneri et al., 2020), a direct link to decline in C availability by girdling can be deduced from observations that (i) cell wall thickening in P. sylvestris lasts until September within the study area (Gruber et al., 2010), (ii) reduction in C supply due to insect defoliation similarly reduces cell wall thickening in conifer species (e.g., Axelson et al., 2014; Castagneri et al., 2020; Peters et al., 2020), and (iii) cell wall thickening has a considerable C requirement (e.g., Cuny et al., 2015).