The experimental plantation was located in Saint-Roch-de-l’Achigan (45° 50′ 50″N–73° 38′ 27″W), 57 m above sea level, 55 km northeast of Montreal (Quebec), Canada. The regional climate is humid continental with noticeable seasonal temperature variations, warm, humid summers and cold winters. According to the nearest weather station in Assomption (45° 48 34”N–73° 26 05”W), the annual average minimum and maximum temperatures for the period 2003-2017 are respectively 1 ± 12°C and 11 ± 13° C. During the growing period (from May 1 to October 31, 2017), average minimum and maximum temperatures were recorded by in-field meteorological station and corresponded to 9.8 ± 5.6° C and 22.1 ± 6.5° C respectively (Amiot et al., 2020). The average annual precipitation was 1102 mm (2005-2015).

Four hectares of Salix miyabeana ‘SX67’ were established in 2008 at a density of 16,000 trees ha⁻¹ with 1.8 m and 0.35 m spacing respectively between willow rows and between cuttings within each row (Guidi Nissim et al., 2015; Jerbi et al., 2015). A first experiment aimed at treating secondary municipal wastewater effluent was conducted on the plantation between 2009 and 2012 (Guidi Nissim et al., 2015; Jerbi et al., 2015). The plantation was last coppiced prior the present experiment when plants roots were seven-year-old in the autumn of 2015. In 2016, a randomized block design was set up (plots spatially distinct from those of 2009-2012 study) comprising four irrigation treatments replicated three times: a control without irrigation (UI), irrigation with potable water (PW), and irrigation with two different doses of the same primary effluent wastewater (Lachapelle-T. et al., 2019). Only UI, PW and the lowest dose of wastewater (noted here as WWD) are used in the present study. Nine experimental square plots (3 treatments x 3 plots) of 100 m² were delimited, each containing six rows of willow with the four central rows irrigated. Four randomly chosen plants in each of those latter four rows were identified and served for most of sampling and analysis reported here.

Primary wastewater was obtained from the local municipal wastewater treatment facility and was allowed to rest for at least 24 h in a conventional septic tank prior to irrigation of the plantation with no further chemical or biological treatment. The plantation was irrigated 111 and 163 days respectively for 2016 and 2017 with a daily dose of respectively 14 mm and 13 mm for PW, 10 mm and 12 mm for WWD. The average annual loads of elements given to the WWD plants through fertigation were 594 kg N ha^-1, 58 kg P ha^-1, 174 kg K ha^-1, 1776 kg Ca ha^-1, 515 kg Mg ha^-1, 3566 kg Na ha^-1 and 5491 kg Cl ha^-1. Experimental setup schematic, characterization of primary wastewater, annual precipitation, loads of water and wastewater as well as the loads of all the nutrients through wastewater irrigation (per year and cumulative) are presented in the supplementary material ( Figure S1 , Tables S1–S3 ), further details on the experiment are described in Lachapelle-T. et al. (2019); Amiot et al. (2020) and Jerbi et al. (2020).

At the end of the 2017 growing season, all above-ground biomass from the four labelled trees within each plot were harvested, for a total of 36 plants (3 treatments x 12 replicates). All trees were fresh-weighed on the field and random subsamples of stems were oven dried at 105° C for 72 h to assess moisture content. The biomass yields were estimated based on dry matter yields at planting density of 16,000 trees ha⁻¹.

In August 2017, two 20-25 cm stem sections were collected from the 36 labelled trees. They were immediately fixed in a formaldehyde-acetic acid-alcohol solution (FAA solution: 3.7% formaldehyde, 5% acetic acid and 47% alcohol). Subsections (3-4 cm; one per tree) were labelled, vacuum infiltrated with water for 48 h and used to determine stem volume by water displacement. Wood specific gravity was later calculated from green volume and oven-dry weight at 105° C (Berthod et al., 2015; Brereton et al., 2015).

As fertigated trees were more developed than UI and PW trees, stem sections were collected at different stem height (e.g. sections at breast height for UI and PW, but at a higher level for WWD) in order to compare developmentally similar wood (Pitre et al., 2007a). Stems were sectioned where wood had already transitioned to secondary growth before irrigation began, thus allowing analysis of the wood (secondary xylem) that was formed before and during the irrigation period to assess the effect of nitrogen fertilization on growth, development and on histologic wood properties (Pitre et al., 2007a). The diameter of the sections were between 9 and 10 mm.

Prior to compositional analysis, 5 g of ODW milled and sieved biomass was extracted with 95% ethanol according to the NREL protocol (Sluiter et al., 2008), using a Dionex^® Accelerated Solvent Extractor (ASE150) (the biomass in 33 ml cell size at 100° C under a pressure of 100 bar during 3 static cycles of 5 min for each extraction). The extracted biomass was then analyzed for structural carbohydrates and lignin in accordance with Sluiter et al. (2012). All sugars were quantified by high-performance liquid chromatography ‘HPLC’ system (Shimadzu Corporation, Kyoto, Japan) with a Bio-Rad Aminex HPX-87H column and refractive index detector. The HPLC data was corrected for the standard anhydro i.e. the contribution of water to the molecule weight of sugars between the monomer and the polysaccharide form (Serapiglia et al., 2009).

Analysis of variance testing was followed by multiple comparisons of means according to Tukey’s Honestly Significant Difference (HSD) (α= 0.05) using JMP statistical software version 9.0 (SAS Institute, Cary, NC), unless otherwise stated. Pearson correlations were calculated for all pairwise combinations of xylem properties and biomass composition.

After two years of growth, the total harvested biomass yields for the plants irrigated with the primary effluent WWD were higher than those of the least productive PW irrigated trees (but not statistically distinct for UI plants) with an annual yield per hectare of 18.3 ± 3.5, 13.1 ± 1.6 and 28.8 ± 6.3 Mg ha^-1 yr^-1 respectively for UI, PW and WWD ( Figures 1A–D ) [previously published results in Jerbi et al. (2020)].

Wood density differed significantly between the different irrigation treatments, with the control UI and PW having a > 18% higher specific density than WWD with respectively 0.37, 0.39 and 0.32 g cm^-3 ( Table 2 and Figure 1H ).

Although the density of fiber and ray cells did not vary significantly between UI, PW and WWD plants ( Table 2 ), the proportion of fiber per secondary xylem area was higher in WWD plants compared to UI and PW plants, with respectively 84, 81 and 81% while vessel to fiber area and density ratio were significantly lower for PW plants in comparison to UI ( Table 2 ).

Under all irrigation treatments, wood reacted with both dyes. A higher proportion of the wood of the wastewater irrigated plants reacted more intensely with the Chlorazol black, resulting in a larger black staining region than the other treatments, especially the UI ( Figures 1E–G , I–K ). Visual observations showed that for all treatments, regions of the stem section displayed the presence of an additional layer in the inner part of the wall of some fiber cells (more pronounced for WWD), contributing to their thickness ( Figures 1M–O ). Analysis of the proportion of wood presenting such formation (recognized as tension wood), showed a significant difference between irrigation treatments, especially between the UI control and the WWD plants, with the proportion respectively of 27.6, 34.1 and 39.8 % for UI, PW and WWD ( Table 2 and Figure 1L ).

The average vessel lumen area did not vary significantly between treatments, measuring 1112, 1083 and 1064 µm² respectively for UI, PW and WWD, nor did average vessel lumen diameter with quite similar values, i.e. 36.2, 36 and 35.5 µm ( Table 2 ). Total vessel density (N mm^-2) did differ between treatments, with WWD showing lower density than those of UI and PW with respectively 152, 172 and 173 vessels mm^-2. As a result, the total theoretical specific hydraulic conductivity K_S (per stem section) did vary significantly between treatments, with UI the highest, PW intermediate and WWD significantly much lower with respectively 10.79, 9.99 and 8.85 Kg. m^-1 Mpa^-1 s^-1.

For all treatments, vessel density per vessel lumen diameter range varied between the different diameter classes, i.e. 10-20, 20-30, 30-40, 40-50, 50-60 and 60-70 µm and showed a unimodal distribution, with class diameters 30-40 µm showing the highest density ( Figure 2A and Table S6 in the supplementary material). Similarly, K_S per vessel diameter range varied between the different diameter classes, with the highest K_S value for classes 30-40 µm, 40-50 µm and 50-60 µm ( Figure 2B and Table S7 in the supplementary material).

The frequency of vessel density per lumen diameter varied between the different classes, with the highest contribution to total vessel density for the group 30-40 µm, with respectively 34.7, 39.4 and 37.6% for UI, PW and WWD. Compared to PW and WWD, UI values for the 30-40 µm was significantly different. The groups 20-30 µm and 40-50 µm, which showed quite similar density frequencies, also contributed greatly to the total vessel density with 22.8, 21.4 and 25.0% respectively for UI, PW and WWD for the former group and 26.5, 27.5 and 22.7% for the latter group ( Table 3 and Figure 2C ). For K_S, the highest contribution was from the group 40-50 µm, with respectively 41.2, 45.1 and 38.1% for UI, PW and WWD ( Table 4 ; Figures 2C, D ).

The extractives content did vary significantly between the different irrigation treatments, with the content of WWD plants, 6.8%, noticeably lower than the content of UI and PW biomass, i.e. 10.5 and 10.7% ( Table 5 ).

Total lignin content as well as its acid soluble and acid insoluble fractions, i.e. ‘‘ASL’’ and ‘‘AIL’’, were statistically similar within the different irrigation treatments, with respectively 27.7, 27.1 and 28.1% for total lignin of WWD, UI and PW, 5.2, 5.1 and 5.3% for acid-soluble lignin and 22.5, 22.1 and 22.8% for acid insoluble lignin ( Table 5 ).

The cellulose content of glucose differed significantly between UI, PW and WWD, with respectively 33.9, 36.2 and 39.5% ( Table 5 ). Hemicellulose content did not differ between treatments UI, PW and WWD and was respectively 16.1, 17.3 and 17.7%. The content of the different hemicellulose sugar monomer components did not vary between treatments, although a statically significant difference in arabinose content between the NW and the WWD biomass was detected. Respectively, for UI, PW and WWD, the content of different monomers was 11.9, 12.6 and 13.3% for xylose, 1.3, 1.4 and 1.3% for galactose, 1.2, 1.3 and 1% for arabinose and 1.7, 2 and 2% for mannose ( Table 5 ).

Fertigation with primary wastewater effluent containing high concentrations of nitrogen significantly increased above ground biomass by 120% on average compared to the PW-irrigated plants. The 57% biomass increase compared the control UI was not significant ( Figure 1D ) (previously published results in Jerbi et al. (2020)). Similar growth conditions may have impacted plant growth as it was observed in young Salix nigra plants (Pezeshki et al., 1998), another flood tolerant willow species. In previous trials using the same plantation (Guidi Nissim et al., 2015; Jerbi et al., 2015), irrigation with different loads of secondary treated wastewater effluent (compared to the primary tested here) also led to an increase in biomass production over two years of growth. The maximum yield of 18.2 Mg ha^-1 yr^-1 then recorded (Guidi Nissim et al., 2015; Jerbi et al., 2015) signifies 63.2% less production of biomass compared to the present study [28.8 Mg ha^-1 yr^-1 over a season of growth (Jerbi et al., 2020)]. No such difference between the two trials was observed on control trees than were supplement with potable water only. Such difference in the yield of fertigated plants is likely due to the higher difference in the nitrogen loads within the two trials. Thus, depending on the nature (i.e. treated or untreated municipal or industrial), wastewater load and composition (nutrient content, EC, pH) the effect of fertigation on wood biomass may lead to drastic differences in terms of yield, xylem tissue structure and wood chemical composition (Pitre et al., 2007b) and therefore broaden the opportunities for the bioproduct sector.

Ranging from 0.32 to 0.39 g cm^-3 ( Table 2 and Figure 1H ), wood density was slightly lower in the present study than what was reported by Tharakan et al. (2003) for a group of 30 non-fertilized willow varieties (0.36 to 0.48 g cm^–3). This was especially true for WWD plants as could be expected as a result of N fertilization (Pitre et al., 2007a; Curran et al., 2008; Hacke et al., 2010).

Although we did not investigate fiber-cells, the low wood density of WWD plants compared to UI and PW suggests a likely alteration of fiber properties (Pitre et al., 2007a; Watanabe et al., 2008; Hacke et al., 2010; Zanne et al., 2010). Indeed, wood density is related to the morphology of the cells within the secondary xylem tissue i.e. the vessels, the ray parenchyma and the fibers (Poorter et al., 2010) with the latter having the thickest cell walls and thus contributing the most. High N fertilization was reported to reduce wood density by both increasing fiber lumen and decreasing fiber cell wall thickness (Luo et al., 2005; Pitre et al., 2007a; Pitre et al., 2010; Poorter et al., 2010). Thus, besides its consequence on vessel to fiber area and density ratio, irrigation with primary wastewater may have induced such alteration of fiber cells that renders the stems more flexible to leaning under gravitation and wind circumstances (compare Figures 1A–C ) and consequently more susceptible to the formation of tension wood. Indeed, similarly to the morphological alteration described by Pitre et al. (2007a) and Pitre et al. (2010) following N addition in poplar trees, we also observed stronger cellulose staining in the inner part of cell wall fibers of stem section of primary wastewater irrigated trees ( Figures 1M–O ), hence likely suggesting an increase of tension wood proportion within less rigid WWD treated plants.

Contrary to what was first hypothesized, no difference was observed between the different irrigation treatments regarding the average vessel diameter ( Table 2 ) even if it should be impacted by soil water status (Hacke et al., 2010; Plavcová and Hacke, 2012) and/or increased because of nitrogen fertilization (Bucci et al., 2006; Hacke et al., 2006; Hacke et al., 2010) and thus reflected by major changes of the theoretical specific conductance K_S (Hacke et al., 2017).

The control UI plants did not receive water other from precipitation (about 1200 mm over the two-year trial) but showed similar hydraulic parameters (i.e. average vessel diameter and K_S) as those of PW ( Table 2 ) which received high loads of potable water i.e. 3687 mm over the same period (Jerbi et al., 2020). Because water scarcity causes xylem to exhibit narrower but more frequent vessels, this may suggest that the water was not the limiting factor for willow growth and development as both treatments UI and PW showed very similar physiological parameters as well as comparable hydraulic parameters ( Table 2 ). In fact, the region where the experiment was set up is considered as a humid continental climate with generally high precipitation that may be more advantageous for a hydraulic architecture granting efficiency of water conductance (widest vessel lumen with less resistance) rather than to a ‘safe’ architecture to avoid cavitation and embolism caused by a drought that is unlikely to occur (Kotowska et al., 2015).

Intriguingly, the hydraulic conductance K_S was significantly lower for WWD plants than PW and UI ( Table 2 ) albeit various studies reported that high N fertilization increased the production of vessels with larger lumen (Harvey and van den Driessche, 1999; Hacke et al., 2010; Hacke et al., 2017). However, for most of the diameter classes, the density of vessels was lower for WWD fertilized plants than UI and PW, especially for the groups 30-40 µm and 40-50 µm ( Figure 2A and Table S6 in the supplementary material). These results could be somewhat explained by the sampling strategy we used. In an effort to compare stems at equivalent developmental stages (wood that already transitioned to secondary growth before the irrigation began), samples were harvested at different heights for each treatment even though stem conducts widen basipetally and that vessel lumen diameter increases axially from the top canopy towards the roots (Larson and Isebrands, 1971; Anfodillo et al., 2013; Hacke et al., 2017).

Regardless fertigation regime, the majority of vessels were found in the diameter groups 20-30 µm, 30-40 µm and 40-50 µm which represented ~ 86% of the total vessel and contributed to ~ 68% of the hydraulic conductance K_S ( Tables 3 , 4 ). This results is as expected for shrub species; with a higher proportion of narrow vessels than trees, i.e. ≤ 50 µm and practically no wide vessels i.e. > 200 µm (Wheeler et al., 2007; Anfodillo et al., 2013).

Difference between WWD plants from one hand and UI and PW from the other shows that fertigation with N-rich wastewater decreased the extractives content of WWD biomass ( Table 5 ). It is known that for a given species the composition of extractives is affected by growth conditions (Yang and Jaakkola, 2011). Fast-growing shrubs such as willows are generally associated with low bark to wood ratio. Because most of the extractives are generally located in the bark (Sassner et al., 2006; Serapiglia et al., 2009) such difference is likely to be associated with the higher growth rate of WWD plants and their lower bark content.

Several studies on willow cultivars have reported a compensatory relationship between cellulose and lignin synthesis (Guidi et al., 2009; Serapiglia et al., 2009; Ray et al., 2012; Serapiglia et al., 2013b). Although, WWD induced an increase of glucan content, lignin was similar between all treatments ( Table 5 ) and rather coherent to what was reported for other willow cultivars (Serapiglia et al., 2009; Ray et al., 2012; Serapiglia et al., 2013a). Furthermore, in a study comparing several willow genotypes grown at 45 degrees to induce the formation of reaction wood, Brereton et al. (2012) also reported that glucan content did increase for plants with the most reaction wood fraction while lignin remained unchanged. Hence, the similarity between the different irrigation treatment for total lignin may suggest that irrigation with primary wastewater and/or with potable water did not affect much lignin synthesis and deposition during secondary cell wall formation (i.e. lignification phase).

Alike, cellulose content ( Table 5 and Figure 1P ) was similar to what was reported for the same cultivar as well as for other genotypes (Sassner et al., 2006; Serapiglia et al., 2008; Serapiglia et al., 2009; Ray et al., 2012; Serapiglia et al., 2013b). The glucan content did differ between UI, PW and WWD with the higher content for plants that received wastewater ( Table 5 and Figure 1P ). High nitrogen application was previously reported to impact the development of secondary xylem of various poplar genotypes mostly by altering fiber anatomy through the deposition of an additional layer with high cellulose content in the inner part of the fiber cell lumen i.e. the G-layer (Pitre et al., 2007a; Pitre et al., 2007b; Pitre et al., 2010). Although quantitatively assaying tension wood is difficult (Brereton et al., 2011), the strong correlation ( Figure 3 , r² = 0.59, p = 0.0001) between glucan content (biochemical analysis) and the proportion of tension wood (based on numerical image analysis) suggests that the increase of cellulose content is likely due to the increase of the proportion of tension wood in the case of WWD treated plants.