Selection for Low-Nitrogen Tolerance Using Multi-Trait Genotype Ideotype Distance Index (MGIDI) in Poplar Varieties

Abstract

The screening of poplar varieties that demonstrate tolerance to low nitrogen (N) represents a promising strategy for improving nitrogen-use efficiency in trees. Such an approach could reduce reliance on N fertilizers while mitigating environmental pollution associated with their cultivation. In this study, a total of 87 poplar varieties were evaluated in a controlled greenhouse pot experiment. Under both low-nitrogen (LN) and normal-nitrogen (NN) conditions, 18 traits spanning four categories—growth performance, leaf morphology, chlorophyll fluorescence, and N isotope parameters were measured. For 13 of these traits (growth, leaf morphology, chlorophyll fluorescence), genetic variation and parameters, including genotypic values, were analyzed using best linear unbiased prediction (BLUP) within a linear mixed model (LMM). LN tolerance of tested poplar varieties was comprehensively assessed with three MGIDI strategies by integrating means, BLUPs, and low-nitrogen tolerance coefficient (LNindex) to rank poplar varieties. The results exhibited highly significant differences across all traits between LN and NN experiments, as well as among varieties. LN stress markedly inhibited growth, altered leaf morphology, and reduced chlorophyll fluorescence parameters in young poplar plants. Among the selection strategies, the MGIDI_LNindex approach demonstrated the highest selection differential percent (SD% = 10.5–35.23%). Using a selection intensity (SI) of 20%, we systematically identified 17 superior genotypes across all three strategies. In a thorough, comprehensive MGIDI-based evaluation, these varieties exhibited exceptional adaptability and stability under LN stress. The selected genotypes represent valuable genetic resources for developing improved poplar cultivars with enhanced low-nitrogen tolerance.

1. Introduction

The genus Populus L. is widely distributed throughout temperate and boreal regions of the Northern Hemisphere. Recognized as one of the fastest-growing temperate tree species, poplars hold substantial commercial value globally [1]. Their ecological and economic significance continues to increase, particularly in timber production, bioenergy feedstock, climate regulation, and ecosystem services [2]. Among the six taxonomically distinct sections within the genus, Aigeiros and Tacamahaca contain the most economically valuable species [3]. These sections demonstrate sexual compatibility, facilitating natural interspecific hybridization. The resulting heterosis has been effectively exploited in breeding programs to enhance economically important polygenic traits in poplar plantations, while allowing the evaluation of clonal variation in critical characteristics, including adaptive capacity, growth dynamics, wood properties, and stress resistance [4,5]. Despite the genus’ broad distribution, species exhibit distinct habitat preferences. For example, P. nigra (black poplar) thrives in moist environments, particularly in sunny locations with medium-to coarse-textured soils. This species of Sect. Tacamahaca achieves optimal growth in nutrient-rich, humus-containing, fertile soils that remain consistently moist yet well-drained [6]. Another notable hybrid, P. simonigra, derived from the crossing between P. simonii (Sect. Tacamahaca) and P. nigra var. italica (Sect. Aigeiros), exhibits exceptional tolerance and has been widely planted for shelterbelt and timber production in arid, cold regions such as Inner Mongolia, Northern China [7]. To maintain optimal productivity in poplar plantations, nitrogen (N) fertilization often becomes necessary. Consequently, N nutrition in poplars has emerged as an important research focus in recent years [8].

Nitrogen (N) is an essential nutrient for plant growth and a key limiting factor of tree growth and development in nutrient-poor soils [8,9,10,11,12]. Insufficient N availability severely impairs normal plant growth and physiological processes. Studies on poplar (Novaes et al. (2009) [13], Mamashita et al. (2015) [9], and Chen et al. (2022) [11]) have demonstrated that N deficiency restricts root, stem, and leaf growth, reduces chlorophyll synthesis, and triggers phenotypic plasticity under low nitrogen (LN) stress [9,11,13]. Typical N deficiency symptoms include stunted growth, reduced lateral branching, leaf chlorosis, and diminished biomass yield [12,14,15]. Globally, high N fertilizer inputs are used to maximize productivity. However, excessive N application not only imposes significant economic costs but also leads to severe environmental degradation, ultimately threatening human health. Paradoxically, N overuse can inhibit plant growth [16,17], increase management expense, and intensify ecological challenges [15,18]. Therefore, maintaining optimal N supply is crucial for sustainable plant development. To minimize resource waste, environmental pollution, and promote sustainable agriculture, improving nitrogen use efficiency (NUE) is imperative [19]. One of the promising solutions is to breed poplar varieties with enhanced N acquisition and utilization efficiency, enabling high yields under reduced N fertilization [5,11,12,14].

Low-nitrogen (LN) tolerance represents a complex quantitative trait governed by multiple genetic and physiological factors. While various parameters have been employed to screen for LN-tolerant genotypes, the inherent complexity of this trait—characterized by genetic polymorphism, genotype-environment interactions, and low heritability—poses significant challenges for reliable phenotyping. Consequently, no consensus exists regarding optimal screening indicators [13,19,20,21,22]. This underscores the critical need for an integrated evaluation framework that can comprehensively assess LN stress tolerance. The identification of LN-tolerant genotypes constitutes a fundamental objective in breeding programs aimed at developing stress-resistant cultivars [10,11,19,23,24]. However, selection efficacy depends critically on the choice of appropriate evaluation metrics [20,22,25]. In poplars, LN tolerance exhibits particularly intricate multidimensionality, where single-parameter assessments fail to capture the holistic stress response. Multi-trait evaluations, while necessary, inevitably introduce trait correlations [14,23,25,26,27]. As posited by Olivoto et al. (2019) [28], such multicollinearity can compromise analytical validity through biased regression coefficients and consequent erroneous interpretations [28,29].

Multiple multi-trait evaluation methods have been developed to identify stable ideotypes based on integrated variate analysis, facilitating the selection of genotypes suitable for diverse environmental conditions [27,30,31,32]. Olivoto and Nardino (2021) proposed the multi-trait genotype-ideotype distance index (MGIDI), an innovative and robust statistical approach that enables comprehensive genotype selection across multiple traits, including productivity, quality, physiology, and phenology [28,33,34]. This method evaluates genotypes by integrating multiple characteristics into a single quantitative index, ranking them according to their proximity to an ideal genotype [28,33,34]. Unlike conventional approaches that rely primarily on phenotypic data—which may not fully capture genetic potential [35,36]—MGIDI has emerged as a powerful breeding tool, enabling the selection of superior genotypes with optimal performance and predictable selection gains while simultaneously assessing their strengths and limitations [35,37]. The versatility of MGIDI is evidenced by its successful application across numerous crop improvement studies. As systematically reviewed by Debnath et al. (2024) following the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines, this method has significantly contributed to enhancing crop performance, quality, and adaptability [35]. Recent applications include screening for low-nitrogen and drought tolerance in various crops such as rice [38,39], wheat [26,27,36,40,41,42,43,44], maize [30,37], rapeseed [31], barley [32,45], tomato [46], and fenugreek [47]. Despite these widespread applications in herbaceous crops, the use of MGIDI for evaluating abiotic stress tolerance in woody plants remains limited, with only a few studies reported in poplar varieties [48] and hybrids [49].

This study investigated nitrogen use efficiency (NUE) in 87 poplar varieties, including hybrids from sections Tacamahaca and Aigeiros, to evaluate responses to low nitrogen (LN) stress. We hypothesized that significant genotypic variation exists in LN tolerance, with certain hybrids exhibiting both high NUE and sustained growth vigor under nitrogen limitation. By analyzing growth traits, leaf morphology, chlorophyll fluorescence parameters, and nitrogen isotope composition, we aimed to (1) assess phenotypic stability under LN stress, and (2) identify superior genotypes demonstrating both LN tolerance and growth vigor. The findings provide a theoretical basis for selecting poplar cultivars suited to nitrogen-limited environments.

2. Materials and Methods

2.1. Plant Materials

The study employed 87 poplar varieties as plant materials (Supplementary Table S1), comprising 45 from Section Tacamahaca, 17 from Section Aigeiros, and 25 interspecific hybrids of these sections. These varieties originated from five major administrative regions (Beijing, Shandong, Inner Mongolia, Heilongjiang, and Jilin) within China’s primary poplar cultivation zones, including (1) the Haihe Plain and Bohai Coastal region, (2) the Huanghuai Plain, (3) the Songliao Plain, and (4) the Songnen and Three Rivers Plain. Uniform stem cuttings (15 cm length × 1 cm diameter) were collected from one-year-old donor plants. These cuttings were individually planted in 9 cm × 18 cm pots under greenhouse conditions, resulting in a total of 522 experimental units (6 replicates per variety × 87 varieties).

2.2. Potting Experiment in Greenhouse

The pot experiment was conducted in the greenhouse at the Chinese Academy of Forestry, Beijing, under natural sunlight conditions without shading. Stem cuttings collected in February 2023 were potted in March using a 9:2 (v/v) peat-perlite substrate mixture. The pear medium (Type 422, Klasmann–Deilmann Co. Ltd., Geeste, Germany) had a pH value of 6.0 and contained no added nutrients. Two nitrogen treatments were established following modified protocols [11,50] on May 20–25, as below.

• Low nitrogen (LN): No nitrogen application
• Normal nitrogen (NN): Three-stage fertilization:
  • Day 30: ¹⁵N-labeled urea (¹⁵N-CH₄N₂O; 10.14% atomic abundance, 48% N content; Shanghai Institute of Chemical Technology, China) applied at 0.05 g/plant
  • Days 40 and 50: Conventional urea (CO(NH₂)₂; ≥46.0% total N; Kunlun Brand, China National Petroleum Co. Ltd., Daqing, China) applied at 1 g/plant per application (total 2 g/plant)

The experiment employed a completely randomized block design with three replicates per treatment (single plant per experimental unit).

2.3. Measurements of Traits

Plant morphological and physiological measurements were conducted following a modified standardized protocol [11]. Initial measurements of plant height (H₁) and basal diameter (BD₁) were recorded using a graduated ruler and digital caliper, respectively, prior to nitrogen application.

During peak growing season from August 10 to 20, leaf trait analyses were performed on mature leaves (5th to 7th node from the apex) of representative plants per variety. Leaf dimensions, including length (LL), width (LW), area (LA), and perimeter (PER), were quantified using a CI-203 laser leaf area meter (CID Bio-Science, Inc., Camas, WA, USA).

After that, chlorophyll fluorescence measurements were conducted during evening hours (20:00–23:00) under stable atmospheric conditions [10]. Using a FluorPen FP100 portable fluorometer (Photon Systems Instruments, FluorCam Co. Ltd, Bmo, Czech Republic), we recorded:

• Initial fluorescence (F₀)
• Maximum fluorescence (Fₘ)

Derived parameters were calculated as follows:

• Variable fluorescence (Fᵥ = Fₘ − F₀)
• PSII excitation energy capture efficiency (Fₘ/F₀)
• PSII potential activity (Fᵥ/F₀)
• Maximum quantum yield of PSII (Fᵥ/Fₘ)

Final measurements (H₂, BD₂) were taken at the end of the growing season (September 10), with net growth increments calculated as height increment (HS = H₂ − H₁) and basal diameter increment (BDS = BD₂ − BD₁).

Following the end of the growing season on September 10–20, an average sample plant from three blocks of each variety under both LN and NN treatments was selected for biomass and isotopic analysis. Immediately after harvest, stems were weighed to determine fresh weight (FW, g) and subsequently oven-dried at 75 °C to constant weight for dry weight (DW, g) measurement.

For nitrogen isotope analysis, NN-treated samples were processed as follows:

• Plant material was homogenized using a laboratory grinder
• Ground samples were sieved through a 60-mesh screen
• Processed samples were stored in airtight plastic bags prior to analysis

Isotopic measurements were conducted at Wako Jingxin Isotope Laboratory in Shenzhen, China using

• A Vario Macro Cube Elemental Analyzer coupled with
• A Delta V or MAT253 Continuous Flow Isotope Ratio Mass Spectrometer (IRMS)

The analytical procedure involved

• N₂ generation in the elemental analyzer’s redox tube at 1000 °C
• Gas separation via adsorption/desorption chromatography
• Isotopic ratio determination by IRMS

Measured parameters included

• Total nitrogen content (%)
• δ¹⁵N values (‰)
• ¹⁵N:¹⁴N ratios (R₁₅N/₁₄N)
• ¹⁵N atomic percentage (AT%)
• Absolute ¹⁵N abundance (mg ¹⁵N/g DW)

2.4. Data Analysis

In this study, we employed the low-nitrogen tolerance coefficient (LNindex) that effectively characterizes poplar varieties’ low-nitrogen (LN) adaptability [51]. The LNindex for each phenotypic trait was calculated as the ratio of its expression under LN condition to that under normal nitrogen (NN) condition, according to the following equation:

$$\mathrm{LNindex}={\displaystyle \frac{\mathrm{trait}\mathrm{value}\mathrm{under}\mathrm{LN}\mathrm{treatment}}{\mathrm{trait}\mathrm{value}\mathrm{under}\mathrm{NN}\mathrm{treatment}}}$$

(1)

The variability for 13 traits (growth, leaf morphology, chlorophyll fluorescence) was analyzed using a linear mixed model, with genotypic values predicted via restricted maximum likelihood/best linear unbiased prediction (REML/BLUP). The following linear mixed-effects model was employed following the methodology of Olivoto and Lúcio [52].

$$Y_{ijk}=\mu +{\alpha }_i+{\beta }_j+{\left(\alpha \beta \right)}_{ij}+{\gamma }_{k\left(j\right)}+{\epsilon }_{ijk}$$

(2)

where Y_ijk is the responsible variable observed in the kth block of the ith genotype in the jth environment (growth condition); µ is the grand mean; α_i and β_j are the effect of the ith genotype and the jth growth condition, respectively; (αβ)_ij is the genotype-by-environment interaction effect; Y_ijk is the effect of the kth block within the jth growth condition; and ε_ijk is the random error term. In this model, α_i and (αβ)_ij were treated as random effects, whereas β_j and γ_k(j) were considered fixed effects.

We analyzed low-nitrogen tolerance coefficient (LNindex) and genotypic values (BLUPs) of 13 traits for growth, leaf morphology, and chlorophyll fluorescence parameters under LN and LN conditions. These datasets, along with values of nitrogen isotope parameters, were subsequently used for Pearson correlation analysis, with results visualized through a heatmap and network plots. Genotype selection for low-nitrogen tolerance was conducted using the Multi-Trait Genotype-Ideotype Distance Index (MGIDI) approach [28].

In the context of the selection strategies under consideration, three distinct selection strategies were implemented in this study, as follows:

• Scenario 1: Integrated analysis incorporating low-nitrogen tolerance coefficient (LNindex) for 13 traits, and measured values of nitrogen isotope parameters.
• Scenario 2: Evaluation based on genotypic values (BLUPs) for 13 traits, and measured values of nitrogen isotope parameters.
• Scenario 3: Assessment using genotype means for 13 traits, and measured values of nitrogen isotope parameters.

For each scenario, MGIDI was calculated as the Euclidean distance between the factor scores of each genotype and the ideotype, following the methodology of Olivoto et al. (2022) [34] as follows:

$${MGIDI}_i={\left[\sum _{j=1}^f{\left({\mathsf{\gamma }}_{\mathrm{i}\mathrm{j}}-{\mathsf{\gamma }}_{\mathrm{j}}\right)}^2\right]}^{0.5}$$

(3)

where MGIDI_i is the MGIDI index for the ith genotype; γ_ij is the score of the ith genotype in the jth factor (i = 1, 2, …, g; j = 1, 2, …, f); g and f dente the genotypes and factor number, respectively; and γ_j is the jth score of the ideotype genotype. A genotype with a lower MGIDI is closer to the ideotype and, therefore, possesses the desired characteristics for all the traits examined.

Strengths and weaknesses of each genotype were assessed by quantifying its Euclidean distance from the ideotype across all evaluated factors. Genotypes exhibiting shorter distances to the ideotype for specific factors demonstrated a more favorable expression of the corresponding traits. To elucidate the multivariate relationships among traits and their genotype-specific associations, we performed principal component analysis (PCA), with genotypes as observational units and measured traits as variables.

Using the MGIDI values derived from three selection scenarios, we ranked all poplar varieties and identified superior genotypes by applying a consistent 20% selection intensity (SI) across all scenarios. For the selected genotypes, we calculated genetic gain (GG) using the following formula:

$$\mathrm{G}\mathrm{G}\mathrm{\%}={\displaystyle \frac{\left(X_s-X_0\right)}{X_0}}\times 100$$

(4)

where GG (%) is the genetic gain of selecting the best genotype for the trait from the population, X_s is the mean of the selected genotypes, and X₀ is the overall mean of all genotypes for the trait.

All statistical analyses and data visualization were performed utilizing the “metan” package v1.18.0 [52], in the software R Studio, R version 4.4.1. The variations between the treatment means were compared using the paired t-test and illustrated in boxplots with the ggplot2 v3.5.2 and ggpubr v0.6.0 packages. The network plots of the pairwise correlation data frame were constructed by the corrrplot v0.95 package.

3. Results

3.1. Mean Comparison Between LN and NN Experiment

Paired Student’s t-tests revealed highly significant differences (p < 0.01) in means between low-nitrogen (LN) and normal-nitrogen (NN) treatments across all measured traits and parameters (Figure 1, Figure 2 and Figure 3). For growth performances (Figure 1), leaf morphological characteristics (Figure 2), and chlorophyll fluorescence parameters (Figure 3), mean values under LN conditions were consistently and significantly lower than those under NN conditions. These results demonstrate that LN stress substantially impaired poplar growth performance, leaf development, and photosynthetic efficiency.

3.2. Mixed-Model Analysis

Table 1. Mixed-model analysis, estimated variance components, and genetic parameters of 87 varieties. Note: H₂, height in the end of growing season; HS, height increase; BD₂, base diameter in the end of growing season; BDS, base diameter increase; FW, fresh weight; DW, dry weight; LL, leaf length; LW, leaf width; LA, leaf area; PER, leaf perimeter; FmFo, efficiency of PSII in capturing excitation energy; Fv/Fm, maximum photosynthetic efficiency of PSII; Fv/Fo, potential PSII activity; GEIr², coefficient of determination of the interaction effects; h²_mg, average heritability; CV_g (%), genotypic coefficient of variation; CV_r (%), residual coefficient of variation.

Genetic Parameters	H2	BD2	HS	BDS	FW	DW	LA	LL	LW	PER	Fm/Fo	Fv/Fo	Fv/Fm
Phenotypic variance	119	1.02	27.2	0.254	3.24	0.848	67	2.98	1.34	428	0.0895	0.0895	0.000287
Heritability	0.62	0.487	0.268	0.133	0.611	0.624	0.676	0.74	0.724	0.723	0.0265	0.0265	0.0068
GEIr2	0.152	0.245	0.42	0.458	0.758	0.769	0.2	0.101	0.162	0.184	0.456	0.456	0.481
h2mg	0.867	0.771	0.532	0.335	0.871	0.877	0.86	0.921	0.889	0.879	0.0892	0.0892	0.0234
Accuracy	0.931	0.878	0.729	0.579	30.2	30.9	0.927	0.96	0.943	0.938	0.299	0.299	0.153
CVg (%)	13.8	12.6	12.3	12	1.25	1.29	29.9	28.5	18.8	32.2	1.06	1.36	0.18
CVr (%)	8.40	9.36	13.3	21.1	3.24	0.848	12.8	13.2	7.45	11.6	4.70	6.01	1.56

presents the mixed-model analysis for the 13 traits of growth, leaf morphology, and chlorophyll fluorescence under low-nitrogen (LN) and normal-nitrogen (NN) conditions. Mean heritability values (h²mg) ranged from 0.0234 (Fv/Fm) to 0.921 (leaf length, LL), with the heritability values ranging from 0.0068 (Fv/Fm) to 0.724 (leaf width, LW). The accuracy values ranged from 0.153 (Fv/Fm) to 0.96 (LL) and. The genotypic coefficients of variation (CVg) ranged from 0.18% (Fv/Fm) to 32.2% (leaf perimeter, PER), with more than 10% for all traits except fresh and dry weight (FW, DW), and chlorophyll fluorescence parameters.

Figure 4 presents the variance components estimated by BLUP for 11 traits under both low-nitrogen (LN) and normal-nitrogen (NN) conditions. The GEI (genotype × environment interaction) variance proportions ranged from 10.05% (leaf length, LL) to 48.14% (Fv/Fm). Notably, the three chlorophyll fluorescence parameters demonstrated higher variance components (45.6–48.14%) than genotypic variances (0.68–2.65%). The most substantial genetic variance was observed in leaf morphology traits, particularly leaf length (LL, 74.03%), followed by leaf width (LW, 72.43%) and leaf perimeter (PER, 72.27%). Interestingly, the three chlorophyll fluorescence parameters exhibited the highest GEI variance proportions (45.6–48.14%), followed by the three growth increase parameters (24.5–45.84%), while leaf morphological characteristics showed comparatively lower GEI variances (10.05–19.96%) (Figure 4).

3.3. Correlation Analysis

Pearson correlation analysis was performed between low-nitrogen tolerance coefficients (LNindex) and best linear unbiased predictors (BLUPs) of growth, leaf morphology, and chlorophyll fluorescence traits under low-nitrogen (LN) and normal-nitrogen (NN) conditions, as well as measured values of N isotope parameters under NN conditions (Supplementary Figure S1). The analysis revealed 147 statistically significant correlation pairs (p < 0.05, p < 0.01, p < 0.001) alongside 318 non-significant correlations. Among nitrogen isotope parameters, the strongest correlations (r = 1.000) were observed between δ15N and R¹⁵N/¹⁴N, δ15N and AT%¹⁵N, AT%¹⁵N and R¹⁵N/¹⁴N. Similarly, BLUPs of Fm/Fo (FmFo_BLUP) and Fv/Fo (FvFo_BLUP) exhibited a positive correlation (r = 1.00). The LNindex values for Fm/Fo (FmFo_LNindex), Fv/Fo (FvFo_LNindex), and Fv/Fm (FvFm_LNindex) showed strong intercorrelations (r = 0.987–0.999). Notably, highly significant (p < 0.001) correlations were detected between FW_LNindex and DW_LNindex (r = 0.978) as well as between FW_BLUP and DW_BLUP (r = 0.995).

Significant correlations were identified both within and between LNindex and BLUPs for most of the six growth traits (Supplementary Figure S1). In contrast, no significant associations were found between five nitrogen isotope parameters and LNindex or BLUPs of leaf morphology and chlorophyll fluorescence traits, except for LA_LNindex and LL_LNindex. Among leaf morphological traits, 14 correlation pairs reached statistical significance (p < 0.05 or p < 0.01), while 14 others were non-significant. The only exceptions were a weak negative correlation between LL_BLUP and LW_LNindex (r = −0.253, p < 0.05) and a weak positive correlation between LL_LNindex and PER_BLUP (r = 0.255, p < 0.05).

The correlation patterns among LNindex, BLUPs of growth, leaf morphology, and chlorophyll fluorescence traits, and measured values of nitrogen isotope parameters were visualized using a heatmap (Figure 5a) and a network plot (Figure 5b). Highly associated traits clustered together, connected by stronger paths, with blue and red lines indicating positive and negative correlations, respectively. The heatmap revealed distinct groupings of LNindex, BLUPs, and nitrogen isotope parameters, with further subclustering of growth, leaf morphology, and chlorophyll fluorescence traits, each exhibiting consistent correlation patterns. Network plot analysis, based on multidimensional clustering, demonstrated predominantly strong positive inter-trait relationships, though these did not fully align with the trends observed for LNindex or trait BLUPs.

3.4. Loadings and Factor Description for MGIDI

Principal component analysis (PCA) followed by exploratory factor analysis (EFA) identified five factors (eigenvalue > 1) that collectively explained 83.6% of total variability when analyzing trait means under normal-nitrogen (NN) conditions. In contrast, five factors accounted for 83% and 80.2% of variability, respectively, when analyzing low-nitrogen tolerance indices (LNindex), best linear unbiased predictors (BLUPs) of growth, leaf morphology, and chlorophyll fluorescence traits under low-nitrogen (LN) and NN conditions, and measured N isotope parameters under NN condition (

Table 2. Eigenvalues, explained variance, cumulative variance, and final loadings of factors retained after superposition by exploratory factor analysis.

Factors	Trait Means					LNindex					BLUPs
	FA1	FA2	FA3	FA4	FA5	FA1	FA2	FA3	FA4	FA5	FA1	FA2	FA3	FA4	FA5
H2	−0.891	0.0555	0.135	0.135	−0.0635	0.79	0.0277	0.159	0.0721	−0.416	−0.923	0.0452	−0.106	0.112	0.0562
HS	−0.508	−0.198	−0.234	0.26	0.128	0.547	−0.148	0.211	0.294	−0.0541	−0.58	−0.27	0.161	0.252	0.0251
BD2	−0.855	0.041	0.192	0.186	−0.063	0.793	0.172	0.00614	−0.222	−0.0881	−0.872	0.134	−0.126	0.255	0.00138
BDS	−0.73	0.0392	−0.0489	−0.185	−0.0526	0.825	0.00354	0.0415	−0.0802	0.0908	−0.778	0.0638	0.1	−0.0507	0.0779
FW	−0.889	0.116	0.168	0.107	−0.0675	−0.0991	0.0093	−0.991	0.0184	−0.0444	−0.15	−0.0287	−0.955	−0.0737	−0.0883
DW	−0.89	0.0989	0.181	0.0909	−0.0728	−0.0913	0.00377	−0.989	0.00385	−0.0348	−0.00353	0.00489	−0.468	0.0865	0.234
LA	−0.0387	−0.0909	−0.00562	0.813	−0.0271	0.245	−0.0618	−0.00959	−0.152	−0.884	−0.107	−0.132	0.0725	0.8	−0.0243
LL	−0.0803	−0.0549	−0.0437	0.843	−0.118	0.246	0.153	−0.014	−0.289	−0.618	−0.0826	0.000572	0.0244	0.864	0.0791
LW	−0.0444	0.116	0.0144	0.837	−0.00414	0.013	0.0244	−0.0787	0.0505	−0.798	−0.123	0.134	−0.136	0.82	0.00228
PER	−0.17	−0.0962	0.154	0.756	0.122	0.195	−0.134	−0.0463	0.05	−0.875	−0.22	−0.159	0.0543	0.715	−0.0969
Fm/Fo	−0.124	−0.0201	0.987	0.0299	0.0192	0.752	0.139	0.0398	−0.0416	−0.409	−0.858	0.174	−0.25	0.149	0.0177
Fv/Fo	−0.125	−0.0198	0.987	0.031	0.0192	0.741	0.129	0.0419	−0.0311	−0.44	−0.866	0.156	−0.262	0.129	0.0259
Fv/Fm	−0.124	−0.0274	0.985	0.0392	0.0235	−0.0939	0.00922	−0.99	0.0171	−0.0503	−0.15	−0.0287	−0.955	−0.0737	−0.0883
δ15N	0.0612	−0.995	0.0184	0.0441	−0.0157	−0.0746	−0.992	0.00774	0.0434	0.00082	0.0943	−0.989	−0.0158	0.0493	0.0171
N content	0.0867	0.373	0.0325	−0.0292	0.916	0.0907	0.314	0.0353	−0.898	−0.103	0.0521	0.369	0.0309	0.0242	−0.912
R15N:14N	0.0605	−0.995	0.0193	0.0428	−0.0168	−0.075	−0.992	0.00649	0.0443	0.00092	0.0936	−0.989	−0.0175	0.0476	0.0179
AT%15N	0.0615	−0.995	0.0189	0.0452	−0.0149	−0.0737	−0.992	0.00665	0.0424	0.0025	0.0946	−0.989	−0.0169	0.05	0.016
15N abundance	0.159	−0.516	0.04	−0.0136	0.827	0.00723	−0.569	0.0455	−0.789	−0.104	0.159	−0.513	0.0261	0.0427	−0.823
Eigenvalues	4.86	3.60	2.79	2.30	1.50	5.15	3.46	3.14	1.71	1.48	5.11	3.58	2.43	1.81	1.50
Variance (%)	27	20	15.5	12.8	8.31	28.6	19.2	17.4	9.49	8.23	28.4	19.9	13.5	10	8.32
Accumulated (%)	27	47	62.5	75.3	83.6	28.6	47.8	65.3	74.7	83	28.4	48.3	61.8	71.8	80.2

FA1, factor 1; FA2, factor 2; FA3, factor 3; FA4, factor 4; FA5, factor 5; H₂, height at end of growing season; HS, height increase; BD₂, base diameter in the end of growing season; BDS, base diameter increase; FW, fresh weight; DW, dry weight; LL, leaf length; LW, leaf width; LA, leaf area; PER, leaf perimeter; Fm/Fo, efficiency of PSII in capturing excitation energy; Fv/Fm, maximum photosynthetic efficiency of PSII; Fv/Fo, potential PSII activity; N content, elemental N content (%); δ¹⁵N, δ¹⁵N value (‰); R¹⁵N:¹⁴N, ¹⁵N:¹⁴N ratios; AT%¹⁵N, atom percent (AT%) of ¹⁵N (%); ¹⁵N abundance, absolute abundance of ¹⁵N.

).

As the result of PCA and EFA with trait means (NN condition), six growth traits (H₂, HS, BD₂, BDS, FW, and DW) belonged to FA1; three N isotope parameters (δ¹⁵N, R¹⁵N:¹⁴N, and AT%¹⁵N) were included in FA2; three chlorophyll fluorescence parameters (Fm/Fo, Fv/Fo, and Fv/Fm) belonged to FA3; four leaf morphological traits (LA, LL, LW, and PER) belonged to FA4, and two additional N isotope parameters (N content and ¹⁵N abundance) were included in FA5.

Among the five factors retained from PCA and EFA with LNindex (LN and NN conditions) and N isotope parameters (NN condition), FA1 included six growth traits; FA2 included three N isotope parameters; FA3 included three chlorophyll fluorescence parameters; FA4 included two additional N isotope parameters, and FA5 included four leaf morphological traits.

Similarly, based on the BLUPs (LN and NN conditions), and measured value of N isotope parameters (NN condition), FA1 included two growth traits and two parameters of chlorophyll fluorescence parameters; FA2 included three N isotope parameters; FA3 included two biomass traits (FW, DW) and one chlorophyll fluorescence parameter; F4 included four leaf morphological traits; and FA5 included two additional N isotope parameters.

3.5. Genotype Ranking Based on MGIDI Analysis with Means, LNindex, and BLUPs

Figure 6 presents genotype rankings determined through three Multi-Trait Genotype-Ideotype Distance Index (MGIDI) strategies: (1) trait means (Figure 6a), (2) low-nitrogen tolerance indices (LNindex; Figure 6b), and (3) best linear unbiased predictors (BLUP; Figure 6c) for growth, leaf morphology, and chlorophyll fluorescence traits under low-nitrogen (LN) and normal-nitrogen (NN) conditions, along with nitrogen isotope parameters under NN conditions.

Genotypes with lower MGIDI values were considered closer to the ideotype, exhibiting more desirable trait combinations. Using a selection intensity of 20%, we identified 17 top-performing genotypes under each strategy: (1) MGIDI_means strategy (Figure 6a): G7, G73, G13, G62, G8, G67, G81, G58, G48, G3, G78, G54, G71, G5, G72, G53, and XY-1; (2) MGIDI_LNindex strategy (Figure 6b): G61, G1, G73, G82, G24, G88, X1, G10, G35, G21, G67, G5, N1, G64, G16, G81, and G7; (3) MGIDI_BLUP strategy (Figure 6c): G73, G67, G7, G81, G13, G8, G82, G58, G10, G5, G78, G3, G62, G79, G48, G53, G88.

The genotype screening results from three MGIDI strategies were comparatively analyzed using a Venn diagram (Figure 7). Our study revealed that five genotypes (G7, G73, G67, G81, and G5) were consistently identified as superior across all three selection strategies. Meanwhile, eight genotypes (G13, G62, G8, G58, G48, G3, G78, and G53) were co-selected by both MGIDI_mean and MGIDI_BLUP. Three genotypes (G82, G88, and G10) were uniquely identified by both MGIDI_LNindex and MGIDI_BLUP. No overlapping selections were observed between MGIDI_mean and MGIDI_LNindex strategies. Furthermore, MGIDI_mean exclusively selected four genotypes (G54, G71, G72, and XY-1), while MGIDI_LNindex identified nine genotypes (G61, G1, G24, X1, G35, G21, N1, G64, and G16). MGIDI_BLUP singularly recognized one genotype (G79) as excellent.

All three MGIDI selection strategies demonstrated statistically significant selection differentials (SDs) for the target traits (

Table 3. Selection gain achieved with three MGIDI scenarios for 87 poplar genotypes under two nitrogen experiments.

Scenarios	Traits	Factor	X0	Xs	SD	SD%
MGIDI_mean	H2_mean	FA1	66.1	73.5	7.40	11.2
	HS_mean	FA1	26.1	28.7	2.55	9.76
	BD2_mean	FA1	5.90	6.57	0.672	11.4
	BDS_mean	FA1	1.85	1.93	0.0839	4.53
	FW_mean	FA1	1.72	2.02	0.309	18
	DW_mean	FA1	0.865	1.03	0.168	19.5
	δ15N	FA2	2.54	2.98	0.435	17.1
	R15N:14N	FA2	1.30	1.46	0.16	12.3
	AT%15N	FA2	1.28	1.44	0.156	12.2
	Fm/Fo_mean	FA3	5.18	5.28	0.104	2.02
	Fv/Fo_mean	FA3	4.18	4.28	0.105	2.50
	Fv/Fm_mean	FA3	0.807	0.811	0.00402	0.499
	LA_mean	FA4	24.6	28	3.32	13.5
	LL_mean	FA4	5.51	6.33	0.82	14.9
	LW_mean	FA4	5.39	6.02	0.628	11.7
	PER_mean	FA4	59.4	71.9	12.5	21
	N content	FA5	1.89	1.94	0.0483	2.55
	15N abundance	FA5	0.239	0.273	0.0345	14.4
MGIDI_LNindex	H2_LNindex	FA1	0.877	0.97	0.0932	10.6
	HS_LNindex	FA1	0.685	0.771	0.0852	12.4
	BD2_LNindex	FA1	0.905	1.01	0.106	11.7
	BDS_LNindex	FA1	0.683	0.856	0.174	25.4
	FW_LNindex	FA1	0.843	1.13	0.291	34.5
	DW_LNindex	FA1	0.85	1.13	0.278	32.7
	δ15N	FA2	2.54	2.91	0.37	14.6
	R15N:14N	FA2	1.30	1.44	0.136	10.4
	AT%15N	FA2	1.28	1.42	0.133	10.3
	Fm/Fo_LNindex	FA3	0.771	0.788	0.0168	2.18
	Fv/Fo_LNindex	FA3	0.716	0.736	0.0201	2.81
	Fv/Fm_LNindex	FA3	0.927	0.933	0.00616	0.665
	N content	FA4	1.89	1.96	0.07	3.70
	15N abundance	FA4	0.239	0.273	0.0345	14.4
	LA_LNindex	FA5	0.853	1.07	0.213	25
	LL_LNindex	FA5	0.898	0.994	0.0964	10.7
	LW_LNindex	FA5	0.964	1.04	0.0713	7.39
	PER_LNindex	FA5	0.855	1.03	0.171	20
MGIDI_BLUP	H2_BLUP	FA1	62	68	6.01	9.69
	HS_BLUP	FA1	21.9	23	1.07	4.90
	BD2_BLUP	FA1	5.59	6.05	0.461	8.24
	BDS_BLUP	FA1	1.53	1.56	0.0299	1.96
	FW_BLUP	FA1	4.66	5.59	0.924	19.8
	DW_BLUP	FA1	2.36	2.85	0.493	20.9
	δ15N	FA2	2.54	2.95	0.409	16.1
	R15N:14N	FA2	1.30	1.45	0.15	11.6
	AT%15N	FA2	1.28	1.43	0.147	11.4
	Fm/Fo_BLUP	FA3	4.58	4.59	0.00667	0.145
	Fv/Fo_BLUP	FA3	3.58	3.59	0.00667	0.186
	Fv/Fm_BLUP	FA3	0.777	0.777	0.000023	0.00296
	LA_BLUP	FA4	22.5	25	2.46	10.9
	LL_BLUP	FA4	5.21	5.91	0.7	13.4
	LW_BLUP	FA4	5.25	5.78	0.533	10.1
	PER_BLUP	FA4	54.7	66.9	12.3	22.4
	N content	FA5	1.89	1.95	0.0595	3.14
	15N abundance	FA5	0.239	0.273	0.0339	14.2

FA1, factor 1; FA2, factor 2; FA3, factor 3; FA4, factor 4; FA5, factor 5; H₂, height in the end of growing season; BD₂, base diameter in the end of growing season; HS, height increase; BDS, base diameter increase; FW, fresh weight; DW, dry weight; LL, leaf length; LW, leaf width; LA, leaf area; PER, leaf perimeter; Fm/Fo, efficiency of PSII in capturing excitation energy; Fv/Fm, maximum photosynthetic efficiency of PSII; Fv/Fo, potential PSII activity; N content, elemental N content (%); δ¹⁵N, δ¹⁵N value (‰); R¹⁵N:¹⁴N, ¹⁵N:¹⁴N ratios; AT%¹⁵N, atom percent (AT%) of ¹⁵N (%); ¹⁵N abundance, absolute abundance of ¹⁵N; SD, selection differential; SD%, percent selection differential; X₀, mean of the original population; Xs, mean of the selected varieties.

), though their effectiveness varied substantially across trait categories. For strategy-specific performance, the MGIDI_mean approach yielded the most substantial improvement in the mean of dry weight (19.5% SD for DW_mean), MGIDI_LNindex showed optimal performance for low-nitrogen tolerance coefficient of fresh weight (22.6% SD for FW_LNindex), and MGIDI_BLUP was most effective for BLUPs of leaf morphological traits (22.4% SD for PER_BLUP). For trait-dependent response patterns, chlorophyll fluorescence parameters (Fm/Fo, Fv/Fo, and Fv/Fm) consistently exhibited minimal response to selection (the lowest SDs across all strategies), while biomass-related traits (fresh weight and dry weight) demonstrated the strongest selection responses (the highest SDs) regardless of selection strategy.

3.6. Strengths and Weaknesses View of Selected Genotypes

The radar plots (Figure 8a–c) illustrated the strengths and weaknesses of genotypes selected through three MGIDI strategies. For each genotype, factor contributions to MGIDI were radially ranked from most influential (near plot center) to least influential (toward plot periphery). Factors positioned closer to the outer edge indicated traits more closely approximating the ideotype.

Analysis using the MGIDI_mean strategy (Figure 8a) revealed distinct performance patterns among genotypes. G78, G58, and G81 exhibited strengths in FA1, comprising six growth traits (H₂, HS, BD₂, BDS, FW, and DW). Conversely, G73, G62, G53, G3, and G81 showed strong performance in FA2, containing three N isotope parameters (δ¹⁵N, R¹⁵N:¹⁴N, and AT%15N). For FA3, genotypes G8, G7, and G5 demonstrated notable performance, while G54 and G73 excelled in FA5, which includes two additional N isotope parameters (N content and ¹⁵N abundance).

The MGIDI_LNindex strategy (Figure 8b) showed that most genotypes primarily contributed to MGIDI through growth-related FA1, except G61 and G35. Specific strengths were observed for G81, G73, G35, and X1 in FA2 (δ15N, R¹⁵N:¹⁴N, AT%¹⁵N), G10, G67, and N1 in FA3 (chlorophyll fluorescence), and G64 (followed by G67 and G82) in FA4 (additional N isotope parameters). Notably, only G1 demonstrated strength in FA5, containing four leaf morphological traits.

Similarly, the MGIDI_BLUP strategy (Figure 8c) indicated most selected genotypes exhibited strengths across all factors except FA5. Within FA1, G78 and G79 showed excellent performance, while G3 performed poorly with maximal MGIDI contribution. All 17 selected genotypes consistently demonstrated strengths in FA2 (N isotope parameters) and FA3 (chlorophyll fluorescence), coupled with weaknesses in FA5 (leaf morphology).

4. Discussion

This study represents a comprehensive evaluation of low nitrogen (LN) tolerance in 87 poplar varieties using a novel multi-trait selection approach. The integration of the MGIDI with BLUP and LNindex provides a robust framework for identifying superior genotypes under N limitation. The findings demonstrated that LN stress significantly suppresses growth, leaf development, and photosynthetic efficiency in poplar seedlings, consistent with previous reports in poplar [11,13] and other woody species [8]. The highly significant differences (p < 0.01) observed between the LN and NN treatments across all measured traits (Figure 1, Figure 2 and Figure 3) validate the effectiveness of our experimental design in inducing nitrogen stress and highlight the phenotypic plasticity of poplar under nutrient limitation.

4.1. Methodological Advancements in Trait Integration

The MGIDI approach pioneered by Olivoto and colleagues [28,33,34,52] proved exceptionally valuable in addressing the complexity of LN tolerance evaluation. By transforming multidimensional trait data into a single selection index, MGIDI effectively circumvented the multicollinearity issues inherent in traditional multi-trait selection [27,29]. The present study extends the application of MGIDI beyond annual crops (as reviewed by Debnath et al., 2024) [35] to perennial tree species, thereby demonstrating its versatility. The integration of BLUP-derived genotypic values further enhanced the analysis by accounting for genotype-by-environment interactions (G × E), which is particularly crucial given the significant G × E effects observed for several traits (Table 1). This combined approach minimized the environmental noise common in phenotypic selection for tree breeding [36,53].

The LNindex emerged as a particularly powerful metric, with the MGIDI_LNindex strategy achieving the highest percent selection differential (10.5–35.23%) across traits (Table 3). This finding is consistent with the conclusions of Liu et al. (2020) [51], who proposed the utilization of relative performance measures for the evaluation of stress tolerance. The LNindex effectively normalized genotypic responses to nitrogen availability, enabling more accurate identification of genotypes with inherent LN tolerance rather than those merely exhibiting superior performance under optimal conditions. The differential responses observed among the three selection strategies (MGIDI_mean, MGIDI_LNindex, and MGIDI_BLUP) underscore how selection criteria shape breeding success in target environments. The strategic approaches have yielded novel insights into genotype screening that can be directly applied to deepen our understanding of LN tolerance mechanisms in poplar.

4.2. Trait Relationships and Biological Implications

The correlation analyses revealed complex relationships among traits (Supplementary Figure S1 and Figure 5). The near-perfect correlations (r ≈ 1.00) among N isotope parameters (δ¹⁵N, R¹⁵N/¹⁴N, AT%¹⁵N) were anticipated in view of their mathematical interdependence. More importantly, the strong correlations between LNindex and BLUP of fresh and dry weights (FW_LNindex/DW_LNindex: r = 0.978; FW_BLUP/DW_BLUP: r = 0.995) suggest biomass allocation as a key indicator of LN tolerance. Similarly, the tight integration among chlorophyll fluorescence parameters (Fm/Fo, Fv/Fo, Fv/Fm) indicates coordinated photosynthetic responses to N limitation, consistent with Mamashita et al. (2015) and Zhang et al. (2018), who reported similar relationships in nitrogen-stressed poplar [9,54].

Notably, the limited correlations between N isotope parameters and LNindex or BLUP of other trait categories (with the exception of LA_LNindex and LL_LNindex) suggested that N assimilation dynamics operate somewhat independently from growth and photosynthetic traits under LN conditions. This finding has important implications for breeding, as it suggests that simultaneous selection for efficient N uptake/utilization and stress-tolerant growth patterns may require independent screening approaches. The factor analysis (Table 2) further supported this hypothesis, with growth traits consistently loading on Factor 1 (FA1), N isotopes on FA2, chlorophyll fluorescence on FA3, and leaf morphology on FA4 across different selection strategies.

4.3. Heritability and Genetic Architecture

The broad range of heritability estimates (h²: 0.0068–0.724) revealed the differential genetic control of traits under nitrogen stress (Table 1). Growth traits generally exhibited moderate to high heritability (e.g., LL: h² = 0.74; LW: h² = 0.724), thus supporting their utility in selection programs. Conversely, chlorophyll fluorescence parameters exhibited remarkably low heritability (Fv/Fm: h² = 0.0068), suggesting strong environmental influences on these traits. This finding is consistent with the observations made by Kalcsits et al. (2016), who emphasized that physiological traits with low heritability may be less reliable for selection unless measured with a high degree of precision [24]. The high accuracy values for growth and leaf morphology traits (up to 0.96 for LL) validate the precision of our phenotyping protocols, while the lower accuracy for chlorophyll fluorescence parameters (as low as 0.153 for Fv/Fm) indicates measurement challenges that warrant methodological refinement in forthcoming studies.

The genotypic coefficients of variation (CVg) revealed substantial genetic diversity for most traits (PER: 32.2%; LA: 29.9%), particularly in leaf morphology and growth characteristics. This extensive variation within the germplasm provides a strong foundation for selection progress. Interestingly, the residual coefficients of variation (CVr) were generally higher for chlorophyll fluorescence parameters, thereby reinforcing their heightened sensitivity to microenvironmental fluctuations.

4.4. Validation of Selected Genotypes

The convergence of five genotypes (G7, G73, G67, G81, and G5) across all three selection strategies (Figure 6) provides strong evidence of their superior and stable LN tolerance. These genotypes represent prime candidates for advanced testing in field conditions. As demonstrated in Figure 8, the radar plots further revealed the complementary strengths of the poplar varieties under investigation. G73 excelled in N isotope parameters (FA2), G81 demonstrated balanced performance across multiple factors, while G5 showed particular strength in chlorophyll fluorescence maintenance (FA3). This diversity in adaptive strategies suggests that the presence of different genetic mechanisms that may underlie LN tolerance, thus offering multiple pathways for breeding programs.

The MGIDI_LNindex strategy identified nine unique genotypes (G61, G1, G24, X1, G35, G21, N1, G64, G16) that had not been prioritized by alternative approaches. The performance of these genotypes was found to be exceptional in the context of N stress, as demonstrated by the elevated LNindex values for growth traits (FW_LNindex: 34.5%; DW_LNindex: 32.7%) and leaf morphology (PER_LNindex: 20%). It is evident that their specialized adaptation makes them a valuable genetic resource for marginal lands that are subject to chronic N limitation.

The factor contribution analysis (Figure 8) provided insights into genotype-specific strengths that were previously unattainable. For instance, G78 (selected via MGIDI_mean) demonstrated exceptional growth maintenance (FA1) yet exhibited weakness in N isotope parameters (FA2). Conversely, G81 exhibited balanced contributions across factors. This detailed understanding enables breeders to make strategic selections based on specific breeding objectives, such as prioritizing growth under stress, N use efficiency, or photosynthetic stability.

4.5. Practical Applications and Breeding Implications

The 20% selection intensity yielded 17–18 genotypes per strategy, a practical number for advanced testing. The substantial genetic gains achieved, particularly through MGIDI_LNindex (up to 35.23% for FW_LNindex), demonstrate the efficacy of this approach for rapid genetic improvement. The findings of this study support the hypothesis that MGIDI is a superior alternative to conventional selection indices for complex traits such as LN tolerance. This conclusion is supported by the observations of Zhao et al. (2019) and Ararisa et al. (2024), who noted limitations in the use of conventional indices [20,22].

From a physiological standpoint, the synchronized decrease in growth, leaf development, and photosynthetic parameters under LN stress (Figure 1, Figure 2 and Figure 3) lends further support to the integrated nature of nitrogen-responsive traits. The strong performance exhibited by hybrids (25 in the study) lends support to the exploitation of heterosis for the purpose of enhancing tolerance to LN, as proposed by Ren et al. (2020) and Novaes et al. (2009) [5,13]. The geographic diversity of the germplasm (across five Chinese provinces) enhances the environmental relevance of selected genotypes for major poplar-growing regions.

4.6. Limitations and Future Research

Despite the considerable advances represented by this study, it would be remiss to ignore the limitations that require attention. First, it must be noted that the greenhouse potting experiment may not fully replicate field conditions, particularly with regard to soil N dynamics and mycorrhizal associations [48,49,55]. Secondly, the focus of our evaluation was on juvenile growth over the course of a single growing season. The long-term performance of the subject under recurrent nitrogen stress remains untested. Thirdly, the exclusion of root architecture traits represents a significant gap, given their importance in the acquisition of nutrients [10]. Fourthly, the molecular mechanisms underlying the superior performance of selected genotypes remain unexplored [56].

Future research should (1) validate selected genotypes in multi-location field trials; (2) incorporate root phenotyping and rhizosphere microbiome analyses; (3) investigate transcriptomic and metabolic bases of LN tolerance; (4) extend MGIDI application to other abiotic stresses and tree species; (5) develop genomic selection models integrating MGIDI outputs.

5. Conclusions

This study establishes an integrated MGIDI-BLUP-LNindex framework for evaluating low-nitrogen (LN) tolerance in poplar, yielding key insights. LN stress significantly impairs growth, leaf development, and photosynthesis (p < 0.01), with marked trait reductions compared to normal N conditions. Combining MGIDI (multi-trait integration), BLUP (G × E adjustment), and LNindex (stress tolerance) provides an efficient selection system for complex traits, overcoming multicollinearity issues in conventional breeding. Growth and chlorophyll fluorescence traits are strongly linked but show minimal association with N isotope dynamics, suggesting distinct genetic regulation of N assimilation and stress response. High heritability in growth (LL h² = 0.74) and leaf morphology (PER CVg = 32.2%) supports strong breeding potential, while low heritability in photosynthetic traits warrants indirect selection. Five genotypes exhibit consistent LN tolerance, while nine others show specialized adaptation under MGIDI_LNindex selection, offering valuable genetic resources. The MGIDI_LNindex approach delivers the highest genetic gain (35.23%), confirming its superiority for LN environments. Complementary genotype sets identified across strategies enable trait-specific breeding. This framework enables the selection of N-efficient poplar varieties, reducing fertilizer dependence while maintaining yield—promoting sustainable forestry, and minimizing N pollution. By merging multivariate statistics with physiological phenotyping, this work advances perennial tree breeding, providing a scalable model for complex trait selection. Future research should validate field performance and elucidate molecular mechanisms of LN tolerance.