Eight large-seeded Andean dry bean genotypes (Table 1) which represent the parents of five recombinant inbred populations available for advanced genetic studies were tested in a pot study for P efficiency. The G122 (ADP-610) × Cardinal (ADP-643) and G122 × Taylor Horticulture (Thort) populations were used to map Beet curly top virus (BCTV) resistance [15] [16] . The ADP indicates an accession within the Andean Diversity Panel (Cichy et al., 2015) which is described in detail on the website (http://arsftfbean.uprm.edu). G122 is a red mottled small cranberry seed type landrace from India with heat tolerance and partial resistance to BCTV. Cardinal is a large seeded cranberry bean cultivar with wide adaptation in the US and with major resistance to BCTV [15] . Taylor Horticulture is an old US cultivar with early maturity but is susceptible to most viruses. The G122 × Montcalm (ADP-636) population was developed for genetic analysis of heat tolerance [17] . Montcalm is a dark red kidney bean cultivar from Michigan State University that is most productive under high inputs. Bukoba (ADP-007) is a yellow seeded “Njano” landrace from Tanzania that is susceptible to terminal drought. Kijivu (ADP-033) is a purpled speckled “Kablanketi” landrace from Tanzania that exhibits drought tolerance. The Rojo (ADP-096) × CAL 143 (ADP-526) population [18] was developed to map resistance to multiple diseases in Eastern and Southern Africa including bean rust (Uromyces appendiculatus), angular leaf spot (Phaeoisariopsis griseola (Sacc.) Ferr.), and

Description of ADP# can be found at http://arsftfbean.uprm.edu/bean/?page_id=179.

halo bacterial blight (Pseudomonas syringae pv. phaseolicola). Rojo is a dark red cultivar from Sokoine University of Agriculture (SUA) in Tanzania and CAL 143 is a red mottled “Calima” breeding line from CIAT (Center for Tropical Agriculture) that has been released as Kalima in Malawi and Zambia.

The experiments were conducted in the USDA-ARS glasshouse at Prosser, Washington. Sunshine mix (American Horticultural Supply, Inc., Oxnard, CA, USA), used as the growing media, was analyzed for P content prior to the experiment to determine suitability for the study. The P content was 0.40 g kg⁻¹, based on the weak acids P extraction method used (Bray-1 designed for non-calcareous soils). This value is low and determined to be in an inorganic form unavailable for plant use [19] . The clay pots (20 cm diameter by 19 cm depth) were filled with sunshine mix, different MAP rates were incorporated in the sunshine mix of individual pots prior to planting. The three mono-ammonium phosphate (MAP) fertilizer rates were 0 kg P ha⁻¹ (0 ppm), 50 kg P ha⁻¹ (22.5 ppm) and 100 kg P ha⁻¹ (45 ppm).

Three bean seeds of an individual genotype were planted per pot and later thinned to a single plant per pot about 10 days after planting. Watering was done daily but no additional fertilizers were applied for the duration of the experiment. The day temperature varied between 25˚C - 28˚C and night temperature was 23˚C, with a 14 h day light provided by supplementing natural light with artificial lights.

The experimental design was a RCBD with four replications. The first experiment, with 96 pots (8 genotypes × 3 P levels × 4 reps) was harvested at flowering maturity and revealed genotypic differences for different traits. It might be interesting to evaluate shoot, root and seed P as well. Thus, a second experiment was conducted and expanded to 192 pots to include two harvests one at flowering maturity and another at plant maturity (8 genotypes × 3 P levels × 4 replications × two harvests). This second experiment with the two harvest dates was conducted twice to confirm the findings. So, in effect one partial (Experiment 1) and two full experiments (Experiments 2 and 3) were conducted.

The plants from all the pots in experiment 1 and half the pots in experiments 2 and 3 were harvested at flowering maturity. The plant shoots and roots were collected separately, oven dried for 48 h at 75˚C, and then weighed for biomass (g) on an individual plant basis. The root samples were separated from the growing media by rinsing them with tap water and then patting them dry with paper towels. The dry seed was harvested from the remaining pots in experiments 2 and 3 at harvest maturity.

The dried shoot, root and seed samples were ground using a Wiley mill (Thomas Scientific; Swedesboro, NJ, USA) to pass a 1 mm sieve. The internal P concentration (ppm) of the ground tissues (0.5 g) was determined by inductively coupled plasma mass spectrometry (ICP-MS) using analysis of a nitric acid digest [20] . Shoot P (mg) was calculated as the product of the shoot biomass (g) and shoot P (mg g⁻¹) on a per plant basis. The P susceptibility index (PSI) was measured using {PSI = 1 − [Y₁/Y]/1 − [X₁/X]}; where, Y₁ = mean shoot biomass of the low P treatment for genotype₁; Y = mean shoot biomass of the low P treatment for all genotypes; X₁ = mean shoot biomass of the normal P treatment for genotype₁; and X = mean shoot biomass of the normal P treatment for all genotypes. Genotype P-use efficiency (PUE) was calculated as biomass yield divided by tissue P concentration at a given P rate in the growth media [21] [22] , given by; PUE = Y₁/P₁, where, Y₁ = Biomass yield on P level₁ (g) and P₁ = P concentration on P level₁.

An ANOVA combining data across experiments was used for all measured traits using proc GLM procedure in SAS version 9.4 (2002-2012) SAS Institute, Cary, NC). A test of hypotheses for mixed model analysis of variance was performed in which replication was considered random while genotypes and treatments (fertilizer rates) were considered fixed. Mean separation was performed using Tukey (P ≤ 0.05). Simple phenotypic correlations were computed among the different trait means for the genotypes in the no P treatment using PROC CORR in SAS.

There were significant (P ≤ 0.05) genotype and treatment effects for all traits measured except genotype for root-P (Table 2). A significant genotype × treatment

*, **, and *** represents significance at 0.5, 0.01, and 0.001 levels of probability respectively. Treatments = phosphorous levels.

effect was only observed for shoot biomass. There were significant genotype × experiment interactions for root biomass, shoot-P and seed-P, and treatment × experiment interactions for shoot biomass, root biomass shoot-P and for root-P, but these interactions were due more to differences in magnitude than crossover events. Thus, results focus on examining the significant genotypic, treatment, and genotype × treatment effects across experiments.

In a combined analysis across treatments and experiments, Montcalm had the highest mean value for shoot biomass while CAL 143 had the lowest. No significant differences were observed among genotypes for root biomass. For shoot-P Thort had the highest mean value while Bukoba had the lowest, root-P did not show any significant differences among genotypes. Thort had the highest value for seed-P while CAL 143 and Kijivu had the lowest. G122 had the highest mean seed yield while Rojo had the lowest (Table 3).

Reduced P fertilizer rate (0 applied) resulted in reduced shoot biomass, root biomass, shoot-P, root-P, seed-P and seed yield (Figure 1). There was variation between the highest fertilizer treatment 100 (kg P ha⁻¹) with the no P treatment for all traits measured. The greatest effect was between 0 and 50 with little to no difference between 50 to 100 for shoot biomass, seed yield and seed-P. These results indicate that dry beans are not responsive to more fertilizer P application at higher level of P in the soil [23] [24] .

Variations have been observed among genotypes and between fertilizer rates for shoot biomass, shoot P and root P (Table 4). Overall shoot biomass was the lowest in the no P treatment compared to 50 and 100 kg P ha⁻¹ treatments. For the no P treatment, the genotypes Montcalm, Bukoba and Kijivu had the highest value for shoot biomass while G122 had the lowest. No significant differences

Means followed by the same or no letter within the same column are not significantly different (P = 0.05) Tukey comparisons of mean.

were derived among genotypes and between treatments, for root biomass. For shoot P, variation was observed among genotypes and treatments. In the no P treatment, Cardinal had the highest value for shoot P while Bukoba has the lowest. Similarly, variations were observed among genotypes and between treatments for root P. In the no P treatment CAL 143 had the highest value for root P while Kijivu had the lowest value. For seed P except for CAL 143 in the no P treatment which significantly differed with Taylor Hort in the 50 and 100 kg P ha⁻¹, all other genotypes and treatments did not differ significantly. For seed yield G122 in the 100 kg P ha⁻¹ differed from Montcalm, Cardinal, Bukoba and CAL 143 in the no P treatment. No other significant differences were observed among genotypes within and between treatments (Table 4).

To examine the genotypic effect in different soil treatments, the PSI and PUE values were computed (Table 5 and Table 6).

It has been reported that shoot biomass in dry beans will increase with an increase in soil P levels [25] . Mean separation among fertilizer treatments shows that there was no difference between the highest (100 kg P ha⁻¹) and normal P treatments (50 kg P ha⁻¹), (Table 4, Figure 1). Shoot biomass was consistently low when no P was added, and the genotypes G122, CAL 143, Cardinal and Rojo were more susceptible to P deficiency as indicated by significantly lower shoot biomass in no P treatment than in 50 kg P ha⁻¹ and 100 kg P ha⁻¹ P treatments (Table 4).

Means followed by the same or no letter within the same column are not significantly different (P = 0.05) Tukey comparisons of mean.

The PSI calculations for each genotype are presented in Table 5. Values greater than 1.1 indicate susceptibility to P-deficiency [26] . PSI calculations show that G122 and Cardinal have high PSI values of 1.63 and 1.19 respectively, indicating susceptibility to P-deficiency. These results confirm Tukey means comparisons in Table 4 that G122, Cardinal and Rojo were more susceptible to P deficiencies. The genotype Bukoba had the lowest PSI value of 0.49, suggesting that this genotype may be more tolerant to low P deficiency. Montcalm, Kijivu, CAL143 and Taylor Horticulture have PSI values between 0 and 1.1 indicating moderate to high tolerance to P deficiency.

The no-P treatment had lower root biomass compared to normal and high P treatments (Figure 1). The results for shoot:root ratios are presented in Table 5. We expected the ratio to be higher in low P treatment compared to high P treatment. These results follow the trend for the two genotypes Rojo and Bukoba. For all other genotypes the ratio is higher in the normal P treatment (50 kg P ha⁻¹) compared to the low (no-P) and high P treatment (100 kg P ha⁻¹). We speculate that there might be an enhanced root growth response when there was sufficient P compared to limited or excess P conditions.

Tissue P concentrations allow the identification of P tolerant genotypes with higher PUE. A genotype that maximizes shoot dry matter while retaining P is a desired one, because the P retained within the plant can be translocated for seed production. Thus, it was expected that plants with high tissue P before flowering may have the highest seed production and seeds with higher P because more P is available from the vegetative tissue to be mobilized for seed production. Shoot P concentrations increased with the increasing applied P. The results for P concentration (mg kg⁻¹) and P content (mg) are shown in Table 5.

For the no P treatment, Cardinal had the highest shoot P concentration of 119 mg kg⁻¹ while Bukoba had the lowest value of 85 mg kg⁻¹. In the high P treatment (100 kg P ha⁻¹), Thort had the highest value of 299 mg kg⁻¹ and Bukoba had the lowest value of 237 mg kg⁻¹ (Table 3). All plants under the 50 kg ha⁻¹ P treatment doubled the amount of tissue P relative to plants under the no-P treatment that was not reflected in shoot biomass.

Shoot P concentrations are known to decrease with increasing nutrient deficiency and crop age [27] . The percent P in shoots (g P 100 g shoot tissue⁻¹) for optimal growth is considered to be in the range of 0.3 - 0.5 g P 100 g shoot tissue⁻¹ during vegetative growth [28] . A 1 g P 100 g shoot tissue⁻¹ concentration or higher is considered toxic for plant metabolism. However, this value depends on the crop in question. Most tropical legumes are very sensitive to excess P, as toxicity may occur at much lower levels of P in the shoot, for example toxicity may occur at 0.3% - 0.4% in pigeon pea [11] . Conversely, with decreasing P supply, P concentrations in shoots exhibited severe P deficiency (Table 5). The range of mean shoot P concentrations under the no P treatment across varieties ranged from 79.16 mg kg⁻¹ for Bukoba to 112.38 mg kg⁻¹ for Thort, which are equivalent to 0.0008 to 0.0011 g P 100 g shoot tissue⁻¹ respectively. These levels indicate that all genotypes in the no P treatment had severe deficient shoot P content compared to the genotypes amended with 50 and 100 kg P ha⁻¹.

Cardinal had the highest root P concentration of 183 mg kg⁻¹ while Kijivu had the lowest value of 129 mg kg⁻¹ under the no P treatment. In the 50 kg P ha⁻¹ Thort had the highest value of 260 mg kg⁻¹ and Rojo had the lowest value of 217 mg kg⁻¹. These results indicate that shoot internal P and, root internal P concentrations doubled the amount of tissue P in 50 kg ha⁻¹ P treatments compared to the no P treatment (Table 3).

Considering the no P treatment, the genotype Cardinal had higher value for seed P content while CAL 143 had the lowest value. Results suggest that the high value of seed P for Cardinal may confer an early vigor growth advantage. This early vigor may continue to have a positive effect in later stages as a variety experiences P-deficiency stress.

Genotypes response to different P treatments depends on genetic makeup and the level of stress imposed. Kijivu had the highest seed yield under no P treatment followed by G122 and Taylor Horticulture. Although not in the same order, the same three genotypes are among the highest yielding in the normal P treatment (50 kg P ha⁻¹). Overall seed yield was lower in no P treatment compared to the normal and high P treatments (Figure 1).

Results for PUE are presented in Table 6. PUE was calculated as the mean shoot biomass (g) divided by the mean tissue P (mg). A high PUE score reflects a high shoot biomass obtained by remobilizing a high proportion of P. Considering that tissue P content is later translocated for seed production, a high PUE may or may not be desired for the sake of seed production.

Results show that when no P is added, the genotype Bukoba had the highest PUE while Thort, G122 and Cardinal had the lowest. Under 50 kg ha⁻¹ P treatments, Montcalm exhibited the highest PUE while Thort had the lowest. The genotypes Bukoba and Montcalm may be exhibiting tolerance by efficient reallocation of their tissue P. PUE was of no consequence when P was oversupplied above adequate levels to support proper nutrition (Table 6).

Pearson correlation analysis was used to identify relationships among traits in the no P treatment (Table 7). A significant correlation at P ≤ 0.001 was observed for shoot biomass to PUE, and mean seed yield to mean seed P, suggesting that genotypes with high shoot biomass can utilize P efficiently when this nutrient element is limited. Other traits with significant correlations at P ≤ 0.01 include shoot biomass and root biomass, shoot biomass and shoot P, and shoot P to root P concentration, which indicated that shoot and root biomass and internal P concentrations are associated when beans were grown in P deficient soils.

Results showed a significant correlation at P ≤ 0.05 for several traits: root biomass to PUE, root biomass to shoot P, root biomass to root P, seed P to shoot P, PUE to shoot P and shoot P to root P; suggesting a high root biomass was advantageous as there would be greater root surface area during development that leads to higher P uptake, as represented by high shoot P and root P. A significant and negative correlation at P ≤ 0.01 was observed for the following traits: shoot biomass to PSI shoot, root biomass to PSI shoot, root biomass to PSI root, shoot P concentration to PUE, PUE to PSI shoot, PSI shoot to shoot P and PSI root to root P. These traits negatively correlated with PSI because lower PSI values indicate greater tolerance to P-deficiency. Shoot P concentration and PSI shoot were correlated negatively to PUE, suggesting these two traits are associated with P-deficiency tolerance. The negative correlation between shoot P concentration

*, **, ***Significant at P ≤ 0.05, 0.01, or 0.001, respective.

and PUE occurs because shoot P concentration is used to calculate shoot P which is then used to calculate PUE.