**Additive and Non-Additive Effects on the Control of Key Agronomic Traits in Popcorn Lines under Contrasting Phosphorus Conditions**

**Talles de Oliveira Santos 1,\*, Fábio Tomaz de Oliveira 1, Antônio Teixeira do Amaral Junior 1,\*, Janeo Eustáquio de Almeida Filho 2, Rosimeire Barboza Bispo 1, Marta Simone Mendonça de Freitas 3, José Francisco Teixeira do Amaral 4, Samuel Henrique Kamphorst 1, Valter Jário de Lima 1, Flávia Nicácio Viana 1, Guilherme Ferreira Pena 1, Pedro Henrique Araújo Diniz Santos 1, Wallace de Paula Bernado 1, Messias Gonzaga Pereira 1, Jurandi Gonçalves de Oliveira 1, Ricardo Enrique Bressan-Smith <sup>1</sup> and Roberto dos Santos Trindade <sup>5</sup>**


**Abstract:** Phosphorus is a non-renewable natural resource that will run out of reserves in the upcoming decades, making it essential to understanding the inheritance of nutrient use efficiency for selecting superior genotypes. This study investigated the additive and non-additive effects of commercially relevant traits for the popcorn crop (grain yield—GY, popping expansion—PE, and expanded popcorn volume per hectare—PV) in different conditions of phosphorus (P) availability in two locations in Rio de Janeiro State, Brazil. Six S7 lines previously selected for P use—L59, L70, and P7, efficient and responsive; and L54, L75, and L80, inefficient and non-responsive—were used as testers in crosses with 15 progenies from the fifth cycle of intrapopulation recurrent selection of UENF-14, with adaptation to the North and Northwest regions of Rio de Janeiro State. Using the Griffing diallel analysis, P use efficiency was predominantly additive in the expression of PE, and non-additive effects were prominent for GY and PV. For obtaining genotypes that are efficient for phosphorus use, it is recommended that heterosis with parents that provide additive gene accumulation for PE be explored.

**Keywords:** abiotic stress; genetic control; Griffing diallel analysis; *Zea mays* var. *everta*

#### **1. Introduction**

Phosphorus (P) is the second most consumed nutrient in agriculture, surpassed only by nitrogen, and is limiting for agricultural productivity worldwide [1]. Despite playing a crucial role in crop productivity, its soil reserves (organic and inorganic forms) have limited supply to the plants because of its fixation and formation of complexes with other soil nutrients [2,3].

Maize (*Zea mays*) is one of the most widely cultivated species of commercial interest, both as a staple food and for industrial use, in tropical and temperate climatic soils in the world. Under cultivation, especially in acidic and alkaline soils, large amounts of P fertilizer are applied to maize fields to maximize yields. The consequence is high amounts of high-cost phosphate fertilizers being applied to P-deficient soils to achieve maximum yields to guarantee food security for the growing world population, which will hit 10 billion inhabitants by 2050 [4]. Therefore, to meet this demand, the global production of phosphate

**Citation:** Santos, T.d.O.; de Oliveira, F.T.; Amaral Junior, A.T.d.; de Almeida Filho, J.E.; Bispo, R.B.; de Freitas, M.S.M.; Amaral, J.F.T.d.; Kamphorst, S.H.; de Lima, V.J.; Viana, F.N.; et al. Additive and Non-Additive Effects on the Control of Key Agronomic Traits in Popcorn Lines under Contrasting Phosphorus Conditions. *Plants* **2022**, *11*, 2216. https://doi.org/10.3390/ plants11172216

Academic Editor: Przemysław Barłóg

Received: 26 July 2022 Accepted: 23 August 2022 Published: 26 August 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

fertilizers will require a significant increase in phosphate extraction in the next decades. This is, however, a non-renewable resource that is likely to become scarcer as its use becomes more frequent [5].

Accordingly, promoting phosphorus use efficiency in crop plants is key to sustainable agricultural development; this is especially important in tropical and subtropical regions, where there is greater loss of the nutrient due to high temperatures and heavy rainfall, together with the fixation of the nutrient by iron and aluminum oxides in the soil, resulting in the loss of about 70–80% of the nutrient applied to crops [6–8].

Phosphorus deficiency in soil leads to physiological and biochemical disturbances in plants. As a result, they show a reduction in leaf area, height, dry matter, and major metabolic activities [9]. This is because phosphorus plays a fundamental part in the production of energy and enzyme activation, in addition to being a structural element of nucleic acids and phospholipids and participating in processes such as cell division [8]. P deficiency generates significant impacts on commercially important crops, among them popcorn (*Zea mays everta*), causing large losses in yields.

The knowledge of the genetic basis regarding agronomic traits of interest is extremely relevant for plant breeding programs that seek to increase crop yield in areas where there is low soil-nutrient availability by means of generating superior hybrids and segregants for limiting conditions [10]. To this end, diallel crosses are widely used in cultivated species to provide genetic information by estimating the combining abilities of the parents and hybrids [11]. This strategy enables the estimation of the existence of additive effects from the parents and nonadditive effects in crosses [12]. Diallel analysis, however, may become impractical depending on the number of lines to be used, requiring large experimental areas and labor in manual crosses. To solve this problem, the testcross method proposed by Davis [13] has been used as an option, allowing the evaluation of many lines in crosses with testers. By doing so, lines with inferior agronomic performance may be eliminated. As a result, the crosses show the most promising lines, thus making the procedure more effective [14].

Previous studies carried out for popcorn have proven the genetic action of the main traits of economic importance in the crop. In environments with adequate water and nutrient supply [15–17] and under water-deficit conditions [18,19] as well as P and N stress [10,20,21], additive genetic action prevails for grain popping expansion, while non-additive action has been the most important in the expression of grain yield under water stress [14,18,19]. These results are also consistent with studies conducted with biotic stressors [22–26].

Despite the relevance, only one research paper has been developed so far [10] toward understanding the influence of additive and non-additive effects on popcorn, a crop that earns about USD 1 billion annually in the United States. Given this, this study was considered relevant to contributing to filling the current gaps in knowledge regarding the genetic control of economically important traits for popcorn under phosphorus-limiting conditions. Therefore, the general and specific combining abilities and the genetic merit of 90 testcross hybrids of popcorn were estimated, and inferences were made about their additive and non-additive gene effects on the key agronomic traits of the crop. The goal was to propose breeding strategies for popcorn to develop cultivars adapted to phosphoruslimiting conditions as an option for leveraging Brazilian agribusiness.

#### **2. Results**

#### *2.1. Genetic Variability for Agronomic Traits Evaluated under Contrasting Phosphorus Conditions in Both Environments*

A significant difference between the genotypes (G) at a 1% probability level (*p* ≤ 0.01) was detected using the F test for all the traits (grain yield—GY, popping expansion—PE, and expanded popcorn volume—PV) for both levels of phosphorus availability, in Campos dos Goytacazes and Itaocara. The source of variation in phosphorus availability (P) and the G × P interaction also exhibited significant effects (*p* ≤ 0.01) for all the traits evaluated in the two locations studied (Table 1).


**Table 1.** Summary of the analysis of variance, of the general combining ability (GCA), and of the specific combining ability (SCA) for the agronomic traits evaluated in testcrosses under contrasting conditions of phosphorus in soil in Campos dos Goytacazes and Itaocara.

\*\* and \*—significant at 1% and 5% level probability using the F test; and ns—not significant at 5% probability using the F test; SV—source of variation; DF—degree of freedom; CVe (%)—experimental coefficient of variation; PE—popping expansion; GY—grain yield; and PV—expanded popcorn volume per hectare.

Regarding the effects of the general combining ability of the progenies (GCA I) and the testers (GCA II), the estimates were significant for all traits except for PV in Campos dos Goytacazes. Analyzing the mean square values for specific combining ability (SCA), it was observed that, for the GY, PE, and PV traits, there were highly significant values (*p* ≤ 0.01) in Campos dos Goytacazes. In Itaocara, there was only significance for GY at a 5% probability level (Table 1).

There was high significance for the GY and PE traits in Campos dos Goytacazes and for GY, PE, and PV in Itaocara regarding the progeny interaction with phosphorus availability (GCA I × P). The tester interactions with phosphorus levels (GCA II × A) and between specific combining ability and phosphorus levels (SCA × P) were found to be significant for all the traits evaluated.

#### *2.2. General Combining Ability Effects*

General combining ability (GCA) estimates provide information about the additive effects of genes for the traits studied. From this, it may be stated that lines L682, L688, and L686, in Campos dos Goytacazes, showed the three highest values—338.431, 310.452, and 290.454, respectively—for GY in the environment without induced stress. In turn, for the low-phosphorus environment, the most positive values were assigned to the genotypes L688, L695, and L689 (313.311, 280.976, and 207.294, respectively) (Figure 1).

Considering the testers in question for GY, three—L70, L59, and L75—in the environment with high phosphorus levels and four—L80, L59, L70, and L75—in the environment with low phosphorus levels displayed positive deviations. Three of the best lines used as testers—L70, L59, and L75—presented positive values in both phosphorus conditions, and L59 was the tester that showed the lowest variation when comparing 367 to 739, ranking well in both high and low soil-phosphorus conditions in the Campos dos Goytacazes environment (Figure 1).

For the general combining ability effect for the PE trait, eight lines at the optimal phosphorus level and seven lines at the low phosphorus level showed positive values for PE, especially for L681, L689, and L690 at the optimal level and L681, L688, and L689 at the low P level (Figure 1). Lines L681 and L689 demonstrated the highest effects of GCA and the lowest variations when both environments were analyzed. Favorable performance was seen for the P7 and L54 lines in the environment with high phosphorus levels when evaluating the testers for PE. As for the phosphorus-deficient environment, the testers with

positive values were P7, L80, and L54. Thus, the tester P7 achieved good performance regardless of the environmental conditions.

**Figure 1.** Estimates of general combining ability effects (*giˆ* ) for three traits evaluated in 21 popcorn parents in a partial diallel scheme in Campos dos Goytacazes. HP—high phosphorus level; LP—low phosphorus level; PE—popping expansion; GY—grain yield; and PV—popcorn expanded volume per hectare; (−) signal indicating negative values.

For the PV trait, the GCA estimates identified six lines—L688, L689, L691, L681, L685, and L696—with positive values in the environment with optimal phosphorus availability. On the other hand, in the environment with a low supply of the nutrient for this same trait, eight lines—L688, L689, L695, L683, L681, L691, L696, and L684—showed positive values. In the simultaneous analysis of the environments, the lines with the best results were L688 and L689, respectively. Additionally, considering the PV, of the six testers, three—L70, P7, and L59—expressed positive results in the environment with high phosphorus levels. In the environment with stress, only two testers—L80 and L70—had positive values. Among the genotypes with good results, tester line L70, from the BRS Angela population, showed positive values for PV at both phosphorus levels in Campos dos Goytacazes (Figure 1).

The general combining ability effects for GY in Itaocara indicated that lines L694, L689, L682, L684, L686, L688, L690, and L683 had positive deviations from the mean in the environment with fertilizer recommendations. Under the nutrient-deficiency restriction, the best lines were L689, L694, L685, L688, L693, L682, L684, and L695. Given the best performances, lines L694, L689, L682, L684, and L688 were in both study environments. Thus, it should be stressed that lines L694 and L689 had the best positions in the ranking (Figure 2).

**Figure 2.** Estimates of general combining ability effects (*giˆ* ) for three traits evaluated in 21 popcorn parents in a partial diallel scheme in Itaocara. HP—high phosphorus level; LP—low phosphorus level; PE—popping expansion; GY—grain yield; and PV—popcorn expanded volume per hectare; (−) signal indicating negative values.

As for the *giˆ* effects of the testers for GY in Itaocara, three lines—L80, L70, and L59 showed positive values in the environment with satisfactory fertilization, while in the environment with induced stress, four genotypes—L59, L70, L75, and L80—were seen as the most relevant. It should be observed that, when analyzing the ranking of the two environments, tester L70 remained stable, ranking second, surpassed only by lines L80 and L59 in the environment with fertilizer recommendation and phosphorus restriction, respectively (Figure 2).

The evaluations conducted in Itaocara showed the following hierarchical ranking of the lines with positive values for the PE trait in the environment where phosphorus was applied: L681, L688, L690, L689, L691, L694, L696, L685, and L683. For low levels of phosphorus in the soil, however, there were changes in the performances of the lines, which began to be arranged according to the ranking of best performances, as follows: L691, L681, L688, L683, L685, L690, L696, L694, and L689 (Figure 2). Regardless of the interaction with the phosphorus levels, the genotypes expressing positive performance were the same. It is worth noting, however, that changes were caused in the ranking of the best genotypes, and that lines L681 and L688 displayed good stability, maintaining high positive values. Regarding the effect of the six testers studied, it may be stated that there was no change resulting from the environments, especially for P7, L70, and L80, in the expression of positive values (Figure 2).

The evaluations performed in Itaocara indicated that, for the trait of popcorn expanded volume per hectare (PV), lines L694, L690, L688, L688, L689, L681, L683, and L696; and L689, L688, L685, L694, L683, L691, and L696, were the most prominent in environments with optimal and low phosphorus levels, respectively. Among these, lines L694, L688, and L689 expressed positive values higher than 12.201 in a range from −21.73 to 22.80. This confirms the good performance of these lines when considering the effects of the general combining ability of the S7 progenies (Figure 2). The PV trait also allowed for discrimination of the testers L70, L80, and P7 in the environment with a phosphorus supply. As for the nutrient-deficient environment, the genotypes that expressed the highest positive values were: L70, P7, L59, and L80. Tester L70 showed good performance in both environments, with positive values of 11.713 and 8.391, respectively, when evaluated in Itaocara (Figure 2).

Considering the set of traits evaluated for *giˆ* , the lines with the best performance for GY and PE in the environment with high phosphorus availability in Campos dos Goytacazes were: L682, L688, and L686; and L681, L689, and L688, respectively. For the PV variable, the genotypes with the highest positive values were: L688, L689, L691, and L681. Of special relevance is the phenotypic plasticity of line L688, which exhibited positive values for all the variables, suggesting that there is a range of favorable alleles in this parent.

The best lines for GY in the stressed environment were L688, L695, and L689, and for PE, L681, L688, and L689. It is emphasized that lines L688, L689, and L695 ranked highest for PV. Particularly noteworthy is the good positive deviation performance of L688 and L689 in the nutrient-deficient environment, confirming the presence of favorable alleles for phosphorus use efficiency in the parents from the UENF-14 population (Figures 1 and 2).

When the set of testers was evaluated, it can be seen that, for GY and PE, the genotypes L70 and L59, and P7 and L54, respectively, had good performances in the environment with the optimal phosphorus level in Campos dos Goytacazes. Regarding the PV trait, the most prominent genotypes were L70, P7, and L59, as they showed the highest positive values. In the environment with low phosphorus levels, a change in the performance of the testers was observed, in which the highest positive values were seen for lines L80, L59, and L70 relative to the GY trait. The most relevant genotypes for PE were P7, L80, and L54. The PV allowed tester lines L80 and L70 to stand out in presenting the highest positive values.

When comparing the individual values of GY and PE for the PV trait in relation to the experiments carried out in Itaocara, the genotypes L694, L690, and L688 gave the best performances. The order of the most positive values for variables GY and PE was attributed to lines: L694, L689, L682; and L681, L688, L690, respectively. Based on the considerations, line L688 showed good performance for both PV and PE and is considered a promising genotype. Regarding the environment with low phosphorus levels, the lines with the best performances were L689, L694, and L685 for GY; and L691, L681, and L688 for PE. Considering the PV, the genotypes with the highest positive values were L689, L688, and L685 (Figures 1 and 2).

Upon analyzing the testers when grown in Itaocara, differences in performance could also be identified. The most prominent positive values were found in L80, L70, and L59 for GY; and in P7, L70, and L80 for PE in the environment with the proper adjustments for phosphorus. Considerations carried out for PV allowed us to indicate the parents L70, L80, and P7 as the most promising. For the nutrient-deficient condition, the variable PV enabled us to distinguish lines L70, P7, and L59 as having the most positive values. For GY and PE, the genotypes L59, L70, and L75; and P7, L70, and L80, respectively, were the most relevant in this same environment.

#### *2.3. Genetic Merit of Hybrids*

Specific combining ability effects (SCA—*s*ˆ*ij*) are the result of the presence of dominance gene effects. The effects of *s*ˆ*ij* for GY in Campos dos Goytacazes showed 46 hybrids with positive variances, suggesting that the dominance gene effects prevail, which expressed amplitudes from 2.289 to 616.921 in the environment with phosphorus supplementation. The three hybrid combinations that exhibited the highest heterosis, considering *s*ˆ*ij* for this

trait, were L691 × P7, L686 × L80, and L689 × L59. Among the testers, the parents L75, L54, and L80 were involved in most crosses with positive SCA values, with magnitudes of 21.73%, 17.39%, and 17.39%, respectively (Figure 3).

**Figure 3.** Estimates of the genotypic value (*s*ˆ*ij*) of hybrids for grain yield (GY) in diallel hybrids of popcorn evaluated under high and low phosphorus availability in Campos dos Goytacazes. (−) signal indicating negative values.

On the other hand, in the environment with induced stress, the number of genotypes with positive values for *s*ˆ*ij* was significantly reduced. Among the 90 hybrids tested, only 37 crosses expressed positive magnitudes, which ranged from 2.601 to 1189.951. The most vigorous genotypes for *s*ˆ*ij* were L684 × L75, L691 × L70, and L686 × L54. As for the highest percentages of crosses with positive estimates in this case, they occurred with the testers L80 and L59, with values of 21.62% and 18.91%, respectively (Figure 3).

Therefore, the selection of parents based on additive genetic merit in Campos dos Goytacazes enabled the identification of genotypes with genetic potential to be phenotypically superior for each environment. In the experiment with fertilizer adjustments, the highest genetic merits were expressed in the combinations L88 × L70, L682 × L59, and L696 × L70, whereas in the nutrient-deficient environment, the hybrids L684 × L75, L691 × L70, and L688 × L70 expressed the highest genetic merits, exhibiting good allelic complementation (Figure 3).

The experiments conducted in Itaocara showed that the effects of *s*ˆ*ij* enabled 49 hybrids to be discriminated, with heterosis values estimated from 22.423 to 631.804 in the environment with fertilization recommended for the popcorn crop. This selection indicated the crosses with the highest dominance gene effects, with the hybrids L691 × L59, L685 × P7, and L692 × L70 standing out as having the best performances. Regarding the estimates of the general combining ability, the testers L80, L70 and L59 showed positive values, being in the formation of the best genotypes. Among them, line L59 was the most frequent, present in 20% of the best hybrid combinations (Figure 4).

**Figure 4.** Estimates of the genotypic value (*s*ˆ*ij*) for grain yield (GY) in diallel hybrids of popcorn evaluated under high and low phosphorus availability in Itaocara. (−) signal indicating negative values.

In the experiments for GY carried out in Itaocara, there was no considerable reduction in the number of genotypes obtained in the environment with nutrient reduction. Thus, from 90 combinations, 51.10% of the crosses had positive values. Among them, three hybrids (L691 × P7, L692 × L70, and L688 × L75) stood out with high estimates of *s*ˆ*ij*, and the greatest dominance gene effect was shown by the combination L691 × P7, representing high allelic complementation in the parents. The most frequent tester in the crosses was line L80, as it participated in nine positive crosses, accounting for approximately 20% of the positive combinations in Itaocara. Analyzing the *giˆ* estimates of this tester, the additive effect of the genes is evident, since it showed positive estimates and high magnitudes in the nutrient-deficient environment (Figure 4).

From the estimates of additive genetic merit, it was observed, in the experiments conducted in Itaocara, that the values of GY emphasized the hybrids L694 × L59, L689 × L59, and L88 × L75 as having the highest estimates of heterosis when grown in an environment with nutrient stress. In the environments with added phosphorus, the best allelic complementation was found in L694 × L80 and L694 × L70 (Figure 4).

In relation to the PE variable, the evaluations conducted in Campos dos Goytacazes under conditions with phosphorus supplementation made it possible to point out the testers L59 and L70 due to their positive SCA values and the highest frequencies in the formation of hybrids with positive estimates of *s*ˆ*ij*, corresponding to 20% of the crosses. These parents, however, showed no additive gene effects for their *gˆi* estimates. Additionally, for PE, 50% of the hybrid combinations showed positive magnitudes of *s*ˆ*ij*. However, the values showed a low range, varying from 0.166 to 3.534. The hybrids that exhibited the highest heterosis for SCA were L684 × L75, L691 × L75, and L684 × L54, respectively (Figure 5).

**Figure 5.** Estimates of the genotypic value (*s*ˆ*ij*) for popping expansion (PE) in diallel hybrids of popcorn evaluated under high and low phosphorus availability in Campos dos Goytacazes. (−) signal indicating negative values.

In the low-phosphorus environment, it was possible not only to obtain 47 hybrids with high gene complementation, in which the SCA estimates ranged from 0.053 to 5.483, but also to detect that the crosses L682 × L70, L681 × L59, and L691 × P7 had the highest SCA values. In terms of tester performance, line L59 had the greatest participation in the formation of superior hybrids, being the source of 21.27% of the best combinations (Figure 5).

When the genetic merit of the hybrids for PE was evaluated in both growing conditions in Campos dos Goytacazes, it was found that the hybrid combinations with the highest genetic merit were L691 × P7, L683 × L54, and L681 × L59 in a low-phosphorus environment, and L688 × P7, L689 × P7, and L688 × L54 in an environment with phosphorus supplementation (Figure 5).

In Itaocara, the experiments for PE inferred that the tester P7 had the highest percentages of allelic complementation in the crosses, with high estimates of *s*ˆ*ij* for 22% of the hybrid combinations in the phosphorus supply condition, and 20.90% in the nutrientdeficient environment. Thus, based on the combining abilities, it can be stated that the tester P7 has genes favorable for phosphorus use efficiency, and also expresses more pronounced genetic divergence than the other parents (Figure 6).

**Figure 6.** Estimates of the genotypic value (*s*ˆ*ij*) for popping expansion (PE) in diallel hybrids of popcorn evaluated under high and low phosphorus availability in Itaocara. (−) signal indicating negative values.

Fifty percent of the 90 hybrids evaluated for *s*ˆ*ij* displayed crosses with dominance gene effects in the environment with optimal phosphorus availability, whose combinations with the highest expressions of heterosis were: L684 × L54, L692 × L54, and L684 × L7. As for the environment with low phosphorus level, 47.80% of the combinations exhibited positive values. Of these, the hybrids that performed best were L694 × L54, L685 × L70, and L684 × L80, with the highest values of the specific combining ability effects (Figure 6).

Regarding the genetic merit of the hybrids tested in the environments, it was found that the combination L688 × P7 presented favorable alleles for the highest expression of PE, in addition to having good phenotypic plasticity, since it ranked well in both growing conditions. Among the hybrids with the best performance in the environment with phosphorus supplementation, the following combinations can be considered: L688 × P7, L696 × P7, and L690 × L80. As for the environment without a phosphate fertilizer supply, the hybrids ranking highest were L91 × P7, L685 × L70, and L688 × P7 (Figure 6). The good performance of the tester P7, which was in the best hybrid combinations and expressed the highest *giˆ* estimates in both high and low P level environments, should again be noted.

In the evaluation of PV in Campos dos Goytacazes, 56.70% of the combinations were found to have positive estimates of *s*ˆ*ij* in the environment with phosphorus supplementation. The combinations L689 × L59, L691 × P7, and L686 × L80 expressed the highest estimates of dominance gene effects. Such results corroborate the effects seen for GY when its SCA was evaluated. This result is understandable, since PV is a combination of GY and PE (Figure 7).

**Figure 7.** Estimates of the genotypic value (*s*ˆ*ij*) for expanded popcorn volume per hectare (PV) in diallel hybrids of popcorn evaluated under high and low phosphorus availability in Campos dos Goytacazes. (−) signal indicating negative values.

Regarding the best tester for this index, line L70 provided the highest heterosis in the crosses, totaling 21.56% of the combinations with positive values. This tester, along with line L80, showed the highest additive gene effects of *gˆi* for this trait. In the environment without the nutrient supplementation of phosphorus, the most frequent tester in the crosses with good allelic complementation was the parent L80 when evaluating its SCA estimates, which comprised 25% of the crosses with higher magnitudes of SCA. This result corresponds with the additive gene effects assigned to this genotype when its *gˆi* estimate was analyzed. As for the hybrids with the highest SCA estimates, they were L684 × L75, L682 × L70, and L691 × L70, respectively, when considering the 36 crosses with the highest magnitudes of dominance effect (Figure 7).

The genetic merit of the hybrids was also quantified for the PV index of plants grown in both environments in Campos dos Goytacazes. The evaluations indicated that the most vigorous combinations corresponded to L684 × L75, L682 × L70, and L688 × L80, in the nutrient-stressed environment, and L688 × L70, L689 × L59, and L691 × P7, in the experiment with a phosphorus supply. It should be noted that line L688 and tester L70 exhibited excellent performances per se when analyzing their respective general combining abilities, corroborating the prevalence of additive gene effects in both gene conformations (Figure 7).

The PV supercharacter also allowed for the discrimination of the behavior of the 90 hybrids evaluated in Itaocara. The combinations L694 × L80, L684 × L75, and L685 × P7 stood out with the highest heterosis in a set of 49 hybrids, for which there was a prevalence of dominance gene effects in the environment with phosphorus supplementation. The most heterotic hybrid—L694 × L80—is assumed to have the highest divergence among the parents (Figure 8).

**Figure 8.** Estimates of the genotypic value (*s*ˆ*ij*) for expanded popcorn volume per hectare (PV) in diallel hybrids of popcorn evaluated under high and low phosphorus availability in Itaocara. (−) signal indicating negative values.

In the low-phosphorus environment, 47 hybrids expressed dominance gene effects, with their amplitudes ranging from 0.255 to 21.709 for estimates of *s*ˆ*ij*. After selecting the hybrids with positive values, the ones with the highest heterosis were determined: L685 × L70, L691 × P7, and L688 × L75, respectively. It should be mentioned that tester L70, which comprised the best hybrid, was also found in a greater number of positive combinations for both environments (Figure 8). Therefore, it may be assumed that the higher frequency of this parent may be related to the additive gene effects attributed to it when considering its *gˆi* estimate.

The evaluation of the combining abilities for PV allowed us to know the genetic merit of the hybrids when grown in Itaocara. This analysis pointed out that the combinations L694 × L80, L694 × L70, and L688 × L80 had the best performances for the high-phosphorusavailability conditions. Hybrids L685 × L70, L689 × P7, and L688 × L70 were also identified as the ones with the best performance in the low-phosphorus environment. Moreover, using the evaluations conducted for PV, it could be verified that there was a predominance of the L70 and L80 testers in the low- and high-nutrient-availability environments. Additionally, the combinations L689 × P7 and L694 × P7 presented the greatest phenotypic plasticity since their recommendations may be made in either an environment with low or high phosphorus levels (Figure 8).

#### **3. Discussion**

The genetic variability among the lines under study due to the significant differences among the genotypes for the traits investigated—GY, PE, and PV—in Campos dos Goytacazes and Itaocara is evident (Table 1). Furthermore, the source of variation in phosphorus (P) availability also had significant effects for all the traits; this suggests that the dose of phosphorus applied to the soil was adequate to provide a distinction between the environments and enable correct differentiation of the efficient genotypes from the inefficient ones in phosphorus use in Campos dos Goytacazes and in Itaocara. This points out that the line classification differed

between the two growing conditions (high and low phosphorus levels in the soil), indicating a specific P-related effect. This provides strong evidence that there was significant variation in the ability of the material investigated to explore P in its environment, a prerequisite for selection-based breeding. However, screening this performance is not informative with respect to the number of genes underlying the variation observed, but encourages further research focusing on QTL analysis to provide more information on the genetic architecture of variation in response to different P levels in soil [27].

In addition to this, genetic variability in germplasm collections is an essential factor in obtaining genetic gains in breeding programs. Due to the high consumption of phosphate fertilizers—which may lead to a shortage of P reserves within a few decades—and the need to develop sustainable agriculture, these genotypes have become important sources of tolerance for breeding programs of popcorn to obtain gains in efficiency and responsiveness in phosphorus use. Gerhardt et al. [10] also reported genetic variability for agronomic traits when evaluating popcorn hybrids and lines at contrasting phosphorus levels in soil.

The G × P interaction, in turn, was significant for all the traits, suggesting a dissimilar response of the popcorn hybrids under different phosphorus-availability conditions (Table 1). This interaction may provide changes in the classification of hybrids between the experiments with high and low phosphorus levels. In this regard, Vencovsky and Barriga [28] recommend that the practice of genotype selection should be environment-specific, that is, each environment should have a set of distinct or partially distinct genotypes. Thus, selection should not be made based on average performance, as favorable alleles that control the expression of the character under stress differ, at least partially, from favorable alleles that control the same character under optimal conditions [29].

Significance in the estimates for all variables, except for PV, was observed in Campos dos Goytacazes when the effects of the source genotype variation were unfolded into general combining ability (GCA) and specific combining ability (SCA). This demonstrates that by exhibiting significance for GCA and SCA, GY and PE showed variability resulting from additive and non-additive effects in controlling gene expression. The significance of variation attributed to the additive effects proves that there are promising parents to be used, mainly in intrapopulation breeding programs, as this variation is very useful in pre-breeding to incorporate exotic germplasms into adapted populations or to adapt populations to abiotic stressors (Table 1). The magnitude of the additive variance expressed by the mean squares of the GCA of the progenies and the testers indicates the existence of significant additive effects of the genes [28].

Regarding the effects of specific combining ability—highly significant for the variables GY, PE, and PV (*p* ≤ 0.01) in Campos dos Goytacazes—it may be assumed to be the gene action of dominance in the expression of PE, which is a trait known to be governed by additive effects, as reported by Dofing et al. [30], Pacheco et al. [31], Larish and Brewbaker [32], and Pereira and Amaral Júnior [15]. As for Itaocara, there was only significance for GY at a 5% probability level (Table 1). This indicates that non-additive gene effects also exert an influence on these traits and suggests possible allelic complementation between the parents at the respective loci, with some degree of dominance. This possibility for PE gene expression agrees with more recent results obtained by Coan et al. [33], who support the existence of mixed inheritance in the expression of this trait.

In view of the significant progeny interaction with contrasting phosphorus levels (GCA I × P) for the traits GY and PE in Campos dos Goytacazes, and for the traits GY, PE and PV in Itaocara, it may be assumed that the additive gene effects between the lines provided differences. In this case, then, selection for each level of phosphorus is recommended, as already mentioned. The tester interaction with phosphorus levels (GCA II × P), also significant for all the traits evaluated, suggests that selection should be made for each phosphorus level. Gerhardt et al. [10], likewise, found significance for the interaction between popcorn progenies and contrasting environments in using phosphorus.

From the results, considering the significant interaction between specific combining ability and phosphorus levels (SCA × P), it can be noticed that the classification of SCA effects differed between environments. The results at each fertilization level should, thus, be considered to allow for the effective selection of hybrids that are efficient and responsive to phosphorus use.

As for the experimental precision expressed by the coefficient of experimental variation (CVe), the values found for all the traits were less than 20%, suggesting excellent experimental precision [34]. These results agree with the estimates found by Santos et al. [35] (2017) and Gerhardt et al. [10] in experiments with abiotic stressors in popcorn.

According to Sprague and Tatum [36], informed by Inocente et al. [37], the GCA corresponds to the average behavior of a line in a series of hybrid combinations, and is expressed by the *giˆ* estimate. For Cruz and Vencovsky [38], a low value of *giˆ* suggests that the average of the hybrids in which line *i* participates does not differ much from the overall average of the diallel, meaning that when there are high values, positive or negative, parent *i* is superior or inferior to the other parents in the diallel when compared to the average of their hybrids. Thus, the estimates of the *giˆ* effects for the variables analyzed have values with signs ranging from negative to positive as a function of the performance of the parent. Accordingly, Cruz and Vencovsky [38] and Scapim et al. [39] affirm that the line with the highest frequency of favorable alleles will express a higher *giˆ* .

Based on these references and considering the *giˆ* estimates, it is observed that, in Campos dos Goytacazes, lines L682, L688, and L686 had the three highest values—338.431, 310.452, and 290.454, respectively—in the environment with adequate phosphorus availability, and, in the environment with low phosphorus levels, the most positive values were assigned to the genotypes L688 (313.311), L695 (280.976), and L689 (207.294) (Figure 1).

Line L688 was not affected by a substantial reduction in phosphorus in the soil; thus, its high combining ability and phenotypic stability are relevant for programs aimed at increasing GY. Scapim et al. [39] reported the use of popcorn populations with high GCA estimates to form 211 varieties with high GY and PE, supporting the relevance of working with parents with high *giˆ* estimates for these traits, which are both of major relevance for popcorn trading.

Regarding the testers, for GY, under optimal phosphorus supply conditions, three lines showed positive deviations (L70, L59, and L75) and four stood out for the environment with induced stress (L80, L59, L70 and L75). Among the best lines used as testers, three—L70, L59, and L75—had positive values in both conditions of phosphorus availability. It should be emphasized that tester L59 had the least variation (367 and 739), ranking well in both phosphorus conditions in the soil, in Campos dos Goytacazes. This good performance may be associated with its genealogy (Beija-flor: UFV) and adaptation to a temperate/tropical climate, a factor correlated with the presence of favorable alleles in a good parent (Figure 1).

For PE, eight lines under optimal conditions of phosphorus availability and seven lines under limiting conditions of the nutrient showed positive values for *giˆ* estimates in Campos dos Goytacazes. There is agreement in the prominence of lines L681, L689, L690, L681, L688, and L689 (Figure 1) in both conditions of phosphorus availability; however, lines L681, and L689 had the greatest *giˆ* effects and least variation between the two environments. Therefore, it may be recommended that these parents obtain lines with good PE, since *giˆ* is the estimator that indicates the parents with the best average performance in crosses. These lines, thus, have the highest concentrations of favorable alleles for traits predominantly influenced by additive gene action.

As for the testers that were outstanding in environments with high (P7 and L54) and low (P7, L80, and L54) phosphorus levels, P7 showed the best performance regardless of phosphorus availability in the soil; this represents a line of interest to be included in crosses aimed at obtaining superior hybrids in relation to efficiency and responsiveness in phosphorus use.

Kamphorst et al. [40] further reported that line P7 showed high GY grain yield averages when grown in a water-stressed environment, and were considered agronomically efficient in water use. This suggests that this line has mechanisms to withstand adverse situations imposed by abiotic stress. Santos et al. [20] also described the relevance of P7 for GY and PE traits under optimal and low soil-nitrogen-availability conditions from a panel of ten popcorn lines evaluated in Campos dos Goytacazes and Itaocara. When evaluating 25 popcorn lines under high and low P conditions in soil, Gerhardt et al. [10] pointed out that, because of high *giˆ* estimates for GY and PE, line P7 has high potential for obtaining superior hybrids in terms of efficiency and responsiveness in phosphorus use.

In Campos dos Goytacazes, six lines distinguished themselves, for PV, by having positive values of *giˆ* estimates in the environment with high phosphorus levels—L688, L689, L691, L681, L685, and L696—and, in limiting conditions of the nutrient, eight lines stood out with positive values, namely: L688, L689, L695, L683, L681, L691, L696, and L684. When both environments were analyzed, it was noted that the lines with the best performance were L688 and L689, respectively (Figure 1). This coincidence of results corroborates the good performance of these genotypes and the substantial incidence of favorable alleles for phosphorus efficiency and use in the lines derived from the UENF-14 population (Table 1). The PV trait is considered a supercharacter that is intended to simulate a selection index, which enables concomitant gains for the two main traits of economic importance—GY and PE expansion—for the popcorn crop [41].

Still considering the PV, among the testers investigated, lines L70, P7, and L59 prevailed in the environment of high phosphorus levels in the soil, having shown positive *giˆ* results, whereas in phosphorus-limiting conditions, only two stood out—L80 and L70. Among the genotypes with good results, tester line L70, from the BRS Angela population, distinguished itself by having positive values for PV at both phosphorus levels in Campos dos Goytacazes (Figure 1). These results agree with the work of Schmitt et al. [42], who reported the good performance of the parent L70 in presenting a positive *giˆ* estimate for PE.

As for the effects of GCA for GY, in Itaocara, eight lines were highlighted based on *giˆ* estimates in the environment with high phosphorus level. Similarly, in low P conditions, eight lines also stood out. Considering both conditions, however, five lines distinguished themselves in both environments by presenting positive deviations, as follows: L694, L689, L682, L684, and L688. Within them, lines L694 and L689 were highlighted in the ranking, in that order (Figure 2).

When it comes to the *giˆ* effects of the testers, still for GY in Itaocara under high phosphorus conditions, three lines were prominent, and four genotypes were superior under P-deficient conditions in the soil. When analyzing both conditions of phosphorus availability, tester L70—which has adaptations to tropical environments—displayed greater stability, only surpassed by lines L80 and L59—adapted to temperate and tropical climates which were superior, respectively, in high- and low-P environments. This suggests good adaptation of these genotypes to the conditions studied for GY. By evaluating 15 populations of popcorn from different Latin-American countries under water-stress conditions, Santos et al. [43] related the importance of varieties with temperate and tropical climatic adaptations. Thus, according to the authors, the relevance of these materials in studies with popcorn to obtain superior genotypes is reinforced. Gerhardt et al. [10] pointed out, in a previous study, line L59 as being efficient and responsive in the use of phosphorus. Therefore, it is a genotype that expresses superior yields to the averages in environments of high and low phosphorus levels.

In this context, it is verified that, for the best S7 progenies, as well as for the testers, the *giˆ* estimates for PE were negative. This is due to the negative genetic correlation between GY and PE, a phenomenon already reported in other research conducted in the popcorn crop [15,16,30,43–47]. This correlation, therefore, makes it difficult to obtain genotypes with high GY and high popping expansion concomitantly [15,48,49].

By analyzing the results of the *giˆ* estimates for PE in Itaocara, even with the complex interaction, the best-performing genotypes, under optimal phosphorus conditions in the soil, were also better under limiting conditions. In this case, there were only differences in the ranking order, indicating that the change in the crop environment caused alterations in the performance of the genotypes under different P conditions. As stated before, when the interaction is of the complex type, it makes it difficult to indicate genotypes for a group of environments, requiring the breeder to recommend appropriate genotypes for specific conditions. The performance of lines L681 and L688 proved to be encouraging for this study, since they showed good phenotypic plasticity for the trait in question, ranking well in both environments, regardless of the P supply. Regarding the testers, the positive results of lines P7, L70, and L80 in both environments distinguish them as potential candidates for obtaining superior hybrids in both environments.

With regard to the *giˆ* effects for PV in Itaocara, lines L694, L688, and L689 were superior to the others as they showed positive values under optimal phosphorus conditions and when under nutrient stress, not exhibiting significant differences in performance, and keeping the positive values for *giˆ* . For the testers, L70 showed the same behavior, performing well under both phosphorus-availability conditions.

Given these considerations, it may be stated that the UENF-14 popcorn population has a high frequency of alleles favorable for the efficiency of phosphorus use in the soil. This assertion is supported by the good performance of lines L688 and L689, which showed good general combining ability and high GY, as well as high popping expansion, suggesting that there is a distinct heterotic pattern. Thus, lines L688 and L689 and testers P7 and L59 are highly recommended for use in future hybrid combinations aimed at increasing GY and PE in environments with low soil-phosphorus levels in breeding programs focusing on more sustainable agriculture.

By considering the merit of the hybrids for the GY and PV traits together, it was assumed that the L688 × L70 cross had higher heterosis when evaluated in environments with high P levels in the soil. In the environment with stress induction, heterosis was more evident in the genotype L684 × L75. Therefore, the interaction of these genotypes with the environments provided relative superiority compared to the others, suggesting that there are favorable alleles in these hybrid combinations.

For the traits PE and PV together, changes in the ranking of the crosses were observed, highlighting the superiority of the genotype L689 × L59, which showed the highest heterosis in the environment with high phosphorus levels. Thus, the hybrid combination L689 × P7 may be considered the most relevant one for the low-phosphorus environment in the experiments in Campos dos Goytacazes. As for Itaocara, the hybrid combinations with the highest popping expansion were L694 × L80 in the environment with added phosphorus, and L685 × L70 in the environment with low phosphorus levels. In relation to the testers, lines P7 and L54 showed additive genetic effects for GCA in the deficient environments, increasing popping expansion. The other testers distinguished themselves through their good allelic complementation in the crosses in which they participated.

When considering GY and PV traits together in Itaocara, the hybrid L694 × L80 was chosen as the most prominent for the environment with fertilizer recommendation for popcorn. For the low-phosphorus environment, the combination with the highest heterosis was L685 × L70; this hybrid also showed positive a magnitude of *s*ˆ*ij* when its SCA was evaluated.

More generally, the genotypes utilized present genetic potential for obtaining hybrids with high heterosis for efficiency and responsiveness in phosphorus use. Considering the prevailing gene effects of the GY, PE, and PV traits in the contrasting environments, the hybrid combinations that displayed promising results for efficiency and responsiveness in phosphorus use were L688 × L70, L694 × L80, L688 × L70; L694 × L80, L689 × L59, L694 × L80; and L689 × P7, and L689 × L59, respectively.

From the information obtained by means of the combining ability and having the most promising combinations, the parents from the UENF-14 population with the best performance—L688, L694, and L689—may be selected as standards for the creation of a new heterotic group and as testers of new lines from this heterotic group. Thus, this new heterotic group should be maintained and used separately for the generation of variability within the group, and for the selection of new lines.

Accordingly, the implementation of a procedure called "line recycle", originating from the UENF-14 population, is also proposed; this will form new biparental populations directly or via backcrossing, in which the elite lines L688, L694, and L689 from the UENF popcorn breeding program will be crossed among themselves, in subsequent cycles, within this same heterotic group. Thus, new lines will be obtained with superior traits to their parents because of the increased frequency of favorable alleles and the consequent higher level of heterosis when crossed to form a heterotic group.

In conclusion, to meet the agroeconomic aspects of the North and Northwest regions of Rio de Janeiro State—which tend to expand—on the basis of the combining ability and of the presence of favorable alleles, we propose obtaining a triple hybrid derived from a simple hybrid from parents originating from the heterotic group of the UENF-14 population—L688 × L689, for example, whose parents showed high GCA; this will be used as the female parent in a cross with the L80 line. Such a line has been demonstrated to be sufficiently vigorous to ensure good pollination and, consequently, satisfactory GY in female plants.

#### **4. Materials and Methods**

#### *4.1. Plant Material, Experimental Design, and Environmental Conditions*

Six S7 lines from the Germplasm Bank of the *Universidade Estadual do Norte Fluminense Darcy Ribeiro* (UENF)—previously classified by Gerhardt et al. [50] regarding P use—were used as testers, of which three (L59, L70, and P7) were efficient and responsive and three (L54, L75, and L80) were inefficient and non-responsive. They were used as testers in crosses with 15 other progenies (L681, L682, L683, L684, L685, L686, L688, L689, L690, L691, L692, L693, L694, L695, and L696) (Table 2) from the UENF-14 open-pollinated variety. This population was selected after five cycles of intrapopulation recurrent selection, adapted to the soil and climate conditions of the North and Northwest regions of Rio de Janeiro State [51]. Ninety testcrosses were generated from the crosses in a partial diallel scheme of the six testers with the 15 UENF-14 progenies.

**Table 2.** Description of the 21 popcorn genotypes from the Active Germplasm Bank of UENF used in the experiments in the North and Northwest regions of Rio de Janeiro State, Brazil.


UFV—Universidade Federal de Viçosa; UENF—Universidade Estadual do Norte Fluminense Darcy Ribeiro; EMBRAPA—Brazilian Agricultural Research Corporation; and IRS—Intrapopulation Recurrent Selection.

The experiments were conducted in a randomized block design with replication arrangements within sets. Five sets of 18 treatments (15 progenies plus three controls: UENF N 01, UENF N 02, and UENF HS 03) in each set were utilized, with three replications. These experiments were carried out during the harvest period, between October 2019 and March 2020, in two locations and under two contrasting conditions with respect to phosphorus availability (high and low phosphorus). The sites were the Experimental Station of the Colégio Estadual Agrícola Antônio Sarlo, in the municipality of Campos dos Goytacazes (latitude: 21◦42 48 S, longitude: 41◦20 38 W, 14 m above sea level), and the Experimental Station of Ilha do Pomba, in the municipality of Itaocara (latitude: 21◦38 50 S, longitude: 42◦03 46 W, 58 m above sea level); these correspond, respectively, to the Northern and

Northwestern regions of Rio de Janeiro State. The climate in Campos dos Goytacazes and Itaocara is classified as humid tropical (Aw), with hot summers and mild winters, with rainfall tending to be concentrated in the summer months.

Sowing was performed according to the conventional planting system, with a stand of 15 plants per plot, or 55,555 plants per hectare. Each experimental plot consisted of a 3.00 m row with 0.90 m spacing between rows and 0.20 m spacing between plants.

Prior to the experiments, soil chemical analysis was conducted to characterize the environments in terms of nutrient availability from samples collected in the 0–10 and 10–20 cm layers, forming a sample composed of ten subsamples (Table 3). The available P content was determined using the Mehlich-1 extractor. According to the clay soil content of Campos dos Goytacazes and Itaocara, the phosphorus levels were classified as low [52] (Ribeiro et al., 1999).

**Table 3.** Chemical and particle-size analysis of the soil at 0–10 and 10–20 cm depths in Campos dos Goytacazes and Itaocara.


Two doses of this fertilizer were used to simulate contrasting environments and to stimulate the genotypes to express the genes responsible for phosphorus efficiency and responsiveness. In the environment with high phosphorus availability, 30 kg ha−<sup>1</sup> of N (in the form of urea), 85 kg ha−<sup>1</sup> of P2O5 (triple superphosphate), and 40 kg ha−<sup>1</sup> of K2O (potassium chloride) were applied. For the environment with low phosphorus availability, 30 kg ha−<sup>1</sup> of N, 0 kg ha−<sup>1</sup> of P2O5, and 40 kg ha−<sup>1</sup> of K2O were used, which means the fertilizer in the stressed environment did not contain phosphorus. Topdressing fertilization was performed in partial applications in both environments when the plants reached the phenological stage of four (V4) and six (V6) fully expanded leaves at a concentration of 100 kg ha−<sup>1</sup> of N (in the form of urea). The other phytosanitary treatments were performed according to the recommendation for the crop in the North and Northwest regions of Rio de Janeiro State [53]. The experiments received additional irrigation whenever necessary to avoid water stress.

#### *4.2. Evaluated Traits*

The following traits were evaluated: (i) Grain yield (GY)—expressed by the average grain yield of the experimental unit in grams per plot, adjusted to 13% moisture and extrapolated to kg ha<sup>−</sup>1. (ii) Popping expansion (PE)—obtained using the ratio between the expanded popcorn volume and the mass of 30 g, expressed in mL g−1, utilizing the average of two samples per plot. Popping was conducted in a microwave oven, with 1200 Watts of power for 2 min, and the expanded popcorn volume was quantified in a 2000 mL beaker. The resulting value was divided by the initial weight of the 30 g grain, and the final result expressed in mL g−1. (iii) Expanded popcorn volume per hectare (PV), determined using the product between GY and PE, with the final value divided by 1000 and expressed in m3 ha<sup>−</sup>1.

#### *4.3. Data Analysis*

Adjustments of the set effects for all the hybrids were made according to the average of the controls common to all the sets before conducting the analysis of variance. We estimated the average of the controls for each set (ACS) and the general average of the controls (GAC); using the GAC/ACS relationship, the adjustment factor for each set was obtained, following the procedure used by Ribeiro et al. [54] and Guimarães et al. [55].

After these adjustments, the analyses were performed following the randomized block model. At first, the individual analyses of variance were conducted considering the environments with high and low phosphorus, according to the following statistical model: Yij =m+Bj + Gi + eij, in which Yij is the observation of the i-th genotype in the j-th block; m is the general constant; Bj is the effect of the j-th block; Gi is the effect of the i-th genotype; and eij is the experimental error associated with the observation Yij, which is normally and independently distributed (NID—0,*σ*2).

Subsequently, a joint analysis of variance was conducted to determine possible interactions between the genotypes and the levels of phosphorus in each location. The joint analysis of variance was conducted following the statistical model: Yijk = m + B/Ajk + Aj + Gi + GAij + eijk, in which Yijk is the observation of the i-th genotype in the j-th block in the k-th block; m is the general constant; B/Ajk is the effect of the k-th block in the j-th environment; Aj is the fixed effect of the j-th environment (P level); Gi is the fixed effect of the i-th genotype; GAij is the fixed effect of the interaction between the i-th genotype and the j-th environment; eijk is the experimental error associated with the observation Yijk with NID (0, *σ*2).

Based on the averages of the 90 testcross hybrids, a diallel analysis was carried out in line with the methodology suggested by Griffing [56], adapted to partial diallels, as follows: Yi j =m+gi + gj + sij + ak + gaik + gajk + saij + e(k)ij, in which Yij: The average value of the hybrid combination, testers (gi), and lines (gj); m: The overall average of the hybrid combinations; gi: The general combining ability (GCA) effect of group i (progenies); gj: The GCA effect of group j (testers); sij: The specific combining ability (SCA) effect for the crosses between parents of the orders i and j; ak: The effect of the environments k; gaik and gajk: The interaction effects between the GCA associated with the i-th and j-th parents in the environments k; saij: The interaction effect of SCA associated with parents i and j and environments k; and e(k)ij: The average experimental error.

The statistical analyses were conducted using the software GENES (v.1, Universidade Federal de Viçosa, MG, Brazil) [57].

#### **5. Conclusions**

Phosphorus use efficiency is mostly dependent on additive gene effects in the expression of popping expansion, and conversely, dominance gene effects in the expression of GY and PV.

The best strategy to obtain efficient and responsive genotypes in phosphorus use involves the exploration of heterosis, with parents that provide an accumulation of additive genes for popping expansion.

Lines L688 and L689 and testers P7 and L59 showed additive gene effects and are highly recommended for use in future hybrid combinations looking to increase GY and PE in environments with low soil-phosphorus levels.

The genetic merit of the hybrids enabled us to know the combinations with the highest dominance gene effects for efficiency and responsiveness in phosphorus use for the traits GY, PE, and PV, as follows: L688 × L70, L694 × L80, L688 × L70; L694 × L80, L689 × L59, L694 × L80; L689 × P7, and L689 × L59, in that order.

Due to the good performance of the parents originating from the UENF-14 population, L688, L694, and L689 are selected as the standards for forming a new heterotic group and for "line recycle", creating new biparental populations.

The production of a triple hybrid by crossing the simple hybrid L688 × L689—from the UENF-14 heterotic group—with the L80 line—from the Viçosa: UFV genealogy—is a good alternative for meeting the needs of the North and Northwest regions of Rio de Janeiro State and for achieving sustainable agriculture.

**Author Contributions:** Conceptualization, A.T.d.A.J., J.E.d.A.F., J.F.T.d.A., P.H.A.D.S. and M.G.P., methodology, T.d.O.S., F.T.d.O., M.S.M.d.F., J.F.T.d.A., S.H.K., V.J.d.L., F.N.V., G.F.P. and P.H.A.D.S.; software, T.d.O.S., F.T.d.O., J.E.d.A.F., S.H.K., V.J.d.L., F.N.V., G.F.P. and P.H.A.D.S.; validation, T.d.O.S., F.T.d.O., A.T.d.A.J., J.E.d.A.F., R.B.B., M.S.M.d.F., J.F.T.d.A., W.d.P.B., M.G.P., J.G.d.O. and R.E.B.-S.; formal analysis, T.d.O.S., F.T.d.O., R.B.B., M.S.M.d.F., J.F.T.d.A., W.d.P.B., J.G.d.O. and R.E.B.- S.; investigation, T.d.O.S., F.T.d.O., A.T.d.A.J., J.E.d.A.F., R.B.B., M.S.M.d.F., J.F.T.d.A., S.H.K., V.J.d.L., F.N.V., G.F.P., P.H.A.D.S., W.d.P.B., M.G.P., J.G.d.O., R.E.B.-S. and R.d.S.T.; resources, A.T.d.A.J., J.E.d.A.F., P.H.A.D.S. and M.G.P.; data curation, T.d.O.S., F.T.d.O., A.T.d.A.J., S.H.K., V.J.d.L., F.N.V., G.F.P., P.H.A.D.S., W.d.P.B., M.G.P., J.G.d.O., R.E.B.-S. and R.d.S.T.; writing—original draft preparation, T.d.O.S., F.T.d.O., A.T.d.A.J., J.E.d.A.F., R.B.B., M.S.M.d.F., J.F.T.d.A., S.H.K., V.J.d.L., F.N.V., G.F.P., P.H.A.D.S., W.d.P.B., M.G.P., J.G.d.O., R.E.B.-S. and R.d.S.T.; writing—review and editing, T.d.O.S., F.T.d.O., A.T.d.A.J., J.E.d.A.F., R.B.B., M.S.M.d.F., J.F.T.d.A., S.H.K., V.J.d.L., F.N.V., G.F.P., P.H.A.D.S., W.d.P.B., M.G.P., J.G.d.O., R.E.B.-S. and R.d.S.T.; visualization, T.d.O.S., A.T.d.A.J., J.E.d.A.F., M.G.P., J.G.d.O., R.E.B.-S. and R.d.S.T.; supervision, A.T.d.A.J., J.E.d.A.F., M.S.M.d.F., J.F.T.d.A., M.G.P., J.G.d.O., R.E.B.-S. and R.d.S.T.; project administration, A.T.d.A.J.; funding acquisition, A.T.d.A.J. and M.G.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (Capes, Brazil), grant number 001, and by the Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ, Brazil).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**César Santos 1, Marcelo Ribeiro Malta 2, Mariana Gabriele Marcolino Gonçalves 1, Flávio Meira Borém 3, Adélia Aziz Alexandre Pozza 1, Herminia Emilia Prieto Martinez 4, Taylor Lima de Souza 1, Wantuir Filipe Teixeira Chagas 1, Maria Elisa Araújo de Melo 1, Damiany Pádua Oliveira 1, Alan Dhan Costa Lima 1, Lívia Botelho de Abreu 1, Thiago Henrique Pereira Reis 1, Thaís Regina de Souza 5, Victor Ramirez Builes <sup>5</sup> and Douglas Guelfi 1,\***


**Abstract:** The present study had the objective to evaluate the effect of blends of KCl and K2SO4 fertilizers and their influence on the yield and the nutritional state of coffee plants, as well as on the chemical composition and quality of the coffee beverage. The experimental design was in randomized blocks with four repetitions and six treatments (T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; and a control, without application of K). The following analyses were performed: K and Cl content in the leaves and the soil, stocks of Cl in soil, yield, removal of K and Cl with the beans, cup quality of the beverage, polyphenol oxidase activity (PPO), electric conductivity (EC), potassium leaching (KL), the content of phenolic compounds, the content of total sugars (TS), and total titratable acidity (TTA). The stocks of Cl in the soil decreased as the proportion of KCl in the fertilizer was reduced. The fertilization with KCl reduces the cup quality and the activity of the polyphenol oxidase, probably due to the ion Cl. The increase in the application of Cl directly relates to the increase in potassium leaching, electric conductivity, and titratable acidity. Indirectly, these variables indicate damages to the cells by the use of Cl in the fertilizer. The activity of the polyphenol oxidase enzyme and the cup quality indicate that the ion Cl- reduces the quality of the coffee beverage. K content in the leaves was not influenced by the application of blends of K fertilizer while Cl content increased linearly with KCl applied. The application of KCl and K2SO4 blends influenced coffee yield and the optimum proportion was 25% of KCl and 75% of K2SO4. The highest score in the cup quality test was observed with 100% K2SO4.

**Keywords:** blend fertilizers; chlorine; cup test; polyphenol oxidase

#### **1. Introduction**

Coffee is one of the most popular beverages in the world, and its cultivation is widespread in 80 countries. Brazil is the largest producer and second largest consumer of coffee in the world. The gross revenue of "Cafés do Brasil" in the 2022 harvest was R\$61.82 billion with *Coffea arabica* accounting for 77% the total revenue (R\$ 47.48 billion). The state of Minas Gerais is the largest producer, responding for R\$ 33.28 billion or 54% of national revenues [1]. After petroleum, coffee is the second most commercialized product [2,3]. The marketing price is based on the quality of the beverage, which is related to the physical, chemical, and sensorial characteristics of the product [2,4,5].

Fertilization and crop nutrition can influence both yield and the chemical composition of the raw beans, which, consequently, interfere with the quality of the beverage [6].

**Citation:** Santos, C.; Malta, M.R.; Gonçalves, M.G.M.; Borém, F.M.; Pozza, A.A.A.; Martinez, H.E.P.; de Souza, T.L.; Chagas, W.F.T.; de Melo, M.E.A.; Oliveira, D.P.; et al. Chloride Applied via Fertilizer Affects Plant Nutrition and Coffee Quality. *Plants* **2023**, *12*, 885. https://doi.org/ 10.3390/plants12040885

Academic Editors: Przemysław Barłóg, Jim Moir, Lukáš Hlisnikovský and Xinhua He

Received: 23 December 2022 Revised: 2 February 2023 Accepted: 4 February 2023 Published: 15 February 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

After nitrogen, potassium (K) is the most accumulated nutrient in coffee plant fruits, where it is demanded in high amounts. K is related to the enzymatic activation of several metabolic processes, such as photosynthesis, proteins, and carbohydrates synthesis, and in the maintenance of cell turgidity [7–9]. In addition, K is directly related to the transport of sugars from the source to the drain (fruits) [10].

The effects of the accompanying chloride ion (Cl) of the potassium chloride fertilizer (KCl) are currently under debate. Cl is demanded by the plants at low amounts, thus being one of the last micronutrients to enter the micronutrient list. Its role is related to the water photolysis on the photosystem II, enzyme activation (amylase, asparagine synthetase, and tonoplast ATPase), and stomatal control [11].

Despite Cl being essential to plant nutrition, when accompanying a highly demanded macronutrient such as K, it can reach excessive concentrations in the soil and plants [12] and consequently reduce the quality of the beverage. In coffee plants, high concentrations of Cl are related to the increase in plant water, which favors an undesirable fermentation of the fruits by microorganisms [13,14].

The study of the Cl influence on the quality of the coffee beverage is not recent, but it is still inconclusive. For example, Dias et al. [15] evaluated an alternative source of K (glauconite silicate mineral) to KCl in coffee fertilization. Despite finding similar yields and polyphenol oxidase activity (PPO) in the beans, the use of the glauconite did not improve the sensorial quality. Silva et al. [16], along two seasons, verified that fertilization with potassium sulfate (K2SO4) increased PPO activity in comparison with KCl, which, according to the authors, is indicative of a better beverage quality. These studies suggest possible negative effects of the Cl on the quality of the coffee beverage and the necessity to use K sources without Cl as the accompanying ion, such as K2SO4 (48% K2O, 16% S) and potassium nitrate (44% K2O, 13% N). Nonetheless, this could increase the production costs, as these sources are more expensive than KCl. In turn, blends of KCl and K2SO4 (a physical mixture of the two less expensive sources in the market) could be an alternative to reduce Cl to thresholds that do not affect the quality of the coffee beverage without excessively increasing the costs of the fertilization.

Therefore, the present study had the objective to evaluate the effect of blends of KCl and K2SO4 fertilizers at different proportions and their influence on the yield and the nutritional state of coffee plants, as well as on the chemical composition and quality of the coffee beverage.

#### **2. Results**

#### *2.1. Effects of the KCl and K2SO4 Blends in the Stocks of Cl in the Soil, Nutrition, and Yield of Coffee Plants*

#### 2.1.1. Harvest of 2017/2018

The initial content of K in the 0–20 and 20–80 cm layers was 91.5 and 58.6 mg dm−<sup>3</sup> while stocks of the element were 201.3 and 386.7 kg ha−1, respectively. The content of Cl in the 0–20 and 20–80 cm layers was 153.8 and 203.8 mg dm−<sup>3</sup> while stocks were 338.3 and 1345.0 kg ha−1, respectively. K and Cl contents in the leaves were 19.2 g kg−<sup>1</sup> and 2880 mg kg<sup>−</sup>1, respectively (Table 1).

**Table 1.** Initial content and stocks of K and Cl in the soil and in the leaves of coffee plants.


BD = Soil density by the volumetric ring method; S.K+ = stocks of K; S.Cl<sup>−</sup> = stocks of Cl. Both stocks were calculated by multiplying the content of the element by the mass of soil in the layer.

Soil stocks of Cl were influenced by the K blends (Figure S1). Overall, the amount of Cl decreased along with the KCl proportion in the treatment. In the 0–20 cm layer, Cl stocks with T1, T2, and T3 were similar (~190 kg ha−1). In the 20–80 cm layer, the highest Cl stock was in the T1 (622 kg ha<sup>−</sup>1), T2 (513 kg ha−1), and T3 (409 kg ha−1; which did not differ from T2). Other treatments showed similar Cl stocks.

In this harvest, K content varied from 20.8 to 34.8 g kg−<sup>1</sup> (Figure S2). Cl content decreased with the increase in K2SO4, with results ranging from 3644 to 5275 mg kg−1. For the statistically non-significant variables, mean values for Cl content in the beans (Figure S3A), yield (Figure S3A), and Cl removal (Figure S3B) were 1778 mg kg<sup>−</sup>1, 3631 kg ha−<sup>1</sup> (Figure S3A), and 2.8 kg ha−1, respectively. The lowest K removal by the beans was in T3 (11.3 kg ha−1). The K removal from other treatments had means close to 24 kg ha−<sup>1</sup> (Figure S3B).

#### 2.1.2. Harvest of 2018/2019

In this harvest, treatments showed Cl stocks near 65 kg ha−<sup>1</sup> in the 0–20 cm layer (Figure 1). The lowest Cl amount was stocked with T4 (50 kg ha<sup>−</sup>1). In the 20–80 cm layer, the average stock of the six treatments was 122 kg ha<sup>−</sup>1.

**Figure 1.** Stocks of Cl in the 0–20 and 20–80 cm layers after application of KCl and K2SO4 blends as cover fertilization on coffee plants. 2018/2019 harvest. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

K content in the leaves varied from 17 to 21 g kg−1, and the lowest value was in the control (Figure 2). Overall, Cl content in the leaves decreased with less KCl applied. Treatments T1 (6950 mg kg<sup>−</sup>1) and T2 (7621 mg kg−1) were far superior from the others.

There was no significant differences among treatments for the following variables: Cl content in beans (693 mg kg<sup>−</sup>1; Figure 3A), yield (1804 kg ha−1; Figure 3A), and K removal (6.0 kg ha<sup>−</sup>1) and Cl removal (0.21 kg ha−1; Figure 3B).

**Figure 2.** K and Cl contents in the leaves of coffee plants, 20 days after application of the second cover fertilization parcel. Harvest of 2018/2019. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

**Figure 3.** Yield of coffee plants and Cl content in the beans at cherry stage (**A**) and K and Cl removal by the beans (**B**) after application of blends of KCl e K2SO4 as cover fertilization. Harvest of 2018/2019. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

#### 2.1.3. Harvest of 2019/2020

The same pattern of the first harvest was observed. Cl stocks in the soil and Cl content in the leaves decreased along with the proportion of KCl in the blend (Figures 4 and 5). In the 0–20 cm layer, Cl stocks were higher for T1 (119 kg ha−1) and T3 (101 kg ha−1) and lower for T5 (57 kg ha−1) and the control (54 kg ha−1); in the 20–80 cm layer, the lowest values occurred with T5 and control (~267 kg ha−1). The highest Cl content in the leaves was in T1 (4919 mg kg−1), and the lowest content was found in T5 (1762 mg kg−1) and control (1819 mg kg<sup>−</sup>1).

**Figure 4.** Stocks of Cl in the 0–20 and 20–80 cm layers after application of blends of KCl e K2SO4 as cover fertilization. Harvest of 2019/2020. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

**Figure 5.** K and Cl content in the leaves of coffee plants. Harvest of 2019/2020. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

There were differences in the yield of the coffee plants depending on the treatment, although they all received the same dose of K (Figure 6A). Yields of T1 (4147 kg ha<sup>−</sup>1) and the control (4055 kg ha<sup>−</sup>1) were the lowest. The treatments that received K2SO4 up to 75% of applied K had similar yields (~5100 kg ha−1). Yields of T2, T3, and T4 were 19, 24, and 24% higher than the yield of T1. Treatment T5 yield was similar to the best yields but did not differ from the yield of T1.

**Figure 6.** Yield (**A**), Cl content in the beans (**A**), and removal of K and Cl (**B**) by the beans of coffee at cherry stage after application of blends of KCl e K2SO4 as cover fertilization. Harvest of 2019/2020. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

There was no significant difference among treatments for the other agronomic variables. The average results were as follows: 883 mg kg−<sup>1</sup> for the content of Cl in the beans (Figure 6A), 45 kg ha−<sup>1</sup> for the K removal, and 1.6 kg ha−<sup>1</sup> for the Cl removal by the beans (Figure 6B).

#### *2.2. Effect of KCl and K2SO4 Blends on the Chemical Composition and Quality of the Coffee Beverage*

#### 2.2.1. Harvest of 2017/2018

The highest K leaching (KL) was in T3 (36.7 μg g−1), and the lowest was in T1 (30 μg g<sup>−</sup>1) (Figure S4). In the variables related to the quality of the coffee beverage, there was a similar pattern among the treatments despite not having differences from each other. The resulting averages were: 81 points for the sensorial analysis; 91.8 μS cm−<sup>1</sup> g−<sup>1</sup> for the electric conductivity (EC); 9.7% for total sugars (TS); 1.04% for caffeine content (Caf); 47.6 u min−<sup>1</sup> g−<sup>1</sup> for the activity of polyphenol oxidase (PPO); 186.5 mL NaOH 100 g−<sup>1</sup> of sample for total titratable acidity (TTA); and 6.4% for polyphenols (Pol) (Table S1).

#### 2.2.2. Harvest of 2018/2019

All treatments received over 80 points in the cup quality (Figure 7A). The highest scores were achieved by treatments T3 (83 points), T4 (84.5), and T5 (83).

**Figure 7.** Scores of cup quality (**A**), electric conductivity and potassium leaching (**B**) in coffee beans at cherry stage after application of blends of KCl e K2SO4 as cover fertilization. Harvest of 2018/2019. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

The EC was higher in T1 (221 μS cm−<sup>1</sup> g<sup>−</sup>1) and lower in T5 (132 μS cm−<sup>1</sup> g<sup>−</sup>1) and the control (117 μS cm−<sup>1</sup> g<sup>−</sup>1). T2 was 24% lower than T1 (Figure 7B). Potassium leached (KL) more in treatments where the proportion of KCl was higher than K2SO4 (Figure 7B). T1, T2, and T3 had similar KL (~37 μg g−1) while the other treatments were lower (~29 μg g<sup>−</sup>1). There was no significant variation for the other variables, and the means were: 9.1% for TS, 1.03% for Caf, 46 u min−<sup>1</sup> g−<sup>1</sup> for PPO, 190 mL NaOH 100 g−<sup>1</sup> of sample for TTA, and 5.0% for the content of Pol (Table 2).


**Table 2.** Variables analyzed in the coffee beans for the harvest of 2018/2019.

CV (%) = coefficient of variation; Pol = total phenolic compounds (%); TS = content of total sugars (%); Caf = content of caffeine (%); PPO = polyphenol oxidase activity (u min−<sup>1</sup> g−1); TTA = total tritable acidity (mL NaOH 0.1 N 100 g−1). Means followed by the same letter in column do not differ according to Tukey's test (*p* < 0.05). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

#### 2.2.3. Harvest of 2019/2020

The highest grade was achieved in T5 (89 points) and the lowest was in the control (84) (Figure 8). However, T1 was not different from the control. T3 and T4 had similar scores (86 points) while T2 was similar to T3 and T4, but not different from T1. The other variables were not influenced by the application of the blends of K. The following means were found: 124 μS cm−<sup>1</sup> g−<sup>1</sup> for EC, 9.6% for TS, 1.02% for Caf, 54 u min−<sup>1</sup> g−<sup>1</sup> for PPO, 70.9 μg g−<sup>1</sup> for KL, 195 mL NaOH 100 g−<sup>1</sup> of sample for TTA, and 5.0% for Pol (Table 3).

**Figure 8.** Scores of the cup quality of coffee beans at cherry stage after application of blends of KCl e K2SO4 as cover fertilization. Harvest of 2019/2020. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

**Table 3.** Variables analyzed in the coffee beans for the harvest of 2019/2020.


CV (%) = coefficient of variation; Pol = total phenolic compounds (%); TS = content of total sugars (%); Caf = content of caffeine (%); PPO = polyphenol oxidase activity (u min−<sup>1</sup> g−1); TTA = total tritable acidity (mL NaOH 0.1 N 100 g−1). Means followed by the same letter in column do not differ according to Tukey's test (*p* < 0.05). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O.

2.2.4. Principal Component Analysis (PCA) for the Agronomic Variables, Chemical Composition of the Beans, and Quality of the Coffee Beverage

The PCAs allow one to understand the behavior of the variables related to the chemical composition and quality of the coffee beverage in relation to the treatments, even if some of the variables were not statistically significant. Initially, we attempted to separate the treatments with ellipses in the PCAs calculated with all available data. However, we decided to add these PCAs in the Supplementary Material as we observed a lower percentage of the explained variance in comparison with the PCAs calculated without the agronomic data (those related to the soil). This happens when a set of variables with very different origins (soil, beverage quality, leaf composition) are used in PCA calculations (S5, S6, and S7).

Therefore, the treatments in the following PCAs represent the proportions of KCl and K2SO4 as described before. Thus, the closer they are, the greater the correlation between the variables that constitute these treatment groups. The agronomic variables stocks of K and Cl in the 0–20 and 20–80 cm layers, K and Cl contents in the leaves, yield, K and Cl removal in the beans and Cl content in the beans are supplementary variables; that is, they do not contribute to explaining the variability of the data. These illustrative variables are represented as dashed arrows, and they help to interpret the other data.

The PCA for the first harvest (2017/2018) indicates that the two components (Dim1 and Dim2) responded for 51.7% of the total variability of the data. The variances explained by these two variables were 38.2% and 18.9%, respectively (Figure S8). The variables K and Cl removal by the beans, K content in the leaves, yield, TS, Pol, and Caf were highly correlated. The control treatment was more related to these variables. In addition, these variables were negatively correlated to the stock of K in the 20–80 cm layer, stock of Cl in the 0–20 cm layer, and cup quality. The EC, KL, and TTA variables were highly correlated. The variables Cl content in the beans, stock of K in the 0–20 cm layer, and Cl content in the leaves were lowly correlated with the stock of Cl in the 20–80 cm layer and with the PPO activity. Furthermore, these variables were negatively correlated with the variables EC, KL, and TTA.

In the harvest of 2018/2019, the two PCA components explained 54.9% of the variability. The variances of each component were 39.6% and 15.3%, respectively (Figure 9). The cup quality was positively correlated with TS in the coffee beans and with yield. These variables are also correlated with T2. The KL variable was correlated with the stocks of Cl in both layers and with the Cl content in the leaves of the plants. T1 and T4 were close to these variables. Pol, Caf, and K and Cl in the beans were strongly correlated. Overall, the PCA shows that the quality of the coffee beverage is negatively correlated with the content of Cl in the leaves and beans.

**Figure 9.** Principal component analysis for the harvest of 2018/2019. PPO = activity of the enzyme polyphenol oxidase; KL = K leaching; TTA = total titratable acidity; EC = electric conductivity; Pol = total phenolic compounds; Caf = content of caffeine; TS = content of total sugars; K in the soil 20 = stock of K in the 0–20 cm layer; K in the soil 80 = stock of K in the 20–80 cm layer; Cl in the soil 20 = stock of Cl in the 0–20 cm layer; Cl in the soil 80 = stock of Cl in the 20–80 cm layer, K rem. by beans: K removal by the beans, Cl rem. by beans: Cl removal by the beans. T1: 100% KCl, T2: 75% KCl + 25% K2SO4, T3: 50% KCl + 50% K2SO4, T4: 25% KCl + 75% K2SO4, T5: 100% K2SO4.

In the last harvest (2019/2020), the PCA explained 56.8% of the variability with the first component explaining 36.8%, and the second component explained 20% of the variance (Figure 10). The cup quality and PPO activity were closely related to T5. The stock of Cl in the 20–80 cm layer was strongly related to T1. KL and Cl contents in the leaves were also correlated with T1.

**Figure 10.** Principal component analysis for the harvest of 2019/2020. PPO = activity of the enzyme polyphenol oxidase; KL = K leaching; TTA = total titratable acidity; EC = electric conductivity; Pol = total phenolic compounds; Caf = content of caffeine; TS = content of total sugars; K in the soil 20 = stock of K in the 0–20 cm layer; K in the soil 80 = stock of K in the 20–80 cm layer; Cl in the soil 20 = stock of Cl in the 0–20 cm layer; Cl in the soil 80 = stock of Cl in the 20–80 cm layer, K rem. by beans: K removal by the beans, Cl rem. by beans: Cl removal by the beans. T1: 100% KCl, T2: 75% KCl + 25% K2SO4, T3: 50% KCl + 50% K2SO4, T4: 25% KCl + 75% K2SO4, T5: 100% K2SO4.

#### **3. Discussion**

Despite the long-time fertilization with KCl in the area, the initial K stocks in the soil were considered medium level [17]. For the initial amount of Cl, however, there is no method of extraction and no reference values to relate to the needs of coffee plants. Cl is a micronutrient that is required in low amounts by plants. Under field conditions, Cl deficiency is uncommon while the excess is frequently expressed.

The stocks of Cl reduced during the three years of study due to the leaching of the element to deeper layers in the soil. The Cl ion has low interaction with the soil solid phase [18]; thus, it is easily leachable [12].

There was a tendency to accumulate K in the leaves when plants received more KCl. KCl fertilizer is more soluble than K2SO4. Nonetheless, in all harvests, the foliar content of K remained adequate in the range of 19.7 to 31 g kg−<sup>1</sup> [19,20] except for the low content in the control treatment in the last two harvests.

The Cl content in the leaves in all harvests was reduced from T1 to T5 and the control. The content of Cl usually found in plant tissues ranges from 2000 to 30,000 mg kg−1, which is equivalent to the amount of macronutrients [21,22]. However, plants vary in their tolerance to Cl [23]. According to Marschner [22], plants sensitive to Cl show toxicity symptoms at concentrations higher than 3500 mg kg−1. In tolerant plants, the symptoms appear when the concentration range from 20,000 to 30,000 mg kg<sup>−</sup>1.

Under field conditions, toxicity symptoms caused by Cl excess are uncommon. Symptoms are characterized by the reduction of the width of the leaves, with possible curling, and the presence of wide necrosis with later leaf drying [11,12]. In this study, despite the high content found when KCl was applied (over 2500 mg kg−1), plants did not show toxicity symptoms. However, it is important to emphasize the damages to the metabolism, growth, and yield that can occur even in concentrations below the toxicity threshold. In fact, in the harvest of 2019/2020, when the foliar content of Cl reached 4919 mg kg−<sup>1</sup> in treatment T1, a lower yield was observed. An argument could be made for the higher availability of S in the treatments that received more K2SO4, but despite the source of K, all treatments received 2 t ha−<sup>1</sup> of gypsum, which provided 340 kg ha−<sup>1</sup> of S to the soil. The reduction in the yield is probably more related to the excess of Cl than the lack of S in the fertilization. Another aspect to be taken into account is that, under high concentrations of Cl in the soil, anion–anion competition may occur mainly with phosphate and nitrate ions. This is due to the inability of proteins to differentiate among nitrate, phosphate, and chlorine ions, leading to the absorption of the ion in higher concentration [12].

Conversely, the reduction in the yield is notable, even with no clear statistical separation, between the treatment that received only K2SO4 and the treatment fertilized solely with KCl. It is possible that such difference may be related to the solubility of the K2SO4, 80 g L−<sup>1</sup> at 25 ◦C, which is considerably lower than the solubility of the KCl, 279 g L−<sup>1</sup> [24]. This difference in the availability of K can compromise the yield in harvests of increased productivity as the harvest of 2019/2020. Another explanation is that the high solubility of KCl can benefit the absorption of cations, such as K, Ca, and Mg [25], increasing plant nutrition, even though it is for a short period. A third possibility is the excess of SO4 <sup>2</sup>−, limiting the availability and absorption of H2PO4 − by the plant, since, besides the K2SO4 application, gypsum was also applied.

The results suggest advantages in providing the two sources of K (25 to 75% of K2SO4) to increase yield. In the first two years of the experiment, when the yields were lower, the exportation of K by the beans was less intense and plant production was not limited by the sources of K once they were applied at the same dose.

The removal of K and Cl and the content of Cl in the beans were not different among the treatments since these elements remain in high concentration in the mucilage and the bean peels [26]. The exception is treatment T3 at the first harvest, but that might be related more to the history of the area than to the treatments.

#### *3.1. Effects of the Application of KCl and K2SO4 Blends in the Chemical Composition of the Beans and in the Quality of the Coffee Beverage*

There was a response in the KL in the first and second harvests. This variable is related to the integrity of the cell wall and membrane and, consequently, to the coffee beverage quality. When these structures are less intact, the cell has a higher tendency to lose cytoplasmatic contents as a reflection of the reduced cell organization [6,19,27]. The KL results for the first year of the evaluation showed the opposite effect to what would be expected, but this lower value observed for the T1 treatment is due to the influence of frequent fertilization with KCl before the evaluation. In the second year of evaluation, after the establishment of a new K dynamic in the soil and the reduction of Cl levels, less KL was found in treatments T4, T5, and the control.

Another piece of evidence for the reduction in the quality of the coffee beans and beverage with increasing doses of KCl is the high values of the EC observed in treatments T1, T2, and T3. As KL, CE also has a direct relationship with the integrity of the cell membrane [28,29].

Despite being considered indicatives of the quality of the beverage, these variables should not be decisive to vouch for the quality of the coffee [28]. In fact, the results of the cup quality in the last harvest suggest the same tendency observed for these variables. Notably, there was a response in the cup quality after the application of the treatments in the second harvest. As previously stated, in the first harvest, all response variables were very dependent on the previous fertilization in the area, thus the lack of response in the sensorial analysis.

However, from the second year of evaluation, some important facts should be emphasized about the K nutrition with the blends of fertilizers and the quality of the coffee beverage. Despite the lower scores for T1 and T2, the same behavior was observed for the control without K fertilization. This result suggests that only reducing the application of Cl via KCl is not enough to improve the quality of the beverage, but also maintaining adequate levels of K is essential to produce a high-quality coffee [6,15].

In the last harvest, treatment T5 achieved the highest score (89 points) in the sensorial analysis. T3 and T4, however, reached a few points less (86 points) than T5. Considering the higher cost of K2SO4 in relation to KCl, the choice for the composition of the K fertilizer should consider the economic cost that this difference of 3 points in the cup quality might return. Another important consideration is that there was a tendency for higher yield in T3 and T4 treatments despite the lack of statistical differences among the treatments. The difference between both treatments in relation to treatment T5 yielded more than five sacks of 60 kg of coffee beans, suggesting that yield should also be considered when choosing the best K fertilizer composition.

#### *3.2. Principal Component Analyses for the Agronomic Variables and the Quality of the Coffee*

The PCA results suggest that studies on how the fertilization of coffee plants affects the quality of the coffee beverage should be carried out for a long duration.

Overall, there were increased effects of the treatments after the second year of evaluation, probably due to the previous fertilizations with KCl, which is the most used source of K in Brazil [15]. However, some points should be considered in relation to the first harvest, such as the correlation among the variable Cl content in the beans, K content in the leaves, and yield showing the importance of the K fertilization in coffee plants. Nonetheless, the negative correlation of these variables with the cup quality suggests an unfavorable effect of one of these variables on the quality of the beverage, most probably the Cl content in the beans.

The correlation between EC and KL can be explained by the direct relationship shared by these two variables since they both indicate damages to the cell integrity of the beans [28–30]. The PCA confirms these results. These damages can lead to the loss of compounds related to the quality of the beans and the cup quality; therefore, lower EC and KL indicate lower coffee quality [28,31]. This lower bean quality is confirmed by the high negative correlation between PPO activity and the variables EC, PL, and TTA, showing that higher values of EC, KL, and TTA are associated with low PPO activity. Several studies found a positive correlation between the PPO activity and the sensorial quality of the coffee [32,33]. Thus, it is possible to conclude that there is a reduction in the PPO activity and the quality of the beverage as EC, KL, and TTA increase. In fact, damages to the cell membrane lead to the loss of selective permeability, facilitating the reaction of PPO with the phenolic compounds (the specific substrate of this enzyme). This reaction produces quinones that inhibit the activity of PPO [16,32].

Noteworthy, the high positive correlation between cup quality and the content of TS in the beans indicates a direct relationship where the increase in TS also increases the cup quality. Treatment T2 was close to this correlation, confirming the results for cup quality and indicating that increases in the K2SO4 proportion tend to increase cup quality. These results confirm the study of Silva et al. [16], which also involved doses and sources of K, and they add information about the sensorial quality of the coffee beverage. These findings provide evidence for the increase in the content of TS and better scores on the cup quality test as the proportion of K2SO4 in the blend also increases.

The reduction in the quality of the coffee beans is observed in the high correlation among KL, the stock of Cl in the soil, and the Cl content in the leaves. When the values of the variables related to Cl increase, K also increases, and the quality of the beans and the beverage reduces.

In the last harvest (2019/2020), a direct effect of the Cl in the variables related to the quality of the beverage is notable. Despite the fact that the PPO activity did not show differences for the treatments in both harvests, this variable behavior within the PCA is enough to indicate the direct relationship of this enzyme with the quality of the coffee beverage.

In addition, the high negative correlations of cup quality and PPO in relation to the variables related to Cl (stocks in the soil and content in the leaves) confirm the negative influence of the Cl in the beverage. In addition, treatment T5, which received only K2SO4, is closely related to cup quality and PPO activity.

Previous reports state that Cl increases the water content of the coffee fruits with consequent microbial fermentation [13,14]. We believe this explanation does not relate to this study since the beans were collected manually at the stage of cherry and benefited under controlled conditions, unlikely leading to an undesirable fermentation.

Although we did not perform physiological or morphological analyses of the coffee beans, the results allow us to infer that there is an effect of the Cl in the beans and that it might be related to the loss of quality of the beverage. The rationale is that the Cl can inhibit the activity of the PPO enzyme when reacting with the copper activator, thus reducing the enzymatic activity when KCl is applied [34].

In conclusion, this study showed that the blends of K fertilizers responded positively to the quality of the coffee beverage when the proportion of K2SO4 relative to KCl was increased. There was a tendency for higher KL, EC, and TTA with the increase of the KCl proportion, which might lead to damages in the cell membrane caused by the Cl and the consequent reduction of the PPO activity and quality of the beverage. However, the decision to use a determined blend of K should consider the improvement of the beverage and the economic return to the farm. Moreover, it should also consider the yield. For example, in this study, the blend that provided the best quality for the coffee beverage was not always the same responsible for the highest yield. And finally, it should also consider the economic costs of fertilization with K2SO4, which is more expensive than KCl. Considering these aspects, a management strategy could be the separation of the farm into plots based on the tendency to produce better quality coffees in previous years. In this case, each plot would receive a determined blend of K, and the highest proportions of K2SO4 should be applied to the plots with a tendency to produce higher quality coffees while the higher proportion of KCl would fertilize plots of low-quality coffee. Finally, any investigation with similar objectives to this work should perform a broader study, especially covering the regions of greater coffee production, where each plot/farm or region would receive potassium fertilization depending on the tendency for better or worse quality of the coffee beverage in the cup.

#### **4. Materials and Methods**

#### *4.1. Experimental Area Characterization*

The experiment was performed through three consecutive years (harvests of 2017/2018, 2018/2019, and 2019/2020) in a commercial production system of coffee located in the municipality of Santo Antônio do Amparo-MG, Brazil (20◦53 26.04" S and 44◦52 04.14" W and mean altitude of 1100 m). The plantation of Coffea arabica L., cultivar Catuaí Vermelho IAC 99, initiated in 2012 and spaced at 3.40 m × 0.65 m, is planted on a clayey Dystrophic Red Latosol-Latossolo Vermelho distrófico (Oxysol) [35].

Before the experiment, soil samples were collected for chemical attributes and texture analyses (Table 4). Samples from the 0–80 cm layer of soil were collected to assess K and Cl stocks. Undisturbed soil samples were taken to assess bulk density (BD). For depths over 5 cm, multiple samples were taken followed by the weighted average of the BD values. After determining K and Cl concentrations (mg kg−1), the values were multiplied by the BD to transform them into kg ha<sup>−</sup>1.


**Table 4.** Soil analyses results on September 2017.

P, K+, Fe2+, Zn2+, Mn2+, Cu2+—Mehlich extractor. Ca2+, Mg2+, Al3+—1 mol L−<sup>1</sup> KCl extractor. H<sup>+</sup> + Al3+—SMP extractor. B—hot water extractor. S—monocalcium phosphate in acetic acid extractor. BS = exchangeable bases sum. ECEC = effective cation exchange capacity. CEC = cation exchange capacity at pH 7.0. V = base saturation. m = aluminum saturation. P-rem = remaining phosphorus. OM = organic matter (oxidation with Na2Cr2O7 0.57 mol L−<sup>1</sup> + H2SO4 5 mol L−1).

#### *4.2. Experimental Design*

The experimental design was in randomized blocks, with four blocks disposed at 90 degrees with the slope of the area. The treatments were composed of blends of KCl and K2SO4 (both in terms of K2O) as follows: T1—100% as KCl; T2—75% as KCl + 25% as K2SO4; T3—50% as KCl + 50% as K2SO4; T4 –25% as KCl + 75% as K2SO4; T5: 100% as K2SO4; and a control without K2O application. Each plot was composed of three planting lines with 16 plants, and the 10 central plants were considered a useful area (Figure 11).

**Figure 11.** Schematic representation of the experimental design, number of plants in each plot, and the useful area used to collect the data.

#### *4.3. Experiment Conducting*

4.3.1. Liming, Fertilization, and Gypsum Application

After the coffee harvest of each studied year, soil samples from the 0–10 cm, 0–20 cm, and 20–40 cm layers were collected to evaluate the needs for liming, fertilization, and gypsum application, respectively [17]. Liming was applied at 1.0 t ha−1, 1.2 t ha−1, and 1.5 t ha−<sup>1</sup> on the first, second, and third harvest years, respectively. Gypsum was applied at 1.1 t ha−<sup>1</sup> and 2.0 t ha−<sup>1</sup> in the second and third years, respectively. Both dolomite lime and gypsum were applied underneath the projection of the tree canopies. P was applied at 120, 90, and 90 kg ha−<sup>1</sup> of P2O5 as triple superphosphate on each consecutive year. N was applied at 350, 350, and 400 kg ha−<sup>1</sup> of N as ammonium nitrate on each consecutive year, divided into three applications.

#### 4.3.2. Potassium Fertilization

Before the experiment, the saline index of each blend of K fertilizer was determined by comparing a 10 g L−<sup>1</sup> sodium nitrate solution with solutions prepared with the blends of KCl e K2SO4 at the same concentration (Jackson 1958). The electric conductivity of the solutions and the saline index were calculated according to the equation: SI = [((ECa))⁄((ECb))] × 100, where SI is the saline index, ECa is the electric conductivity of the sample, and ECb is the electric conductivity of the sodium nitrate solution. The SI found were 142, 130, 121, 112, and 104%, for T1, T2, T3, T4, and T5, respectively.

The maintenance fertilization was done according to Guimarães et al. [36] using the abovementioned blends. The doses of K2O applied were 150, 200, and 300 kg ha−<sup>1</sup> for the respective agricultural years of 2017/2018, 2018/2019, and 2019/2020. All K fertilizations were divided into three applications.

4.3.3. Agronomic Variables Assessed on the Three Harvests

• K and Cl content in the leaves

The third and fourth pair of leaves on both sides of the plants were collected from the useful area 20 days after the second application of the cover fertilization. The leaves were washed in deionized water, dried at 65 ◦C, and grounded in a Willey mill. The plant material was digested in a solution of nitric-perchloric acid (4 parts of nitric acid to 1 part of perchloric acid), and K was determined with inductively coupled plasma (ICP). To determine Cl, 1 g of grounded material was added to 50 mL of ultrapure water under agitation for 15 min [37]. After filtering the extract, the content of Cl (mg kg−1) was determined with a selective electrode (Hanna®, model HI4107) coupled to a Hanna® device, model HI2221. The determination curve was built using the concentrations of 2, 20, 200, and 1000 mg L−<sup>1</sup> of Cl.

• Yield

The harvests were done when more than 70% of the fruits were mature. For the chemical analyses, 4 L of beans at the cherry stage were collected two days before each harvest. Fruits were peeled with an electric peeler (Pinhalense®, model DPM-02) and submerged for 24 h to remove the mucilage. After removing the peels and the rotten beans, samples were air-dried until a 10.8% to 11.2% moisture level.

The yield was determined by harvesting all fruits in the useful area. After the harvest, 5 L of a mix of fruits in every stage of maturation were air-dried under sunlight for one day. When the samples reached around 12% of moisture, beans were peeled and weighted. The moisture level was then adjusted to 12%, which is considered adequate for commercialization. To estimate yield, the weight of the beans in the useful area was projected to the number of plants in one hectare (4524 plants).

• K and Cl content and removal in the beans

K and Cl contents in the beans were determined at the cherry stage after air-drying (65 ◦C, until constant weight) and grounding the beans in a Willey mill. K content was determined after nitric-perchloric digestion with measures done by ICP. Cl content followed the same procedures to quantify Cl in the leaves. The amounts of these elements removed from the soil were obtained by multiplying their content in the beans by the yield on each treatment.

• Stocks of Cl in the soil

Stocks of Cl in the 0–20 and 20–80 cm layers of soil were checked during the experiment. Six soil samples were taken from the soil underneath the projection of the tree canopies (three from each side of the parcel). Extraction and determination of Cl followed the same procedure described for leaf Cl content, but with the proportion of 10 g of soil to 50 mL of ultrapure water. The stocks were determined by multiplying the element concentration by the mass of soil in each layer.

The analytical standard Tomato leaves (NIST 1573A), with 0.66% of Cl, was used in both soil and plant material analysis. The mean recovery of Cl was higher than 92%, assuring that the extraction and determination used for Cl were effective for both soil and plant material.

• Chemical analysis of the beans and coffee sensorial analysis

After benefiting the coffee samples, the beans were stored in paper bags in a cold chamber until the chemical and sensorial analyses. The chemical analyses were performed at the Laboratory of Coffee Quality Analysis in the Empresa de Pesquisa Agropecuária de Minas Gerais (EPAMIG). Potassium leaching (KL, in μg g−1) was determined after 5 h of soaking [29], and electric conductivity (EC, in μS cm−<sup>1</sup> g−1) was determined according to Loeffler et al. [38]. The total titratable acidity (TTA, m mL NaOH 0.1 N 100 g−1) was done according to Carvalho et al. [32] in the adaptation of the methodology from the Association of Official Analytical Chemists [39]. The content of total sugars (TS, in %) followed the anthrone method [40]. The activity of the polyphenol oxidase enzyme (PPO, in u min−<sup>1</sup> g<sup>−</sup>1) was determined according to Carvalho et al. [32]. Total phenolic compounds (Pol, in %) were extracted according to Goldstein and Swain [41] and determined by the Folin-Denis method, described by AOAC [39]. Caffeine content (Caf, in %) was determined by spectrophotometry at 273 nm [42]. The coffee beans were frozen in liquid nitrogen and grounded in an IKA mill for the analyses, except for the KL and EC determinations.

The sensorial analysis (cup quality) was performed at the Laboratory of Agricultural Products Processing in the Universidade Federal de Lavras following the Specialty Coffee Association of America (SCAA) protocol. Three professionals with skills to differentiate fragrances, characteristics, and flavors participated in the cup test. The evaluation was based on scores given to the following attributes: fragrance/aroma, uniformity, clean cup, sweetness, flavor, acidity, body, aftertaste, balance, defects, and overall. The coffees were classified as the SCAA [43] according to their final scores (Table 5).


**Table 5.** Coffee beverage classification according to the cup quality.

Source: Specialty Coffee Association of America (SCAA) (2009).

#### • Statistical analyses

After model validation and analysis of variance indicating differences among treatments (*p* < 0.05), the response variables were submitted to Tukey's test (*p* < 0.05) on the R 3.3.1 environment [44]. Principal component analyses (PCA) were performed to correlate the agronomic variables with the coffee beverage variables and yield. In the PCA, two components (Dim1 and Dim2) were used to represent the total data variability. The package Facto MineR (version 1.42) was used in the R software.

#### **5. Conclusions**

The activity of the polyphenol oxidase enzyme and the cup quality indicate that the ion Cl- reduces the quality of the coffee beverage. The increased application of the Clion increases KL, EC, and TTA, indicators of the loss of coffee quality. K content in the leaves was not influenced by the application of blends of K fertilizer while Cl content increased linearly with KCl applied. The application of KCl and K2SO4 blends influenced coffee yield and the optimum proportion was 25% of KCl and 75% of K2SO4. The highest score in the cup quality test was observed with 100% K2SO4. However, other blends

showed close scores. The decision for the fertilizer should consider the cost of the K source. KL and EC can indirectly show that the Cl can damage the coffee beans and reduce the selective permeability of the cell membrane, with possible negative consequences to the coffee beverage.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/plants12040885/s1, Table S1. Variables analyzed in the coffee beans for the harvest of 2017/2018; Figure S1. Stocks of Cl in the 0–20 and 20–80 cm layers after application of KCl and K2SO4 blends as cover fertilization on coffee plants. Harvest of 2017/2018. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O; Figure S2. K and Cl content in the leaves of coffee plants, 20 days after application of the second cover fertilization parcel. Harvest of 2017/2018. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O; Figure S3. Yield of coffee plants and Cl content in the beans at cherry stage (A) and K and Cl removal by the beans (B) after application of blends of KCl e K2SO4 as cover fertilization. Harvest of 2017/2018. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O; Figure S4. Potassium leaching in coffee beans at stage of cherry after application of blends of KCl e K2SO4 as cover fertilization. Harvest of 2017/2018. Means followed by the same letter in the column do not differ according to Tukey's test (*p* < 0.05). Vertical bars indicate the standard error of the mean (*n* = 4). T1: 100% KCl; T2: 75% KCl + 25% K2SO4; T3: 50% KCl + 50% K2SO4; T4: 25% KCl + 75% K2SO4; T5: 100% K2SO4; control did not receive K2O; Figure S5. Principal component analysis for the harvest of 2017/2018. PPO = activity of the enzyme polyphenol oxidase; KL = K leaching; TTA = total titratable acidity; EC = electric conductivity; Pol = total phenolic compounds; Caf = content of caffeine; TS = content of total sugars; K in the soil 20 = stock of K in the 0-20 cm layer; K in the soil 80 = stock of K in the 20-80 cm layer; Cl in the soil 20 = stock of Cl in the 0-20 cm layer; Cl in the soil 80 = stock of Cl in the 20-80 cm layer, K rem. by beans: K removal by the beans, Cl rem. by beans: Cl removal by the beans. T1: 100% KCl, T2: 75% KCl + 25% K2SO4, T3: 50% KCl + 50% K2SO4, T4: 25% KCl + 75% K2SO4, T5: 100% K2SO4; control did not receive K2O. Figure S6. Principal component analysis for the harvest of 2018/2019 considering agronomic data. Figure S7. Principal component analysis for the harvest of 2019/2020 considering agronomic data. Figure S8. Principal component analysis for the harvest of 2017/2018.

**Author Contributions:** Conceptualization, D.G., T.H.P.R., T.R.d.S., V.R.B. and C.S.; methodology, D.G. and C.S.; software, M.G.M.G.; validation, D.G., C.S., M.R.M. and D.P.O.; formal analysis, M.G.M.G. and C.S.; investigation, C.S., T.L.d.S., W.F.T.C., M.E.A.d.M., A.D.C.L. and L.B.d.A.; resources, D.G.; data curation, C.S., M.R.M., F.M.B., A.A.A.P., H.E.P.M.; writing—original draft preparation, C.S., D.G., T.H.P.R. and T.R.d.S.; writing—review and editing, D.G., C.S., D.P.O., M.R.M., A.A.A.P. and H.E.P.M.; visualization, C.S. and D.G.; supervision, D.G.; project administration, D.G.; funding acquisition, D.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Yara International, National Council for Scientific Development and Technology (Process: 310572/2020-7), the Agency for Improvement of Higher-Level Personnel and Minas Gerais Research Foundation.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors thank the Agency for the Improvement of Higher Education Personnel, the National Council for Scientific Development and Technology, and the Foundation for Research Support of Minas Gerais.

**Conflicts of Interest:** The authors declare that there is no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

MDPI St. Alban-Anlage 66 4052 Basel Switzerland www.mdpi.com

*Plants* Editorial Office E-mail: plants@mdpi.com www.mdpi.com/journal/plants

Disclaimer/Publisher's Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Academic Open Access Publishing

mdpi.com ISBN 978-3-0365-9583-2