Mycorrhizal status impacts the genetic architecture of mineral accumulation in field grown maize (Zea mays ssp. mays L.)

Arbuscular mycorrhizal fungi (AMF) establish symbioses with major crop species, providing their hosts with greater access to mineral nutrients and promoting tolerance to heavy metal toxicity. There is considerable interest in AMF as biofertilizers and for their potential in breeding for greater nutrient efficiency and stress tolerance. However, it remains a challenge to estimate the nutritional benefits of AMF in the field, in part due to a lack of suitable AMF-free controls. Here we evaluated the impact of AMF on the concentration of 20 elements in the leaves and grain of field grown maize using a custom genetic mapping population in which half of the families carry the AMF-incompatibility mutation castor. By comparing AMF-compatible and AMF-incompatible families, we confirmed the benefits of AMF in increasing the concentration of essential mineral nutrients (e.g., P, Zn, and Cu) and reducing the concentration of toxic elements (e.g., Cd and As) in a medium-input subtropical field. We characterised the genetic architecture of element concentration using quantitative trait mapping and identified loci that were specific to AMF-compatible or AMF-incompatible families, consistent with their respective involvement in mycorrhizal or direct nutrient uptake. Patterns of element covariance changed depending on AMF status and could be used to predict variation in mycorrhizal colonisation. We comment on the potential of AMF to drive genotype-specific differences in the host ionome across fields and to impact the alignment of biofortification breeding targets. Our results highlight the benefits of AMF in improving plant access to micronutrients while protecting from heavy metals, and indicate the potential benefits of considering AMF in biofortification programs.

Plant roots can take up nutrients directly from the soil (the direct pathway) or acquire 106 them via symbiosis with AMF (the AMF pathway; (Smith et al., 2003)). In the case of P, 107 specific members of the PHT1 phosphate transporter family have been identified to be 108 involved in the direct and AMF pathways, their expression and accumulation patterns transporters implicated in direct P uptake are repressed in response to AMF (Bucher, 2007;115 De Vita et al., 2018). Although currently less well defined, analogous direct and AMF 116 pathways appear to also function in N uptake, driven by specific accumulation of distinct 117 ammonium transporters (Hui et al., 2022;Koegel et al., 2013).  families. Furthermore, we tested the capacity of ionome profiles to predict mycorrhizal status 133 and assessed the potential of breeding AMF-effective hosts to achieve biofortification targets.

136
The castor mutation impacted the leaf and grain ionome of field grown plants 137 To assess the impact of AM symbioses on host mineral element accumulation in the field, we  Table 1). Element response to AMF (Mycorrhizal Response -MR; defined 149 here as the percentage increase in AMF-C compared with AMF-I) ranged from -36% to 150 116% in leaves and from -25% to 65% in grain (Fig. 1A). Zn and Cu concentrations showed 151 the highest positive response in both leaves (116% and 71%, respectively) and grain (50% 152 and 65%, respectively). The concentrations of potential toxic elements, such as As (-36%) 153 and Mn (-24%) in leaves, and As (-20%) and Cd (-23%) in the grain, were significantly 154 reduced in AMF-C plants, consistent with previous reports (Lehmann & Rillig, 2015;155 Ruytinx et al., 2020). We conducted Principal Component (PC) analyses, and AMF-C and 156 AMF-I ionomes were separated by the first PCs in both leaf and grain, which captures AM 157 effect driven by Zn, Cu, and Fe ( Supplementary Fig. S1). Although not the most responsive 158 6 element, leaf P concentration was significantly increased in AMF-C compared to AMF-I 159 families (8.5%). Grain P concentration did not change significantly with AMF status. In our 160 study field, both AMF-C and AMF-I families maintained an average P concentration within a 161 typical critical range (Gagnon et al., 2020), indicating that P limitation was not a major driver 162 of previously reported differences in plant performance (Ramírez-Flores et al., 2020). 163 In addition to impacting individual elements, AM status affected patterns of 164 covariance among elements. AMF-C and AMF-I families were more clearly distinguished by 165 leaves than grain as shown by the magnitude of MR associated with individual elements (Fig.   166 1A). This pattern was supported by significant differences between pairwise correlation 167 matrices for element concentrations in the leaves of AMF-C and AMF-I families (p < 0.001; independent of AM status (e.g., the positive correlations between Sr and Ca, or As and Fe). In 170 contrast, the sign of three correlations changed with AM status, indicating a strong effect of 171 AMF on covariance (e.g., the positive correlation between Zn and the three elements Sr, Ca 172 and Mn in AMF-C became negative in AMF-I). Ten significant correlations were specific to 173 AMF-C (e.g., the positive correlations between K and P) and 10 to AMF-I (e.g., the positive 174 correlation of P and Cu) families. The grain element correlation matrix, unlike that of leaves,  The genetic architecture of element accumulation was impacted by AMF status 185 Having observed a significant general effect of AMF status on the plant ionome, we 186 proceeded to use Quantitative Trait Loci (QTL) linkage mapping to evaluate genetic 187 differences between W22 and CML312 hosts and AMF × QTL effects. As discussed 188 previously, we equated AMF-C and AMF-I specific QTL with mycorrhizal benefit and  3. The AMF × QTL effects we detected were all conditional in nature (i.e., specific to AMF-211 C or AMF-I families). We did not find evidence of any QTL showing strong, but opposing, 212 allelic effects between AMF-C and AMF-I families (antagonistic pleiotropy. See e.g., 213 (Ramírez-Flores et al., 2020)). For the AMF-C specific QTL qCdlf/gr2.05 and qNigr9.01, the 214 effect was sufficiently strong that they were supported by Models 1 and 2, but not Model 3 215 (i.e., the interactive model did not present marked improvement over the additive model).

216
Globally, the parental allelic effect of detected QTLs varies depending on the element and 217 tissue type (Fig. 2). For example, for QTLs detected in bin 4.08 for leaf in AMF-C families, 218 the genotypes that carry the CML312 allele tend to accumulate less Ca, Fe and As, but more 219 Rb. For qCdlf/gr2.05, the allelic effect is conserved across grain and leaf in the AMF-C 220 families, where the genotypes with W22 allele tend to accumulate less Cd than genotypes 221 with the CML312 allele. In contrast, for qNigr9.01, the AMF-C with the W22 allele at this 222 QTL tends to accumulate Ni in the grain. To obtain more rigorous support for our QTL and assess the potential for AMF to impact 227 ionome variation across environments, we compared our findings to a published maize 228 multisite ionome dataset (Asaro et al., 2016). This previous work presented grain ionome 229 data for a biparental maize mapping population (hereafter, the IBM data), evaluated during  Table S1). Nine of the overlapping QTL were detected preferentially in 242 AMF-C plants in our experiments, suggesting AMF communities to be active in the IBM 243 field sites (Fig. 3B). These common QTL also showed evidence of G × E interaction in the 244 previous study (Fig. 3B). Although difficult to substantiate without further data, we speculate 245 that variation in the AMF community between field sites contributed to G × E effects seen in 246 the IBM. For example, strongly supported AMF-C conditional QTL linked to Cd and Ni on 247 chromosomes 2 and 9, respectively, were clearly recovered in IBM data from North Carolina  Having observed clear differences between AMF-C and AMF-I families, we proceeded to 257 assess whether ionome variation within the AMF-C family was a reflection of differences in 258 9 the extent of mycorrhizal colonisation. We used microscopy to quantify mycorrhizal hyphae, 259 arbuscules and vesicles from field sampled AMF-C roots in terms of percent root length 260 colonised (as reported previously, AMF-I families were free from root-internal fungal 261 structures (Ramírez-Flores et al., 2020)). We observed appreciable colonisation in field 262 sampled roots (hyphae = 23.6% ± 2%, arbuscules = 16% ± 1.4%, and vesicle = 3.4% ± 263 0.4%), with individual families ranging from 0 to 74.4% in hyphae colonisation (Fig. 4A). To qFelf4.08 and qRblf4.08 QTL were conditional on AM symbiosis and no significant 293 differences were seen between alleles in AMF-I families (Fig. 4E). The conditionality of 294 qFelf4.08 and qRblf4.08 is consistent with their capturing an aspect of mycorrhizal function 295 also reflected in arbuscule abundance. We compared direct observation and predicted values 296 of colonisation in their relationship to plant phenological traits and yield components.

297
Consistent with previous reports (e.g., (Sawers et al., 2017)), we did not find any simple 298 correlation between the abundance of root internal fungal structures and yield components. 299 We did observe greater arbuscule abundance to be correlated with accelerated flowering (Fig.   300 4F). Similar to the QTL analysis, this correlation was more significant with the predicted 301 arbuscle values than the direct observations.

372
We identified 20 element QTLs that showed evidence of interaction with AM status.

373
All AMF × QTL effects were conditional, i.e., these QTL were specific to either mycorrhizal

526
We calculated pairwise correlations among elements for both AMF-C and AMF-I 527 groups to explore changes in element correlation patterns in response to AMF colonisation.

528
Correlations between elements were performed using stats::cor with spearman correlation 529 method. To test whether correlation matrices are equal between AMF-C and AMF-I, Chi-530 square test was performed using the decorate::delaneau.test (Hoffman 2021).

531
Correlations changed in many pairs of elements from AMF-C to AMF-I in our study.

532
We then manually picked pairs of elements that showed large differences in direction or To understand the impact of AMF colonisation on the genetic architecture of element 583 accumulations and the evolvability to meet future biofortification breeding targets, we 584 applied an evolutionary quantitative genetic approach by quantifying the alignment between 585 genetic variation and the evolutionary trajectory towards maximised biofortification goals.

586
The biofortification goal was defined to obtain maximised grain Zn and Fe as set by 587 HarvestPlus as 38 μg/g dry weight for Zn and 60 μg/g dry weight for Fe (Bouis et al., 2011). 588 We extracted genetic variance-covariance matrices (G-matrix) by fitting Generalised Linear    Table S1. QTLs detected in common with a published multisite ionome analysis.