Identification of growth regulators using cross-species network analysis in plants

Abstract

With the need to increase plant productivity, one of the challenges plant scientists are facing is to identify genes that play a role in beneficial plant traits. Moreover, even when such genes are found, it is generally not trivial to transfer this knowledge about gene function across species to identify functional orthologs. Here, we focused on the leaf to study plant growth. First, we built leaf growth transcriptional networks in Arabidopsis (Arabidopsis thaliana), maize (Zea mays), and aspen (Populus tremula). Next, known growth regulators, here defined as genes that when mutated or ectopically expressed alter plant growth, together with cross-species conserved networks, were used as guides to predict novel Arabidopsis growth regulators. Using an in-depth literature screening, 34 out of 100 top predicted growth regulators were confirmed to affect leaf phenotype when mutated or overexpressed and thus represent novel potential growth regulators. Globally, these growth regulators were involved in cell cycle, plant defense responses, gibberellin, auxin, and brassinosteroid signaling. Phenotypic characterization of loss-of-function lines confirmed two predicted growth regulators to be involved in leaf growth (NPF6.4 and LATE MERISTEM IDENTITY2). In conclusion, the presented network approach offers an integrative cross-species strategy to identify genes involved in plant growth and development.

Figure 1

Outline of the cross-species network approach to identify candidate growth regulators. For Arabidopsis, maize, and aspen, the expression data (Step 1) is used as input to construct a fully connected meta-network per species (Step 2). Subsequently, each meta-network is split into five DSs by applying specific density cutoffs (Step 3). These DSs are the input for two different analyses: they are used first as input to compute cross-species gene neighborhood conservation (Step 4a). Second, they are used to predict functions via GBA (Step 4b). This leads to gene function annotations of query genes based on prior knowledge on growth regulators (guide growth regulators). Edge thickness defines in which subnetwork the interaction is conserved (line thickness represents the DS and ranges from 1, the most stringent DS represented by the thickest line, to 5, the least stringent DS represented by the thinnest line). Finally, the results of these two analyses (Steps 4a and 4b) are integrated to obtain a list of candidate growth regulators (Step 5).

Figure 2

Triplets and their functional enrichments in cross-species conserved leaf networks. A, The number of triplet genes showing cross-species gene neighborhood conservation is plotted for all DSs. B, The biological process functional over-representation at each DS is summarized for two sets: (1) all triplet genes (All) and (2) growth regulators and their network neighbor (growth regulator-related) triplet genes, subset of all triplet genes. Functional categories marked with asterisks (*) belong to leaf growth modules described in Vercruysse et al. (2020b) and to the DEG sets from relevant studies on plant development (Bezhani et al., 2007; Gonzalez et al., 2010; Eloy et al., 2012; Vercruyssen et al., 2014; Vanhaeren et al., 2017). For clarity, long biological process names have been abbreviated (^§). C, Overview of growth regulators with (and without) cross-species neighborhood conservation at different DSs.

Figure 3

Recovery of RARGE II leaf trait genes for each DS split in proliferation and expansion. The dashed line indicates the leaf-related phenotype gene recovery expected by chance within the RARGE II dataset.

Figure 4

Gene-function network of the 34 phenotype-related genes out of the top 100 predicted growth regulators. Predictions are clustered by expression profile (proliferation on the left and expansion on the right). Node label colors from dark green (weak) to yellow (strong) represent the reliability of the gene prediction (GBA score). Node border colors indicate known growth regulators from Arabidopsis (black), known growth regulators from aspen (red), and Arabidopsis known growth regulator paralogs (violet). Diamonds represent TFs. Links from dark orange thick (DS1) to light orange thin (DS5) represent the DS where the genes were found connected. Genes are linked with their respective growth-related pathways (centered if connecting to both proliferation and expansion-related genes) by gray links. Anti-correlation links (connecting proliferation with expansion genes) were removed for clarity.

Figure 5

Mutants of predicted growth regulators NRT1.3 and LMI2 showed altered rosette growth. A and B, Dynamic growth analysis of PRA, compactness, and stockiness over time of wild-type Col-0 and the mutants of NRT1.3 (A) and LMI2 (B) in soil. Values are means ± SD. For phenotypic analysis of mutants of LMI2, sample sizes (n) were n = 16 for Col-0, n = 16 for lmi2-2, and n = 17 for lmi2-1. For phenotypic analysis of mutants of NRT1.3, n = 14 for Col-0, n = 15 for sper3-1, and n = 13 for sper3-3. The asterisks represent the time points at which differences in the PRA become significant between the mutants and wild-type, as determined by Student’s t test (*P < 0.05; **P < 0.01). The experiments were repeated 3 times with similar results, and one representative experiment is shown. C and D, Phenotype of 26-d-old mutants of NRT1.3 (C) and LMI2 (D). Scale bar = 1 cm.