Lipidomic and metabolomic profiles of Coffea canephora L. beans cultivated in South-western Nigeria

Coffee (Coffea spp.) is one of the most popular refreshments globally. Coffee lipid diversity has untapped potential for improving coffee marketability because lipids contribute significantly to both the health benefits and cup quality of coffee. However, there have not been extensive studies of lipids of C. canephora genotypes. In this study, Ultra-performance liquid chromatography coupled with mass spectrometry (UPLC–MS) profiling of lipid molecules was performed for 30 genotypes consisting of 15 cultivated and 15 conserved genotypes of C. canephora in Southwestern Nigeria. We identified nine classes of lipids in the 30 genotypes which belong to the ‘Niaouli’, ‘Kouillou’ and ‘Java Robusta’ group: among these, the most abundant lipid class was the triacylglycerols, followed by the fatty acyls group. Although ‘Niaouli’ diverged from the ‘Kouillou’ and ‘Java Robusta’ genotypes when their lipid profiles were compared, there was greater similarity in their lipid composition by multivariate analysis, compared to that observed when their primary metabolites and especially their secondary metabolite profiles were examined. However, distinctions could be made among genotypes. Members of the fatty acyls group had the greatest power to discriminate among genotypes, however, lipids that were low in abundance e.g. a cholesterol ester (20:3), and phosphotidylethanolamine (34:0) were also helpful to understand the relationships among C. canephora genotypes. The lipid diversity identified among the C. canephora genotypes examined correlated with their overall Single Nucleotide Polymorphism diversity assessed by genotype-by-sequencing, supporting the relevance of this study to coffee cup quality improvement.

, five were found in the C. canephora genotypes studied ( Fig 1A).

184
Each class of lipid contains multiple molecules of varying lengths and degree of saturation.
185 Although these lipid classes are diverse, we constructed a simplistic figure of how they may be related to 186 each other in a metabolic network ( Fig 1B). We identified a cholesterol ester (20:3) in this work but 187 these compounds have not been well studied in plants. 221 The predominant fatty acyls among our C. canephora germplasm were linoleic acid (18:2) followed by 222 palmitic acid (16:0) (S1A Figure), which together encompassed 50% of the total fatty acids. Linoleic and 223 palmitic were also the predominant FAs in a range of C. arabica and C. canephora genotypes studied by

226
The proportion of the major FA among genotypes was also examined, because of their influence 227 on coffee quality. These FAs are also members of highly related biosynthetic pathways (Fig 3). FA (16:0) 228 i.e. palmitic acid, was higher in 'Kouillou' and 'Java Robusta' compared to 'Niaouli' (S1A Figure). This is of 229 interest because palmitic acid was positively associated with the high hedonistic values in some specialty 230 Brazilian C. arabica coffees [9]. Oleic, linoleic, and linolenic acid levels have also been identified as good 231 markers for Brazilian C. arabica coffees which had lower acidity, fragrance, body and flavor [9]. Oleic 232 acid (18:1) was 3-fold lower than linoleic acid (18:2) especially in Nia_11 (S1A Figure). Further, linolenic 233 acid (18:3) was low (10-fold) relative to linoleic acid (18:2). It is tempting to speculate that there is 234 differential regulation of the enzymes that catalyze these steps (Fig 3). In Arabidopsis these two fatty 246 genotypes had a greater proportion of high-carbon triacylglycerols i.e. C56, C58 lipids compared to 247 'Kouillou' which had more C48-C52 compounds (S1B Figure). 255 the other lipid classes (Fig 2), however, levels were notably higher in 'Niaouli' when compared to 256 'Kouillou' and 'Java Robusta' (S1B Figure). 260 Phospholipids are potentially important to food sensory perception [37,40]. PCs accumulated to higher 261 levels than LPCs which makes sense given their relative position the lipid biosynthetic pathway (Fig 1B). 285 palmitic (16:0) arachidic acids (20:0) and TG (50:1). Based on our data analysis we show in Table 4, the 286 lipids that have a high discriminatory power in the coffees in our study, were compared with 287 information from data that was published by others. Fatty acids were the most discriminatory of all 288 lipids, although the types varied with respect to the species examined (Table 4).

289
290 Table 4. Fatty acids with high discriminatory power among coffee genotypes. Included are the two main 291 varieties of C. canephora ('Niaouli' and 'Kouillou') used in this study, two cultivated Coffea species (C. arabica 292 and C. canephora) and specialty or high-quality coffee. Citations to the referenced data are included. 298 identify the lipids that contribute most to differences among genotypes (Fig 4). This analysis predicted 312 Neighbouring compounds may be correlated with each other. This is also true of metabolites that are in 313 chemical equilibrium, show mass conservation, or are highly sensitive to a common parameter [43,45].
314 Using Pearson's correlation coefficient at a cut-off of ≥ | ± 0.75|, we identified those lipids that co-  Table).

317
The most notable observation was the very strong negative correlations between CE (20:3) and 318 other lipids, many of which were uncharacterized (Fig 5). Negative correlations were also found with CE 319 (20.3) and TG (54.1), TG (52.1), but most notably with TG (50.1). Based on the number of carbons in 320 their acyl chains, it seems likely that these three TGs are closely connected in the TG biosynthetic 321 pathway. Negative correlations among metabolites are sometimes due to regulatory mechanisms 322 controlling metabolites that have conserved moieties, or metabolites connected by enzymes with high 323 variance which can cause negative correlations between its substrate and product metabolites [43,48].
324 Based on the simplified pathway in Figure 1B, a relationship between the CE and TGs can be envisaged.