3.3. FA and TAG Biosynthesis Profiling in Oil Palm Mesocarp Development (12 WAA–22 WAA) Using the Oil Palm Mesocarp Microarray
To study the robustness of the oil palm mesocarp microarray, we further compared the expression profile of the selected FA and TAG biosynthesis genes throughout mesocarp development with published datasets derived from transcriptome sequencing at similar development stages. From the expression profiles of the selected genes (Figure 3), the majority of probes exhibited an increase in expression signal throughout mesocarp development. This is in concordance with Bourgis et al. [4] where they reported almost all the transcripts related to FA biosynthesis continue to increase until the end of oil accumulation. From our correlation study, the FA and TAGs genes show reasonable correlation (Pearson correlation) with R2 = 0.569 and p-value < 0.01 (Figure 4). Individually, 56% of these genes assessed with the custom oil palm microarray show high concordance to the published dataset, with R2 > 0.9 and p-value < 0.05 (Table 2).
Nonetheless, there are a few genes that exhibited differences in expression profiles derived using this microarray compared to the published data [4,6] such as LACS, SAD, MAT, PDH(DHLAT), CPT, GPAT, PDAT, DGAT 1, WRI1 and LPAAT (Figure 3). The differences could be a reflection of how the respective experiments were carried out and also differences in the technology used to assess the transcripts (hybridisation versus library sequencing). However, when we compared the WRI1 expression profile from the oil palm microarray here to Bourgis et al. [4] and to Tranbarger et al. [6] we observed similarity between these three different studies (Figure 5). We observed that the expression pattern of WRI1 in Bourgis et al. [4] (Figure 5A) and our microarray (Figure 5C) are similar at initial stages (12 WAA to 16 WAA) but show different trends of expression at the end of the maturation stage. However, the expression trend of WRI1 in our microarray data is similar to that reported by Tranbarger et al. [6] (Figure 5B) with an expression peak at 120 DAP (~17 WAA). The expression of WRI1 then declines as the mesocarp develops towards the maturation stage.
microarrays-03-00263-t002_Table 2 Table 2  Pearson correlation of expression changes between microarray and published data [4]. * Significant at p-value < 0.05.
Figure 3  Expression change comparisons of selected FA genes between microarray and Bourgis et al. [4] throughout mesocarp development.
Figure 4  Coefficient of correlation of transcriptome sequencing [4] and microarray data for FA genes at 16 WAA. R2 = 0.569, p-value < 0.01.
Figure 5  Expression comparisons of WRI1 between Bourgis et al. [4] (A), Tranbarger et al. [6] (B) and oil palm microarray (C).   Differences in sampling methods at different maturation stages and potentially differences in the genetic backgrounds used maycontribute to the observed differences in expression profiles of these genes. In this study, mesocarp samples of each maturation stage were harvested from all 16 selected palms, compared to the published studies where mesocarp samples of different maturation stages were harvested from different palms. We also found that the expression of specific genes can be genotype dependent (data not shown).
In the design of this microarray, we also included paralog sequences from our mesocarp transcriptome sequencing that appeared in the mesocarp tissue, focusing on the genes involved in the FA biosynthesis pathway as reported in Dussert et al. [5]. From the comparison of paralogs of several FA genes (ACC, KAS I, KAS III and SAD), we observed that the level of expression of these paralogs is different and that these correspond to different enzymes (Figure 6). The paralogs corresponding to KAS I are different in expression level at the initial stages (12–14 WAA) but similar in their expression level from 16 WAA until maturation. In the case of KAS III, the paralog 18883 was expressed at higher levels than 26680 at the initial stages of mesocarp development (12–14 WAA) but the relative expression levels switched for these two paralogs at later stages of oil palm mesocarp development (18–22 WAA). However, in some cases paralogs showed only low levels of expression throughout mesocarp development as can be observed in the case of ACC and SAD. From the expression pattern of different paralogs (although within the same tissue, mesocarp) it can be suggested that these paralogs may play different roles in the FA biosynthesis pathway [5]. The ability to design microarray probes to detect paralogs, based on the extensive transcript sequencing potentially allows the detection of functionally different transcripts. Overall, it was observed that the expression levels of individual genes appeared more consistent using the oil palm microarray than reported using the 454 sequencing-based publications [4,6], despite the fact that 454 sequencing is known to be sensitive [40]. However, microarray experiments are based on the hybridization efficiencies of the probes and this will affect the level of signal produced. The probe efficiencies may be affected by sequence differences between transcript and probe sequence. Furthermore, the position of differences in probe sequences may also affect hybridization efficiencies and eventually signal intensity [41]. While being a disadvantage of microarray approaches compared to transcriptome sequencing, the use of siblings in this experiment combined with the 3 × 60 bp probe design should minimize the effect. The development of a custom oil palm mesocarp microarray is also justified by the increased range of response observed when genes involved in lipid biosynthesis were compared with the Arabidopsis microarray. While Arabidopsis is a highly characterized model plant with a full genome sequence that has been highly annotated, the current study suggests that it may have limited use as a cross-species microarray platform for oil palm without validation and cross-species sequence comparisons, even for metabolic pathways that are functionally conserved.
Figure 6  Expression of paralogs of four FA genes throughout mesocarp development.

3