Chapter 3. Results
3.6 Microarray Data Analysis
To elucidate the biological functions that rol genes manipulated in hairy root, Agilent Tobacco Oligo 4x44K microarray were used to analysis transcriptional profiles differences between hairy roots induced by A. rhizogenes strain A4 and its respective rol gene-deficient derivatives. We first compared transcriptome of hairy roots induced
by WTA4 and all respective rol gene-deficient derivatives. Considered the pivotal roles ΔrolB and ΔrolC had on hairy root architecture, we then performed another biological replicate by comparing the transcriptome of hairy roots induced by ΔrolB / ΔrolC and WTA4 using the same kind microarray chip in a dye-swap manner. Each microarray RNA sample was combined of at least 20 independent hairy root clones.
In comparison between ΔrolA and WTA4, 217 genes were differentially up regulated in HR ΔrolA under 2-fold cut off threshold and 352 genes were differentially down regulated (Table 3-1). On the other hand, comparison between ΔrolD and WTA4 showed that 271 genes were differentially up regulated in HR ΔrolD under 2-fold cut off threshold and 297 genes were differentially down regulated (Table 3-1).
The consistent up- and down- regulated genes from the comparison between HR ΔrolB vs. HR WT as well HR ΔrolC vs. HR WT in the two biological replicates were meshed under 2-fold cut off threshold. The number of up regulated genes in HR ΔrolB and HR ΔrolC compared with HR WT were 6 and 42; the number of down regulated genes in HR ΔrolB and HR ΔrolC compared with HR WT were 242 and 208, respectively (Fig. 3-12, and Table 3-1). The number of the constitutive expressed genes was lower in comparison with other microarray analyses. This may due to different hairy root clones have different T-DNA insertion sites and / or insertion copy number, and we used two different sets of hairy root to perform the biological replicates.
Therefore, we are more confident with the constitutive expressed genes as they may be more likely to be regulated by lacking of the rol genes, albeit their T-DNA insertion site and / or insertion copy numbers. The same results can be seen in the constitutive down regulated genes with minor effects. The down-regulated genes were used for gene ontology analysis as they may be positively regulated by rolB and rolC, respectively.
3.7 Gene ontology of HR ΔrolB down-regulated genes
Gene ontology of the 242 down regulated-genes in HR ΔrolB was performed using agriGO with default settings (Du et al., 2010). Half of the 242 genes have been annotated, therefore only these genes were compared. The GO terms with p-value smaller than 0.05 were shown in Table 3-2. To get better impression about the relationships among the GO extracted, diagram according to biological process was produced (Fig. 3-13).
Four main categories of GO were present: localization, cellular process, metabolic process, and response to stimulus. The subgroups in “localization” pin point that the lipid transport was drastically decreased in HR ΔrolB (p = 2.95x10-13). Genes included in this blocks are listed in Table 3-3. The major group “response to stimulus” contained two specific subgroups showing significant differences. These groups were “response to wounding” and “response to ethylene” (p = 1.36x10-4 and 4.43x10-2, respectively; Fig.
3-13). Although there was no significant difference in the group “cellular process and metabolic process”, one of their subgroup “cellular amino acid derivative biosynthesis process” was extracted (p = 3.35x10-2, Fig. 3-13).
3.8 Gene Ontology of HR ΔrolC down-regulated genes
Gene ontology of the down regulated genes in HR ΔrolC was performed as that
described in HR ΔrolB. In this case, 88 of 208 genes have been annotated. Therefore, we chose these 88 genes for further analysis. Table 3-4 provides the complete GO terms with p-value smaller than 0.05.
Three main categories of GO were present: response to stimulus, metabolic process, and localization (Fig. 3-14). The subgroups in “localization” showed that the lipid transport was drastically decreased in HR ΔrolC (p = 6.37x10-10) (Fig. 3-14 and Table 3-5). It is worth noting that the group “lipid transport” showed the smallest p-value among all GO terms in HR ΔrolC and well as in HR ΔrolB. The specific sub-group extracted in metabolic process was the carbohydrate metabolic process (p = 1.52x10-2) (Fig. 3-14 and Table 3-5). Two sub-groups that were significant extracted from the group “response to stimulus” were “response to chemical stimulus (p = 4.11x10-3) and response to wounding (p = 8.10x10-3) (Fig. 3-14). Genes of these groups were listed in Table 3-5.
3.9 qRT-PCR Analysis of the Gene Ontology-extracted genes
qRT-PCR was used to confirm the results obtained from the microarray. We aimed to test the relatively down-regulated genes in HR ΔrolB / HR ΔrolC comparing to HR WT. The comparison results between HR ΔrolB and HR WT confirmed the microarray data where all the successfully tested genes showed lower expression level in HR ΔrolB than HR WT (Table 3-3, Fold change in qRT-PCR analysis). The confident level of the qRT-PCR test was high as permutation analysis showed p-value equals to 1x10-3 in all except primer DW003388 (p-value = 2x10-3). The unsuccessful tested genes mean the genes that showed CT value larger than 0.5 among the 3 technical replicates. These data were excluded and represented as NA in Table 3-3. Also, primers exclusively existed in GO term “cellular amino acid derivative biosynthesis process” had not been designated,
and was shown by gray background.
Comparing the expression level between HR ΔrolC and HR WT, 26 of the genes tested showed the same trend as the conclusion made in microarray analysis with small p-value (Table 3-5). The gray backgrounds in Table 3-5 represent the probes without
known NCBI accessions, and no qRT-PCR data shown. Primers showed CT value larger than 0.5 among the 3 technical replicates were excluded and thus no fold change were shown (represented by NA). Although the expression of EB684101 (belongs to carbohydrate metabolic process) and FG636940 (belongs to chemical stimulus) were 0.9363 and 0.8980 fold, respectively, in HR ΔrolC divided by HR WT, both permutation p-values were larger than 0.05 (p-value was 0.0679 for the former and 0.1568 for the latter). Therefore, these data should be abolished then test again. On the other hand, gene FG137954 (belongs to lipid transport) and AB263747 (belongs to carbohydrate metabolic process) were not differentially expressed in HR WT and showed small p-value. The expression of FG137954 and AB263747 did not obey the trend shown in microarray, and these genes should be excluded for further hypothesis generation. For rolC-regulated-genes exploration, the 26 tested genes that showed the expected trends should be analyzed with higher priority.
Overall, the qRT-PCR analysis reconfirmed the accuracy of the microarray data and helped to exclude some unlikely-to-be-true cases. The information can be used in further biological function exploration of the rolB and rolC.
Chapter 4. Discussions
4.1 rolB and rolC play crucial roles in hairy root initiationAs revealed in DREPI and RL ratio analysis (Fig. 3-5 and 3-6), our results showed that the ΔrolB and ΔrolC were less virulent to N. tabacum than WT A4 was. These observations are similar to the results obtained in K. diagremontiana where rolB Tn5- interrupted mutant results in avirulent traits and rolC interrupted mutant results in attenuate hairy root growth after initiation (White et al., 1985). However, there are some differences. For example, the rolB interrupted mutant resulted in avirulent traits in K.
diagremontiana, but rolB deleted mutant only resulted in significantly decreased
virulence in N. tabacum. Moreover, we did not observe the hyper-hairy root syndrome in ΔrolA in N. tabacum, which was the case in K. diagremontiana (White et al., 1985).
In fact, no significant difference was shown in the RL ratio analysis between HR ΔrolA and HR WT (p = 1.9x10-1). The different hairy root initiation ability between K.
diagremontiana and N. tabacum may due to their different plant origin. This point of
view can be further supported by the fact that although rolD deletion resulted in retard hairy root growth and increase callus in K. diagremontiana, our results showed that rolD was not necessary for hairy root inducing in tobacco system, which consists with
what Vilaine and colleagues had showed (Vilaine and Cassedelbart, 1987). However, the trend that rolB and rolC seem to play a major role in hairy root initiation did not change.
4.2 rolB and rolC affect hairy root architecture more drastically
To our knowledge, there is no comprehensive analysis about how rol genes affect hairy root architecture. In this study, we analyzed the main root length, branch root
number, total branch root length, and branch root density of the rol gene-deficient A.
rhizogenes A4 induced-hairy root. From the analyses, HR ΔrolC showed impairing in
all the parameters measured with the smallest p-value (p = 9.99x10-4) when comparing to HR WT (Fig. 3-9 to 3-12). This indicated that the development of HR ΔrolC were greatly impaired. HR ΔrolB showed reduced branch root number and total branch root length, but not in main root length and branch root density (Fig. 3-9 to 3-12). The p-value of the branch root density between HR ΔrolB and HR WT were 0.0789, which
was slightly higher than the significant level α=0.05. The results indicated that branch root development of HR ΔrolB were impaired. No significant difference was observed between HR ΔrolA and HR WT in hairy root architecture, nor in hairy root initiation.
However, the absence of rolA did show hyper-hairy root morphology in K.
diagremontiana system (White et al., 1985). Therefor, rolA probably plays a milder role
in hairy root syndrome. Similar to HR ΔrolA, only main root length was affected in HR ΔrolD. The decreased MRL in HR ΔrolD reflects the observation that rolD interrupted mutant resulted in retard hairy root growth in K. diagremontiana (White et al., 1985).
Taken together, it suggest that rolB and rolC that play more vital roles in hairy root syndrome.
4.3 Difficult maintenance of the HR ΔrolB and HR ΔrolC clones
We noticed that the available clones of HR ΔrolB and HR ΔrolC were decreasing during persistent sub-culturing. Many hairy root clones of both genotypes slowed or stopped elongation (but the root remained white and healthy) after multiple sub-culturing. The population decreased from 65 to less than 10 independent clones in HR ΔrolB, and from 82 to less than 15 in HR ΔrolC, respectively, over time. Moreover, it should be noticed that the initial population (65 and 82) had already excluded the
clones that showed slow growing rates at the beginning of hairy root emergence. This phenomenon was less pronounced in HR ΔrolA, HR ΔrolD, and HR WT. The ongoing population declining made us failed to have enough biomass from one independent clone for qRT-PCR analysis. Therefore, more hairy root induction should be performed if more HR ΔrolB and HR ΔrolC clones are needed for qRT-PCR analysis.
4.4 Comparison between the microarray and qRT-PCR analyses
We noticed that the expression fold change of HR ΔrolB vs. HR WT measured in qRT-PCR were consistently larger than that detected in microarray (Table 3-3 & 3-5), regardless of what the genes was tested. Similarly, the expression fold change of HR ΔrolC vs. HR WT of the qRT-PCR analysis seemed to lie within that detected in first and second microarray observed. This may resulted from two aspects: different sampling strategies and sampling size. The sampling strategy in microarray analysis was mixed sampling while individual clones were independently analyzed in qRT-PCR.
Also, the mixed sample contained at least 20 clones while 2 clones were used for qRT-PCR. Therefore, more independent clones were needed for qRT-PCR test to know the population down-regulation range. However, the observations that these genes were down regulated in HR ΔrolB vs. of HR WT and / or in HR ΔrolC vs. of HR WT were true, regardless of what the experiments was used.
4.5 Biological processes manipulated in the absence of rolB
Our GO results did not extract any auxin-related groups (Fig. 3-12), although 4 probes did show GO term “response to auxin stimulus”. The observations that neither endogenous free IAA, nor the rate of IAA biosynthesis and metabolism were altered in rolB-transformed tobacco (Nilsson et al., 1993; Schmülling et al., 1993) support our
finding. On the other hand, the group “response to ethylene stimulus” appears out of
expectation since no previous studies have linked rolB to ethylene (Fig. 3-12).
Literature searching shows some interesting clues of the relationship between RolB and ethylene. First, both of them enhance auxin bind to plasma membrane (Maurel et al., 1990; Friml et al., 2002; Negi et al., 2008; Lewis et al., 2011). Second, both of them seem to have a positive regulation of wounding. The effect of ethylene on wounding (and vice versa) have been well demonstrated (Hamilton et al., 1990; O'Donnell et al., 1996). The GO group “response to wounding” was extracted in HR ΔrolB-down regulated genes suggest a positive regulate character of rolB on wounding response.
Therefore, further work may focus on the ethylene content in hairy root and see if the ethylene content were lower in HR ΔrolB than that in HR WT.
Also extracted from GO analysis of the HR ΔrolB down-regulated genes was a Myb58 gene, which is a transcription activator of the lignin biosynthesis pathway (Zhou et al., 2009). Since lignin is one of the main components of cell wall, lacking Myb58 will results in cell wall biosynthesis defects. This may explain our observation that the branch root development were retarded in HR ΔrolB and that it was difficult to maintain HR ΔrolB clones because they stop to elongation (see Discussion 4.2). Future work may focus on whether the stop-to-elongation phenomenon observed in HR ΔrolB does result from cell wall biosynthesis defects.
Another gene extracted from GO analysis were ribonuclease 1 (RNS1) that has been known to regulate anthocyanin, a secondary metabolite, content in plant cell (Bariola et al., 1994). Although no reports directly suggest that anthocyanin content is higher in rolB expression hairy root, study does showed that the gene expression of phenylalanine ammonia-lyase (PAL) and stilbene synthase (STS), key enzymes of anthocyanin biosynthesis, was higher in rolB expressing grape (Vitis amurensis) (Kiselev et al., 2009). Therefore, two questions remain to be elucidated. The one is
whether anthocyanin content is higher in rolB-expression hairy root comparing to the unexpressed ones. The second is that whether rolB regulates secondary metabolite synthesis by RNS1.
Overall, the gene ontology analysis revealed several plausible ways of finding the biological functions of the rolB.
4.6 Biological processes manipulated in the absence of rolC
Gene ontology analysis revealed that lipid transport was affected in HR ΔrolC when compared to HR WT (p = 6.37x10-12, Fig. 3-14). In this group, 11 probes, which matched 6 GO sources, were extracted (Table 3-5). According to the gene name listed in the GO sources, we searched the TAIR database and found out that 4 of them, AT1G12090, AT1G62500, AT1G62510, and AT5G48485, locate in the endomembrane system. Another GO source, AT2G10940, locates in chloroplast thylakoid membrane, plasmodesmata, and apoplast, while GO source AT3G22142 does not have localization information. The results indicated that lipid transport in the endomembrane system was likely to be manipulated in HR ΔrolC. However, previous studies pointed out that RolC lies in the soluble, cytosolic fraction of the transgenic cells (Estruch et al., 1991; Oono et al., 1991). Therefore, RolC might either regulate lipid protein indirectly or some domain of these proteins may lie within the cytosol (like membrane integrated proteins).
Future works may focus on is there possible that lipid transport proteins and RolC have direct interaction.
Ten tobacco probes known to belong to GO group carbohydrate metabolic process were extracted. The probes can be referred to 8 GO sources. Six of them had been evaluated by qRT-PCR and 4 of them had been confirmed this way. The reconfirmed ones were AT1G10640, AT1G60590, AT3G61490, and AT5G55180. According to
TAIR, gene AT5G55180 belongs to O-glycosyl hydrolase family 17, locates in endomembrane system, and function in cation binding. This gene was up regulated in dark-grown seedlings when treated 6 h with 90 mM sucrose (NCBI GEO profile GDS1734 / 248100_at / AT5G55180). Since rolC was specifically expressed in phloem and its expression can be induced by sucrose (Schmulling et al., 1989; Sugaya et al., 1989; Yokoyama et al., 1994), it would be interesting to know if the two genes are related.
GO group response to chemical stimulus contained 21 probes with 16 GO sources.
Genes belonged to this group had different functions and are stimulated by different sources (for example: pathogen, auxin, gibberellin). However, many of them seemed to response directly and / or indirectly to biotic stress like pathogen. Combining this finding and the other extracted GO group “response to wounding”, the ability of hairy root to response to pathogen might be diminished in HR ΔrolC.
4.7 Deduced biological functions of rolB
Our results showed that rolB deletion resulted in retard hairy root initiation (including DREPI and RL ratio) and reduced branch root number, suggesting that rolB may stimulate meristem formation. This viewpoint can be strengthen by the fact that rolB stimulates the formation of flower and root meristemoids and that its promoter
express in all types of meristem in transgenic tobacco plant (Altamura et al., 1991;
Altamura et al., 1994). The observation that total branch root length was also reduced in HR ΔrolB, indicated that rolB might also play a role in meristem maintenance.
Moreover, the slow or no growing phenomenon of HR ΔrolB after multiple sub-culturing strengthened this viewpoint. Therefore, further work may focus on whether root development- and / or meristem- related genes expression are inhibited in
HR ΔrolB.
Bearing the idea that rolB may play a role in meristem formation and / or maintenance and combining the GO analysis data, we come up with a hypothesis of how RolB acts in the plant cell (Appendix 1). In meristematic cells where high auxin level is high, auxin induces the expression of rolB promoter that results in increases RolB content. High content of RolB increases cell auxin sensitivity and resulted in meristem formation (gray background in Appendix 1). In order to prevent hyper-auxin level, RolB activates ethylene biosynthesis pathway by unknown mechanisms. This may be achieved by activating
1-aminocyclopropane-1-carboxylate synthase (ACC
synthase) since this gene was extracted in GO analysis. Increase in ethylene resulted in expression of the down stream pathogen related genes (PR genes) and the wound response genes (lower right in Appendix 1). Some lipid transport proteins are known to be induced by wounding (García-Olmedo et al., 1995; Jung et al., 2003). Therefore, these genes may be regulated by RolB as the wounding response genes do. However, more experiments have to be done to check this hypothesis.4.8 Deduced biological functions of rolC
RolC may play an important role in meristem formation and / or maintenance
because HR ΔrolC was defective in main root length, branch root number, total branch root length, and branch root density consists well. Moreover, the result observed in K.
kiagremontiana where rolC interrupted mutant showed attenuated hairy root growth
after initiation (White et al., 1985) strengthen this viewpoint. GO analysis revealed that carbohydrate metabolic process genes includes a group of cell wall biosynthesis-related genes: polygalacturonase, glycoside hydrolase, xyloglucan:xyloglucosyl transferase, and acidic endochitinase (Fig. 3-13). Therefore, aside from meristem activity, the retard
morphology observed in HR ΔrolC may results from defect cell wall biosynthesis defects (Appendix 2). Another reason that may cause the defect morphology of HR ΔrolC may be referred to the pyruvate decarboxylase, lactate dehydrogenase, and alcohol dehydrogenase 1, which were extracted in the GO group “response to chemical stimulus”. Lack of these genes will cause malfunction in carbohydrate metabolism. It is still unknown how RolC interacts with these genes (Appendix 2). Therefore, future study may focus on these genes to elucidate how they affect root elongation.
The GO results showed that HR ΔrolC may be less response to wounding and
The GO results showed that HR ΔrolC may be less response to wounding and