According to the previous analysis, out of 12,342 detected transcripts, 3,218 transcripts (26%) varied significantly in their translation state (TL) over a diurnal cycle (Supplementary Dataset 6). In this dataset, 39 (out of a total of 49) transcripts encoding Fe-S assembly proteins and 68 (out of 97) transcripts encoding Fe-S cluster-utilizing proteins were detected as expressed (Supplementary Datasets 3, 4). Among these, 13 transcripts encoding Fe-S assembly proteins (33%) and 13 transcripts encoding Fe-S cluster-utilizing proteins (19%) showed diurnal fluctuations in TL (Table 1). The apparent over-representation of Fe-S cluster assembly genes among genes with diurnal changes in TL was not statistically significant according to a Fisher’s exact test.

Thirteen transcripts encoding Fe-S assembly proteins varied significantly in their TL over the diurnal cycle. Among these 13 transcripts, 7 correspond to mitochondrial, 5 to plastidic and one to cytosolic proteins. These transcripts clustered into two groups of similar TL profiles over the diurnal cycle (Figure 3A). TL of transcripts in the first cluster peaked during and at the end of the night (ZT18 or ZT0). Interestingly, 7 out of 9 transcripts in this group encode proteins associated with Fe-S cluster carriage/transfer, including mitochondrial ISCA2, ISCA4, NFU4, and GRXS15 and chloroplast SUFA, NFU1 and GRXS14. The remaining two transcripts in this cluster encode the candidate Fe donor protein mitochondrial frataxin (FH) and the alpha-helical mitochondrial ferredoxin, respectively. This ferredoxin protein is thought to provide an electron for Fe-S cluster synthesis on ISU1, the major scaffold protein in mitochondria. TL of transcripts in the second cluster peaked during the day (ZT6 or ZT12). Among the four transcripts that reached peak TL during the day, two encode plastidic cysteine desulfurase activators and interactors of CpNifS, namely SUFE2 and SUFE3 (Narayana Murthy et al., 2007). The other two transcripts encode the cytosolic Fe-S assembly scaffold protein NBP35 (Bych et al., 2008; Kohbushi et al., 2009; Bastow et al., 2017), and HSCA2, one of two abundant mitochondrial HSP70-type chaperones which was proposed to function in Fe-S cluster transfer (Balk and Pilon, 2011). The estimation of protein biosynthesis rates (PBR) (Supplementary Figure 1A and Supplementary Dataset 7) resulted in diel profiles that were similar to TL for most Fe-S assembly proteins, but considerably more diverse than diel TL profiles for Fe-S carrier/transfer proteins (see Figure 3A).

Twenty-three transcripts encoding Fe-S proteins varied significantly in TL over the diurnal time course. Among these, 13 transcripts encode Fe-S cluster-utilizing proteins (Figure 3B) and 10 transcripts encode Fe-S proteins that are involved in cluster assembly and Fe-S protein biogenesis (Figure 3A). The 13 transcripts encoding Fe-S cluster-utilizing proteins comprised 6 plastidic, 4 nuclear, 2 cytosolic, and 1 mitochondrial localized proteins. Proteins detected to exhibit diurnal regulation of TL have various functions, in particular DNA and RNA metabolism as well as isoprenoid and leucine biosynthesis. All except two of these proteins are predicted to bind the most abundant [4Fe-4S] cluster, which is less stable than the [2Fe-2S] cluster. Only DWARF27B, a homolog of a rice β-carotene isomerase protein acting in strigolactone biosynthesis, and Xanthine the dehydrogenase XDH1 bind [2Fe-2S] clusters (Zhang et al., 2012; Balk and Schaedler, 2014). Fe-S cluster-utilizing proteins fell into three groups of differing diel TL profiles (Figure 3B). TL of 8 transcripts peaked during the day (ZT6 or ZT12), TL of two transcripts peaked at ZT0 and ZT6, and TL of three transcripts peaked at the end of the night (ZT0) (Figure 3B). TL of none of the Fe-S cluster-utilizing proteins peaked at ZT18 during the night, in contrast to the Fe-S carrier/transfer proteins that exhibited diel TL differences (see Figure 3A). When, instead of TL, we considered estimates of PBR, we mostly observed no changes or only minor shifts for few proteins (Supplementary Figure 1 and Supplementary Dataset 8).

Among the transcripts showing diel changes in TL, ten transcripts encode Fe-S proteins that are also involved in Fe-S protein biogenesis (see Figure 3A). The TL of 7 of these 10 transcripts peaked during the night (ZT18) and one peaked at the end of the night (ISCA4). Only the cytosolic Fe-S cluster scaffold protein NBP35 and SUFE3 peaked during the daytime.

To compare Fe-S cluster-containing proteins with other metalloproteins, we determined the proportion showing diurnal variation in TL among proteins binding heme/cytochrome cofactors (29%) and proteins binding cationic Fe (28%), Zn (22%), Cu (30%), or Mn (39%) (Table 2). Among these, in comparison to all functions (26%) only Zn-dependent proteins are significantly differently represented, with an underrepresentation in contrast to the trend for over-representation of heme/cytochrome-dependent, Cu-dependent and Mn-dependent proteins. In order to compare the timing of biogenesis of metalloproteins among those exhibiting significant variation in TL over the diurnal cycle, we determined the proportion of transcripts peaking in TL at each time point (Figure 4). Whereas TL of 62% of transcripts encoding Fe-S cluster assembly proteins peaked at ZT18, TL of none of the transcripts encoding Fe-S cluster-utilizing proteins peaked at this time point (Figure 4A; statistically significantly different from all detected transcripts at P < 0.01, Fisher’s exact test), in accordance with observations described earlier (see Figure 3). The diel TL profile of transcripts encoding heme/cytochrome- as well as Fe, Zn, and Cu cation-binding proteins was similar to that of all detected transcripts, with a maximum of 30–50% peaking at ZT18, a less pronounced secondary maximum of 25–42% peaking at ZT6 and a minimum of 0–17% peaking at ZT12 (Figures 4A,B). Notably, none of the transcripts encoding Cu-dependent proteins showed an afternoon peak in TL at ZT12 (statistically significantly different from all detected transcripts, P < 0.05, Fisher’s exact text), and as many as 50% peaked during the night at ZT18. The diel course of TL of transcripts encoding Mn-dependent proteins differed from that of all others, with a maximum of 39% of proteins peaking at the end of the night (ZT0).

In our analysis, we separated Fe-S cluster-associated functions into Fe-S cluster assembly and Fe-S cluster utilization. By comparison to all genes detected as expressed, those with functions in Fe-S cluster assembly were slightly, ca. 1.3-fold overrepresented among the genes exhibiting diurnal variation in TL. By contrast, genes associated with Fe-S cluster utilization were slightly, ca. 0.7-fold, underrepresented (Table 1). Expanding this analysis, genes encoding Mn-dependent proteins were the most overrepresented, ca. 1.5-fold, among genes exhibiting diurnal variation in TL, followed by genes encoding Cu-dependent proteins (1.2-fold) (Table 2). The proportion of genes exhibiting diurnal variation in TL among heme/cytochrome-dependent and Fe cation-dependent proteins was similar to that of all expressed genes. It is important to note that none of these deviations were statistically significant, possibly because of a lack of sensitivity of statistical testing given the small number of genes in each of the functional classes. The only significant difference was an underrepresentation, approximately 0.85-fold, of genes encoding Zn-dependent proteins among those exhibiting significant diurnal variation in TL. The considerably larger number of genes in this class is likely to have facilitated the statistical detection of this effect. This result may reflect the high proportion of transcription factors of overall low abundance among Zn-dependent proteins, by comparison to genes encoding highly abundant proteins functioning in metabolism. Taking these results together, diel dynamics in TL are relevant for all classes of metalloproteins.

However, based on this preliminary analysis, the pervasiveness of such diel dynamics in TL among metalloproteins is similar or only slightly higher than for non-metalloproteins. Moreover, our results suggest differences between different types of metalloproteins and metalloprotein functions, with overall more pronounced diel dynamics in TL among Mn-dependent proteins and Fe-S cluster assembly proteins and overall less diel dynamics in TL among Zn-dependent proteins. Evidently, TL analyzed in this study cannot be seen as a proxy of the overall rate of protein biogenesis (PBR), which is better described by an estimate of the total amount of ribosomal occupancy on transcripts derived from a given gene in the future (e.g., in this study: PBR_est = 2 × SP + 7 × LP). A re-analysis of data shown in Figure 3 suggested little difference between the diel dynamics in TL and the diel dynamics in PBR (Supplementary Figure 1) except for Fe-S carrier/transfer proteins, for which diel PBR profiles were rather variable.

Among the metalloprotein-encoding genes exhibiting diurnal variation in TL, we observed some striking differences in the diel profiles (Figure 4). The deviating profiles for Mn and, very pronouncedly, Cu-dependent metalloproteins are consistent with an overall metal-specific differentiation of TL over the diel cycle for these metals. In addition, there was a clear difference between genes encoding Fe-S cluster assembly proteins and those encoding Fe-S cluster-utilizing proteins. More than 60% of the former peaked during the night at ZT18, whereas none of the latter peaked at this time point. Instead, TL of genes encoding Fe-S cluster-utilizing proteins peaked mostly ZT6 (42%) and ZT12 (33%). A possible interpretation of this observation is the occurrence of a temporal gap between the biogenesis of Fe-S clusters and their incorporation into freshly synthesized polypeptide chains of Fe-S cluster-utilizing apo-metalloproteins.

A more detailed examination of the diel TL profiles of Fe-S cluster assembly proteins identified two distinct clusters (Figure 3A). The Fe-S assembly proteins in the first cluster, for which TL peaked during the night (ZT18 and to a lesser degree ZT0) mainly participate in the third step of Fe-S cluster protein biogenesis, which is Fe-S cluster transfer to target apoproteins via carrier/transfer proteins (Figure 1). By contrast, the three Fe-S cluster assembly proteins SUFE2, SUFE3, and NBP35 in the second cluster, which peaked during the day (ZT12 and ZT6), contribute to the first two steps of Fe-S protein biogenesis, persulfide formation and Fe-S cluster formation on the scaffold protein complex (Figure 1). These data suggest that some functions in Fe-S cluster synthesis and Fe-S cluster transfer to apoproteins have opposing diel dynamics of association with ribosomes. Taking all our observations together, the biosynthesis of apometalloproteins that require Fe-S clusters appears to occur preferentially during the day, as well as just before dawn for some cytosolic/nuclear Fe-S proteins (Figure 3 and Supplementary Figure 1).

In contrast to cytoplasmic NBP35, the main scaffold proteins SUFB, C and D of plastids and ISU1 of mitochondria did not show diurnal dynamics at the level of translation state. In the context of the model described above, it is possible that their basal transcript levels and translation states are sufficiently high throughout the day, which can satisfy the need of Fe-S biogenesis even during the peak phase of Fe-S protein synthesis. It is worth noting that besides stimulating the cysteine desulfurase activity of NFS2 alike SUFE2 (SufE activity), SUFE3 possesses quinolinate synthase activity which is required for NAD cofactor synthesis (Narayana Murthy et al., 2007). The quinolinate synthase activity of SUFE3 is dependent on the presence of a highly oxygen-sensitive [4Fe-4S] cluster (Cicchillo et al., 2005; Ollagnier-de Choudens et al., 2005; Narayana Murthy et al., 2007). SUFE3 cannot complement the embryo lethal phenotype of a sufe1 knock-out mutant, suggesting the SufE activity of SUFE3 is likely dedicated to Fe-S cluster formation in its own quinolinate synthase domain (Narayana Murthy et al., 2007; Balk and Pilon, 2011).

Three mitochondrial proteins are less easily integrated into the general patterns observed here, frataxin (FH), ferredoxin and HSCA2. FH and ferredoxin are thought to have a role in the second step of Fe-S cluster assembly by analogy to human and yeast (Balk and Schaedler, 2014), but their TLs peak at night alongside the TL of several proteins with Fe-S carrier/transfer functions (see Figure 3A, in agreement with Supplementary Figure 1). FH was proposed to deliver Fe to ISU1 by direct interaction with the NFS1/ISD11/ISU1 complex in mitochondria (Balk and Schaedler, 2014). An alternative model has also been discussed in which FH promotes the interaction of the cysteine desulfurase NFS1 with the scaffold ISU proteins to favor the sulfur transfer reaction (Couturier et al., 2013). FH was reported to localize to both mitochondria and plastids (Turowski et al., 2015). HSCA1 can functionally complement the yeast ssq1 knockout mutant and its ATPase activity is enhanced by HSCB and ISU1 (Xu et al., 2009). The role of HSCA2 in Fe-S protein biogenesis remains to be investigated. Additional and more detailed information will be required to test the possible models arising from this study. In particular also, diel variation in TL was only detected for a subset of proteins in each functional class.

Among the 13 Fe-S assembly proteins cycling in translation state, 7 are Fe-S carrier/transfer proteins. ISCA4, NFU4, GRXS15, and ISCA2 localize in mitochondria, whereas NFU1, SUFA, and GRXS14 are localized in plastids. So far, a number of Fe-S transfer/carrier proteins have been identified in A. thaliana and they are thought to form a flexible network for Fe-S cluster delivery (Couturier et al., 2013). The function of some Fe-S transfer/carrier proteins may be redundant, because knock-out mutants for most genes encoding Fe-S transfer/carrier proteins are viable (Couturier et al., 2013). For instance, plants deficient in GRXS14 or GRXS16 did not display any growth defect, whereas the growth rate of plants lacking both was reduced (Rey et al., 2017). However, some reports showed that Fe-S carrier/transfer proteins also have specificity for a particular Fe-S target protein or group of targets (Balk and Schaedler, 2014). For instance, NFU2 and HCF101 are involved in the maturation of one or several proteins belonging to PSI and some other stromal proteins (Lezhneva et al., 2004; Stöckel and Oelmüller, 2004; Touraine et al., 2004; Yabe et al., 2004). Thus, we presently cannot decide whether these 7 translationally cycling Fe-S carrier/transfer proteins may preferentially function in the biogenesis of specific Fe-S proteins.

Another feature of Fe-S carrier/transfer proteins is that they act in different manners depending on their redox environment. Mutants for some genes encoding Fe-S carrier/transfer proteins only showed lethal phenotypes specifically under oxygenic conditions. The maturation of two Fe-S proteins in E. coli, IspG and IspH, requires a different combination of Fe-S carrier/transfer protein isoforms under different oxygenic conditions. Under aerobic conditions, it uses IscU, IscA and then ErpA for cluster insertion into apo-IspG and apo-IspH, whereas it uses IscU and either ErpA or IscA under anaerobic conditions (Vinella et al., 2009). Another example is cytoplasmic NAR1 in eukaryotes, an Fe-S carrier/transfer protein in CIA pathway. NAR1 forms a complex with CIA1, AE7, and MET18 to transfer Fe-S clusters from the scaffold protein NBP35 to target apoproteins. Interestingly, nar1 mutants in plants, yeast, and nematodes display phenotypes only under normal oxygen pressure but not under low oxygen pressure (Fujii et al., 2009; Mondy et al., 2014). Therefore, the function of some Fe-S carrier/transfer proteins has a close relationship with the oxygen levels in vivo and in vitro, which might explain the existence of extensive cycling of TL of Fe-S transfer/carrier proteins over the diurnal cycle as observed here.

Here we report that the TL of subsets of Fe-S assembly and Fe-S utilizing proteins, as well as of other metalloproteins, vary over a diurnal cycle. It was concluded earlier that Fe-S assembly proteins are not diurnally regulated at the transcript level (Balk and Schaedler, 2014). But there is no investigation so far into whether the rates of Fe-S protein biosynthesis are under circadian or diurnal regulation. Knowledge on the regulation of Fe-S protein biogenesis in plants grown under various regimes of temperature and photoperiod would also help us better understand the mechanisms of de novo synthesis or repair of Fe-S clusters (Couturier et al., 2013).

Although much progress was made in elucidating the Fe-S protein biogenesis machinery in plants in the past decades, we are still lacking plant-specific knowledge on the function of a number of predicted Fe-S cluster assembly proteins. Another issue is that the environment-dependent regulation of Fe-S cluster protein biogenesis is largely unknown. Initially, it is important to understand whether and how Fe-S cluster assembly proteins and Fe-S cluster-utilizing proteins are regulated at the transcript, translational and protein levels under various environmental conditions, such as iron and sulfur starvation and other environmental stresses known to enhance internal levels of ROS.