top of page

TL-Jewellery 群組

公開·18 位會員

DECODE RED - Part 222:26 Min

Spatial partitioning of molecules into functional compartments is a basic principle for the organization of a cell (Bauer et al., 2015). Compartmentation determines the chemical environment, and thus profoundly affects kinase activity and substrate selectivity. Kinases were found in most organelles, and study of subcellular localization on the proteome level facilitated mapping of the kinome (Thul et al., 2017). In imaging-based approaches, antibodies were used to label proteins for visualization. For instance, Cell Atlas of the Human Protein Atlas (HPA) project mapped more than 12,000 proteins to 13 major organelles using antibodies against endogenous proteins (Thul et al., 2017). However, reliability was limited by antibody specificity in such one antibody one protein approaches (Baker, 2015). Alternatively, organelle proteome could be identified by mass spectrometry after fractionation (Calvo et al., 2016; Orre et al., 2019). Although empowered by proximity-labeling technologies (Branon et al., 2018; Lönn and Landegren, 2017), compartments accurately analyzed this way are still limited. Furthermore, curation of subcellular localization data from UniProt, literatures, and open resources has resulted in comprehensive databases such as COMPARTMENTS (Binder et al., 2014). However, accuracy is often affected by complex data source. As a result, understanding of subcellular localizations of the kinome is still fragmented.

DECODE RED - Part 222:26 Min

(A) Distribution of the kinome plasmid library on the kinome tree. Each red dot represents a kinase in the library. (B) Ratio of N-terminal- and C-terminal-tagged kinases in each kinase family. (C) Flowchart of kinome subcellular localization mapping. (D) Distribution of 10 subcellular compartments in each kinase family (left panel) and in kinases with different tag positions (right panel).

KA revealed kinases with previously unknown localizations to major or sub-compartments. Identification of SGK3 as an exosomal kinase suggests a role in intercellular communication. KA also found MAP4K3 in tubule structures overlapping with ER and microtubules, which is different from previous reports (Hsu et al., 2018). While the nature of this structure is not entirely clear, it is specifically regulated by LKB1, an upstream component of the mTOR pathway. It is possible that LKB1 regulates the structure itself or the localization of MAP4K3 to this structure. Intriguingly, MAP4K3 was reported to mediate the regulation of mTOR activity by amino acid sufficiency. Consistently, studies in Drosophila and mice indicated a role of MAP4K3 in regulation of body size (Bryk et al., 2010), immune response (Chuang et al., 2011), and longevity (Chuang et al., 2019), although the mechanism of action is not clear. Further elucidation of this LKB1-regulated localization of MAP4K3 may help solve the mystery. KA has also identified kinases enriched at nuclear peripheral regions, which may play a role in the maintenance and functions of this domain featured by gene-poor and compact chromatin with gene silencing markers (Lieberman-Aiden et al., 2009; Stevens et al., 2017). KA has also documented a collection of new kinases on plasma membrane, in ER, and in nucleolus, which all worth further studies. Except for cytosolic and nuclear localizations, prediction of these new localizations were largely ineffective, suggesting novel sorting mechanisms such as signal patches.

Here, the authors have carried out a systematic survey of the subcellular localizations of a large fraction of the protein kinases in the human kinome, which they have catalogued in a Kinase Atlas (KA). To do this they tagged 464 out of the 538 human protein kinases with Flag, HA, HA-Flag, Myc or GFP at either the N- or C-terminus, and transiently expressed them by transfection in HeLa cells. 95% of the tagged kinases could be detected, with about half being expressed at high level. Staining for the pertinent tags and comparison with cells stained for organellar marker proteins allowed the authors to assign each protein kinase to one of 10 subcellular compartments: cytosol (C), nucleus (N), plasma membrane (PM), mitochondrion (MI), endoplasmic reticulum (ER), Golgi apparatus (GL), vesicle (V), cytoskeleton (CS), centrosome (CT), and aggresome (AG). Half the kinases were localized to the cytosol, 15% to the nucleus, and 10% to the plasma membrane, and the rest were divided among other compartments, with several kinases having more than one location. Families of kinases tended to have preferred locations; for instance, as expected the tyrosine kinase family was enriched on the plasma membrane, largely due to the large number of RTKs. Most of the assigned mitochondrial kinases were atypical protein kinases, whereas the RGC and PKL subfamilies were ER/Golgi localized. They found about 70% overlap between their KA localizations and other databases, such as COMPARTMENTS and the Human Proteome Atlas. Some of the discrepancies may depend on whether the tags were N- or C-terminal and the nature of the tag. They found 104 kinase localizations that were uniquely annotated by KA, with 7 localized to the plasma membrane, as confirmed by cell fractionation studies. POMK, the protein-O-mannose kinase, was confirmed to be ER localized, as expected. SGK3 was annotated as specific for the Golgi compartment, but could also be detected in conditioned medium. MAP4K3/MEKK3 was found to decorate tubular structures that were shown to overlap with the ER. Several nuclear kinases were unevenly distributed in puncta of varying size, Including TRIB3 and HIPK2, whereas BMP2K formed puncta in the cytoplasm. Treatment with hexanediol disrupted puncta formed by several of these protein kinases but not puncta formed by other protein kinases, suggesting that the hexanediol-sensitive puncta might form as a result of liquid-liquid phase separation (LLPS). Deletion of predicted IDRs in TRIB3 and BMP2K eliminated their puncta formation. To determine if endogenous BMP2K formed LLPS puncta, they generated BMP2K-HA tag knock-in HeLa cells, and observed BMP2K present in cytoplasmic and/or nuclear puncta, which were abolished by hexanediol treatment. They also found four previously unidentified mitochondrial kinases not annotated in MitoCarta2: MOK, LIMK2, PKN3 and the TNK1 tyrosine kinase. They confirmed the mitochondrial localization of MOK by generating a MOK-Flag knock-in HeLa cell line, and showing that both a full-length 50 kDa form and shorter 20 kDa form were enriched in a mitochondrial fraction. Fractionation and IF analysis indicated that MOK and TNK1 were localized to the mitochondrial intermembrane space, whereas LIMK2 was on the outer mitochondrial membrane. They mapped two short regions in MOK (aa53-103 and aa301-330), required for mitochondrial import. To explore a possible mitochondrial function of MOK, the authors generated MOK knockout A375 melanoma cells and stable MOK knockdown Caki-1 cells, and found that MOK KO/KD cell mitochondria had fewer cristae based on EM analysis, and that both respiration rate and ATP level were reduced, and could be rescued by expression of the MOK2 splice variant but less well by full length MOK1. They showed that the MOK2 splice variant, missing 30 residues near the N terminus of the catalytic domain, lacked in vitro kinase activity, and yet was able to rescue the mitochondrial phenotype of the MOK KO A375 cells. Finally, in investigating possible mechanisms of MOK mitochondrial import, they showed that depletion HSPA9, which is responsible for inward translocation of matrix proteins downstream of the TIM23 complex, reduced MOK import, and that MOK was associated with the IMM protein ATAD3A, which spans between the IMM and OMM, and the SLC25A13 and SLC25A11 solute carriers.

1. It is unclear to what extent the observed protein kinase localizations require intrinsic protein kinase activity (apart from MOK). This could be checked by expressing kinase-dead mutants of a subset of the protein kinases with interesting localizations. Such studies may be beyond the scope of this paper, but this issue should be discussed.

4. Based on Figure 1D, it appears that the authors did not co-stain for a protein kinase of interest and a compartment specific marker to localize the protein kinase, but rather relied on patterns of kinase localization similar to that of an organelle marker protein. If this was what was done, it needs to be explained.

In this manuscript, Zhang et al. describe the Kinome Atlas (KA), a map of the subcellular localization of 456 kinases representing 85% of the human kinome. The KA is based on experimentally acquired imaging data; the authors have assessed the localization of these kinases using immunofluorescence of overexpressed tagged cDNAs. Importantly, the authors define 10 subcellular compartments that are used to catalog each kinase based on their localization.

Using KA, the authors define the subcellular localization of a collection of kinases that had not been previously studied. In addition, they further report a particular subcellular pattern in some kinases, which the authors attribute to molecular condensates.

It would be interesting to decode consensus linear motifs from newly identified kinases with specific localizations. In order to do so, we have searched for known motifs using Motif Scan ( -bin/motif_scan/), MOTIFS (, and searched for new linear motifs using MEME Suite ( To exclude the interference of the kinase domain, which is highly homologous, we have also performed analysis using sequences excluding the kinase domain. However, no new motifs related to mitochondrial or plasma membrane localizations were found. Other localizations were not analyzed. We postulate that unconventional localization signals may be diverse and present at low ratios. There may be a higher chance to identify such motifs when analysis was performed on the proteome level beyond the kinome. In addition, as we discussed in page 17, line 545, novel sorting mechanisms such as signal patches are possible, which could not be predicted from secondary sequence. 041b061a72


bottom of page