Since it is important that the non-canonical motifs are also plausible in terms of TF-DNA binding affinities, we wanted to avoid those binding sites that show weaker affinity than the currently validated motifs.
As such, we chose to avoid rarer sequences. This is a conservative approach that likely will miss some non-canonical motifs. But we can be confident that the non-canonical motifs we report are bound and likely functional. Of note, since the HT-SELEX experiments were performed on individual TFs, by design the experiments preclude the possibility of co-factor binding affecting the observed specificity signals.
Nevertheless, it is useful to distinguish the current study from Slattery et al. We note that, Slattery et al. Rather, their study established that the Hox proteins recognize variants of the canonical motif by utilizing co-binding with Exd Extradenticle-Homothorax to bind at different genomic loci.
As such, Slattery et al. Also, Slattery et al. We found that 13 of the 19 TFs discussed above are homeodomains. Forkhead TFs have been discussed by Bulyk and colleagues [ 20 , 28 ] for their ability to recognize multiple motifs. We found only two of nine Forkhead TFs in this dataset have a non-canonical motif, and those too belong to the cases where the non-canonical motif appears to be similar across several different families Supplementary Text 3 , Table 1 , Supplementary Table 1.
It is also worth mentioning that we did not find a non-canonical motif for any of the 14 nuclear receptor factor TFs, suggesting that some TF families may have a characteristic lack of non-canonical motifs. How can these non-canonical motifs be important if they are generally less abundant than canonical motifs?
Secondly, in a tissue and cell-type specific manner, the occurrences of some of these non-canonical motifs show as strong an evolutionary conservation as the corresponding CIS-BP motifs. Non-canonical motifs are important also from the perspective of biochemical mechanisms. Such differences in recognition mechanisms may play a role in tissue and cell-type specificity of a TF. As such, it is important to consider the entire set of possible motifs for a TF.
We analyzed the quality filtered datasets for human TFs released with Yang et al. For detecting motif matches, we used the FIMO [ 10 ] program with the commonly adopted and relatively liberal significance threshold of 1e-4 [ 7 , 12 , 15 , 16 , 21 , 31 , 34 ]. To investigate the existence of non-canonical motifs and their difference from canonical motifs, we developed a pipeline for analyzing HT-SELEX data.
The additional criteria aim to ensure that any observed signal of non-canonical motifs is likely not an artifact of the HT-SELEX procedure [ 23 ].
The following is an outline of our pipeline; we describe the steps in detail in the following sub-sections. Flow diagram of the current analysis pipeline to identify non-canonical motifs. The pipeline combines the standard practices of HT-SELEX data modeling [ 29 ] with an additional set of conservative filtering criteria to eliminate experimental and statistical artifacts shown in red. These criteria include selecting L-mers sequences of length L with a minimum count and a minimum di-nucleotide entropy, followed by selecting L-mers that have a minimum enrichment, positive enrichment over successive rounds, and alone occurrence i.
Step 1. From the TF-bound oligos from the round 0 library, one then constructs the round 1 library and repeats this process for several rounds. Step 2. Within the oligos of the selected round, we identify the occurrences of the known motifs of the TF. As the known motifs, we use every CIS-BP motif [ 33 ] that is derived based on direct binding evidence in human. In a post-analysis checking, we confirmed that the discovered non-canonical motifs do not match with CIS-BP motifs derived from indirect evidence.
Step 3. Following the approach of Slattery et al. Step 4. We perform an initial filtering on L -mers based on the following two criteria. Secondly, we computed the entropy based on dinucleotide frequency of each L -mer, and discarded all L -mers that have an entropy lower than the minimum entropy of a CIS-BP motif occurrence.
Step 5. To estimate the count of an L -mer in round 0, we build higher-order Markov models of the round 0 oligos following Slattery et al. Step 6. We discarded all non-motif L -mers that fail to satisfy the following three criteria. First, we discard all non-motif L -mers that have an enrichment lower than the least enriched motif L -mer. We discard a non-motif L -mer if its round-over-round enrichment for any pair of successive rounds is less than 1.
Step 7. We separately cluster the motif L -mers and the non-motif L -mers into canonical and non-canonical motifs. To ensure the reliability of the clustering algorithm, we confirm that the canonical motifs are similar to the CIS-BP motifs of the corresponding TF.
When counting the number of oligos where a non-canonical motif occurs alone, we also ensure that the same oligo is not counted more than once for different non-canonical motifs. Step 8. Based on the similarity scores of the canonical motifs, we then compute the empirical one-tailed p -values for the similarity scores of the non-canonical motif.
We followed Slattery et al. For each round 0 library, we first shuffled the order of the sequences and partitioned those into two equal sized datasets for training and validation. We then computed the optimal order of a Markov model for that library by fitting Markov models of order between zero and an integer M on the training sequences, and comparing the model performance coefficient of determination, R 2 on the validation sequences.
We say that a k -mer occurs in a DNA sequence if the sequence has a k -length substring either in the forward or the reverse complement direction exactly matching the k -mer. According to Slattery et al. We then set L equal to the value of k for which the above KL divergence was the maximum. Like Slattery et al. S is the set of all k -mers that occur at least times in the selected round R ,. Minimum count: we eliminate an L -mer if it occurs less than times in the library sequences.
We defined the di-nucleotide based entropy for a given sequence as follows. We identify the lowest ranked motif L -mer and discard every lower-ranked L -mers. Round-over-Round enrichment is at least one: for each non-motif L -mer, we computed its enrichment between every pair of consecutive rounds Round-over-Round, RoR enrichment. We take the filtered lists of motif L -mers and non-motif L -mers, and cluster the L -mers into canonical and non-canonical motifs.
The key idea is to iteratively identify a cluster head defined below and cluster all the L -mers that: a have not been assigned to any other cluster yet and b are covered by the current cluster head defined below. These choices were adopted from previous string-kernel based support vector machine models of TF binding specificity [ 2 ]. We say that a cluster head covers an L -mer if it occurs in the L -mer with up to m mismatches.
Intuitively, a cluster head identifies a core region within the L -mers that it covers. After we cluster the L -mers covered by the current cluster head, we identify a new cluster head for the remaining L -mers and repeat the same process. We continue this iterative process until every L -mer has been assigned to a cluster or we have identified a maximum number of clusters we set the limit at five.
We next align the L -mers in every cluster. We identify the position within each L -mer where the cluster head occurs with the fewest number of mismatches.
We call these positions the anchor positions for alignment. If there are more than one anchor position for an L -mer, we choose the one that is closest to the middle position of the L -mer. Of note, we always count mismatches by considering l -mers in both the forward and the reverse complement orientation. From these alignments, we finally create the position weight matrices or motifs by counting the number of occurrences of each nucleotide at each position of the alignment.
We visually confirmed each canonical motif constructed from the above process and confirmed their similarity with the CIS-BP motifs of the same TF Supplementary Table 1.
It is useful here to mention a final point about the non-canonical motifs constructed in the above process. At any stage during cluster construction, if we find multiple cluster heads i.
In such cases, the same L -mer will be assigned to more than one cluster and thus, will contribute to more than one motif. It is not clear how this may influence our downstream analyses. Therefore, after performing multiple test corrections on the non-canonical motifs see below , we manually check if there is any pair of significant non-canonical motifs that includes the same L -mer and keep the motif that is more different from the CIS-BP motifs see below.
Thus, in our results, an L -mer never occurs more than once in the non-canonical motifs. We say that a motif canonical or non-canonical of a TF occurs in a sequence an oligo or a ChIP-Seq peak if any of its constituent L -mers occur in the sequence. When a non-canonical motif of a TF occurs in a sequence, but none of the canonical motifs of the TF occurs in that sequence, we say that the non-canonical motif occurs alone in that sequence.
To compute the distance D between two motifs, we first trim the motifs by eliminating non-informative positions information content less than 0. Then we consider the every possible l -length sub-motifs see below of the two trimmed motifs, compute their Euclidean distances normalized by l , and report the minimum of these normalized distances as D. As we did for cluster heads above, the l -length sub-motifs capture the similarity between the two motifs in a core region. While computing the Euclidean distances, we always consider one of the motifs in both forward and reverse complement orientation, and take the smaller of the two distances.
Next, we compute the statistical significance of the D min value of each non-canonical motif by computing a p -value using a normal distribution with mean and variances computed from the D min values of the canonical motifs. We report this p -value as the statistical significance of the non-canonical motif. The role of non-canonical signalling in the context of CKD is not well understood relative to canonical signalling; however, the integrated relationships of these pathways suggest that normalizing both Wnt signalling pathways may be an attractive direction for future research and therapeutic development.
Specifically, excessive canonical signalling appears to promote the fibrosis and decline in kidney function in CKD and increasing non-canonical signalling to inhibit canonical signalling may be protective in this context. However, effective pharmacological control of RAS has been difficult given no overall reduction in CVD risk; this is likely due to complex compensatory regulation of RAS.
Progression of atherosclerosis begins with an initial endothelial injury and subsequent endothelial dysfunction. In the context of atherosclerosis, vascular cell adhesion molecule 1 VCAM-1 is critical for recruitment of inflammatory cells to the atheroma. Canonical Wnt3a signalling has been shown to reduce VCAM-1 expression and can limit export of myeloid cells from the bone marrow.
Non-canonical Wnt5a is an inflammatory stimulus that has been shown to regulate macrophage phenotypic shifts in the atheroma. Our group has recently demonstrated non-canonical Wnt5a signalling components correlate with plaque severity and, furthermore, lead to increased lipid accumulation in macrophages and foam cell formation unpublished.
Furthermore, silencing Wnt5a expression results in inhibition of progression of atherosclerosis and a reduction in the size and severity of plaques without affecting blood lipid levels. Proliferation, migration and differentiation of smooth muscle cells into a myofibroblast phenotype are important for plaque stability. Inhibition of canonical Wnt signalling may be particularly beneficial in preventing restenosis and vein graft failure.
Recently, Pandey and Chandravati have suggested that Wnt signalling can complement public health initiatives in predicting occlusive vascular disease restenosis after coronary artery bypass graft surgery which is a major public health problem in the United States. Identification of new molecular mechanisms in coordination with clinical data is crucial in the development of cost-effective biomarkers for this at-risk population. Thus, therapeutics targeting Wnt signalling may provide a novel way to combat atherosclerosis.
These features make the Wnt signalling pathways an attractive therapeutic target. Although evidences presented in this review demonstrate some functions of Wnt signalling pathways in CMDs through homeostatic and inflammatory mechanisms Figure 2 , the precise role of Wnts in a patient with multiple disease co-morbidities remains unknown. Overarching hypothesis. Chronic metabolic disease involving fibrosis and extracellular matrix deposition have dysregulated Wnt signalling and a tendency to have excessive activation of the canonical Wnt signalling pathway.
Chronic metabolic diseases involving inflammation and lipid accumulation have dysregulated Wnt signalling and a tendency towards excessive activation of the non-canonical Wnt signalling pathway. Excessive non-canonical signalling has also been shown to drive insulin resistance in adipose tissue and liver.
The role of canonical and non-canonical Wnt signalling in kidney disease is less clear. The intricate balance of both arms of the Wnt signalling pathways with fibrosis, inflammation and lipid accumulation may contribute to the development of these chronic metabolic diseases. The pervasiveness of aberrant Wnt signalling in CMDs makes it an interesting target as a therapeutic as well as a potential biomarker. The unique aspects of canonical and non-canonical Wnt signalling are additional challenges when considering the development of therapeutic strategies such as avoiding potential off target effects of Wnt signalling modulation.
Discrepancies in the literature concerning overall effects of canonical and non-canonical signalling appear to depend on the affected tissue, reiterating the importance of avoiding off target effects. Additionally, there are mixed reports of the role of Wnt signalling depending on in vitro and in vivo conditions and echoes that the Wnt signalling pathways are highly cell and context dependent. A concept to highlight is the complexity of Wnt signalling at the systemic, receptor and intracellular levels.
Also, the actions of Wnts are implicated in multiple organ systems. The Wnt signalling pathway may also play distinct roles at various stages of each disease over time. These additional levels of complexity represent additional barriers when developing a therapeutic intervention.
In conclusion, the literature supports an interrelationship between the Wnt signalling pathways as a potential pathological mechanism in CMDs, and a greater understanding of the complexity of Wnt signalling will increase the opportunity to develop effective treatments for these burdensome diseases.
Wnt signalling is implicated in lipid accumulation, fibrosis and chronic low-grade inflammation in lifestyle-related chronic metabolic diseases. Targeting the Wnt signalling pathways may be novel approach to treating lifestyle-related chronic metabolic diseases. Given complexity of the Wnt signalling pathways and their role in multiple organ systems, it is important to avoid potential off target effects of Wnt signalling inhibition. Clip art for figures: Library of Science and Medical Illustrations.
National Center for Biotechnology Information , U. Diab Vasc Dis Res. Published online Nov 8. Ian Ackers 1, 2 and Ramiro Malgor 1, 3. Author information Copyright and License information Disclaimer. Email: ude. This article has been cited by other articles in PMC. Keywords: Wnt signalling, metabolic disease, diabetes, non-alcoholic fatty liver disease, kidney disease, cardiovascular disease.
Open in a separate window. Figure 1. Characterization of Wnt proteins Wnt proteins are approximately — amino acids in length with a conserved cysteine-rich binding domain consisting of 23—24 cysteine residues.
Wnts as regulators of metabolism The perpetuation of chronic low-grade inflammation is a common denominator in CMD. Wnt signalling in CMDs Obesity and diabetes Adipogenesis Approximately two-thirds of adult Americans are considered to be obese, and obesity is a risk factor for diabetes, fatty liver disease, CVD and cancer.
Diabetes Non-canonical Wnt signalling has been shown to negatively regulate insulin signalling in white adipose tissue. CKD CKD is defined as progressive loss of kidney function over time, most commonly because of diabetes, hypertension and other chronic diseases.
Endothelial cells Progression of atherosclerosis begins with an initial endothelial injury and subsequent endothelial dysfunction. Macrophages Non-canonical Wnt5a is an inflammatory stimulus that has been shown to regulate macrophage phenotypic shifts in the atheroma. VSMCs Proliferation, migration and differentiation of smooth muscle cells into a myofibroblast phenotype are important for plaque stability. Figure 2. Key messages Wnt signalling is implicated in lipid accumulation, fibrosis and chronic low-grade inflammation in lifestyle-related chronic metabolic diseases.
References 1. Nusse R, Varmus HE. Many tumors induced by the mouse mammary tumor virus contain a provirus integrated in the same region of the host genome. Cell ; 31 : 99— Mutations affecting segment number and polarity in Drosophila. Nature ; : — A new nomenclature for int-1 and related genes: the Wnt gene family. Cell ; 64 : Logan CY, Nusse R. The Wnt signaling pathway in development and disease. Annu Rev Cell Dev Biol ; 20 : — Nugent R.
Chronic diseases in developing countries. Ann N Y Acad Sci ; : 70— Nutr Metab ; 8 : Arch Pharm Res ; 32 : — J Am Soc Nephrol ; 26 : — Take the Wnt out of the inflammatory sails: modulatory effects of Wnt in airway diseases. Lab Invest ; 96 : — Nonalcoholic fatty liver disease induced by noncanonical Wnt and its rescue by Wnt3a. Adipogenesis and WNT signalling.
Trends Endocrinol Metab ; 20 : 16— J Bone Miner Res ; 22 : 19— Endothelial dysfunction in human diabetes is mediated by Wnt5a-JNK signaling. Arterioscler Thromb Vasc Biol ; 36 : — Arthritis Rheum ; 44 : — Activation of noncanonical Wnt signaling through Wnt5a in visceral adipose tissue of obese subjects is related to inflammation. Open Circ Vasc J ; 5 : 1—7. Gordon MD, Nusse R. Wnt Signaling: multiple pathways, multiple receptors, and multiple transcription factors.
J Biol Chem ; : — Still, this correlation is lower than those observed for the non-canonical Wnt signatures compare lines 20 and 21 with lines 1 and 4 in Supplementary Table 2. The presence of tumor-infiltrating immune cells TIICs in the tumor microenvironment is usually associated with poor prognosis Fridman et al. Previous reports have shown higher infiltration of macrophages in GC Wei et al.
Figure 3. Correlation between gene expression, immune cell markers, and EMT. The Spearman correlation coefficients and p -values are indicated in each plot. B Correlation between ROR2 gene expression and immune cell infiltration.
C Correlation between the indicated non-canonical Wnt signatures and gene signatures corresponding to an epithelial blue or mesenchymal red transcriptional program. See main text for details. TPM, transcripts per million. Subsequently, each gene was assessed independently for their correlation with immune cell infiltration using TIMER as an alternative approach.
Collectively, these results prompted the question of whether these immune signatures were correlated with GC prognosis. The analysis focused next on EMT, a process that correlates with cancer initiation and progression Dongre and Weinberg, , including in GC Peng et al.
Recently, a comprehensive analysis of developmental signaling pathways including the non-canonical Wnt pathway , and their correlation with EMT was reported Xue et al.
The epithelial and mesenchymal signatures reported by this study Supplementary Table 1 were used to corroborate whether the non-canonical Wnt signatures correlated with either an epithelial or mesenchymal transcriptional program.
For both signatures, there was a strong correlation with the mesenchymal, but not the epithelial, transcriptional program Figure 3C. Altogether, this analysis suggests that Wnt5a signaling, mediated by FZD2 and FZD7, together with the co-receptor ROR2, might be correlated with EMT and, to a lesser extent, with macrophage infiltration, which might partly explain their correlation with poor prognosis.
To determine the molecular processes that might be differentially regulated in GC patients with high expression of FZD7 and ROR2 , differences in expression at both the gene and protein level were analyzed using cBioPortal. This list, containing genes Supplementary Table 5 , was used for an over-representation analysis in WebGestalt and Metascape.
GO terms related to extracellular matrix constitution and organization, as well as terms related to cell migration, are among the top over-represented terms, according to WebGestalt Figure 4B. Analysis using Metascape revealed additional over-represented terms Supplementary Figure 6. The two methods also identified GO terms related to cell signaling. Figure 4. For simplicity, the size of the circles is not proportional to the number of genes.
The rationale of the analysis is summarized in the diagram. Here, publicly available gene expression and clinical data were analyzed to provide an updated view about the relative contribution of the canonical and non-canonical branches of the Wnt pathway to GC prognosis. On the other hand, the non-canonical Wnt pathway is involved in GC cell migration and invasion Kikuchi et al.
However, most reports addressing potential mechanisms by which the non-canonical Wnt pathway might be involved in GC are limited to animal models and in vitro analysis of cell behavior, or were carried out using specific datasets before the release of TCGA data. Therefore, the role of the Wnt pathway in GC remains insufficiently understood, and a revised overlook is thus necessary to assess whether updated data corroborate previous associations. This study might contribute to advance our understanding of the role of non-canonical Wnt signaling in GC.
Furthermore, other reports have shown that FZD7 can activate non-canonical Wnt signaling triggered by Wnt5a in different cellular contexts Nishita et al. Therefore, FZD2 might be unable to compensate for signal transduction from endogenous canonical Wnt ligands. However, it might compensate for other functions performed by FZD7, like signal transduction from non-canonical Wnt ligands, such as Wnt5a, likely secreted by TAMs Zhao et al.
Notwithstanding, these frizzled receptors might transduce both canonical and non-canonical Wnt signals in GC, and thus pharmacological inhibition of these receptors acquires greater relevance Koushyar et al. Also, a hexapeptide inhibitor of Wnt5a signaling, Box5 Jenei et al. It is important to note some limitations of this study.
First, analysis at the protein level is limited to the data available in the TGCA, and other proteins might also be correlated with the non-canonical signatures, leading to different or additional conclusions. Previous research has shown that KIT is altered in gastrointestinal stromal tumors Hirota et al. Mutations in MYH11 have been reported in gastric and colorectal cancers Jo et al. It has been proposed that MYH11 might be related to stem cell differentiation defects, as well as disturbed energy balance Alhopuro et al.
Moreover, these proteins might activate different signaling pathways leading to cancer progression, as seen for KIT in hepatocellular carcinoma Rojas et al. It must be noted that Disheveled is involved in Wnt5a-mediated integrin adhesion turnover Matsumoto et al. Finally, claudin 7, a tight junction transmembrane protein, has been observed to be overexpressed in GC, correlating with a shorter OS Jun et al. Interestingly, overexpression of claudin 7 leads to increased invasiveness in the gastric cell line AGS Zavala-Zendejas et al.
However, claudin 7 expression is either up- or downregulated in several cancer types, and dysregulated protein levels might interfere with its epithelial barrier function, leading to impaired tissue homeostasis Wang et al. Moreover, cytoplasmic staining for claudin 7 has been reported in GC tissues Johnson et al.
Finally, the precise role of claudin 7 in GC might depend on the histological subtype Johnson et al. Second, the data presented here must be corroborated biochemically, both in vitro and in vivo. Third, this study needs to be extended at the histological level to fully corroborate the findings presented here.
The same applies to proteins involved in the GO categories identified here as enriched in the non-canonical signature. Fourth, the result that a non-canonical Wnt signature is correlated with mesenchymal marker expression suggests that this pathway might be even more relevant than previously proposed.
Future studies should address this correlation experimentally and compare the correlations observed for the non-canonical Wnt signatures to genes associated with pathways previously shown to be altered in GC Shi, Finally, future studies should aim to understand the role of cytoplasmic Wnt components and effectors, such as Disheveled proteins, disheveled associated activator of morphogenesis 1 Daam1 , JNK, and others.
Of note, Daam1 has been associated with cancer progression in other tissues Zhu et al. In summary, the results presented in this study corroborate previous insights about the role of Wnt5a and the non-canonical Wnt pathway in GC, helping to clarify the relevance of this Wnt signaling branch in GC from a prognostic perspective and extending previous knowledge, by showing that a signature associated with this pathway correlates with immune cell signatures and mesenchymal marker expression.
Therefore, this study strengthens the need to better understand the biological and molecular processes modulated by this pathway in the homeostasis of the gastric tissue and cancer. Gene and protein lists are available in the Supplementary Information. PA conceived and designed the study, and wrote, revised, and approved the submitted manuscript. The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Alhopuro, P. Unregulated smooth-muscle myosin in human intestinal neoplasia. Anastas, J. WNT signalling pathways as therapeutic targets in cancer. Cancer 13, 11— Google Scholar. Astudillo, P. Wnt5a signaling in gastric cancer. Cell Dev. Boussioutas, A. Distinctive patterns of gene expression in premalignant gastric mucosa and gastric cancer. Cancer Res. Bray, F. CA Cancer J. Cancer Genome Atlas Research Network Comprehensive molecular characterization of gastric adenocarcinoma.
Nature , — Cerami, E. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data: figure 1. Cancer Discov. Cheng, X. Metastasis 22, 85— Cristescu, R. Molecular analysis of gastric cancer identifies subtypes associated with distinct clinical outcomes. Dong, X. Cancer Lett. Dongre, A. New insights into the mechanisms of epithelial—mesenchymal transition and implications for cancer.
Cell Biol. Flanagan, D. Frizzled-7 is required for Wnt signaling in gastric tumors with and without Apc mutations. Winding back Wnt signalling: potential therapeutic targets for treating gastric cancers.
0コメント