Harvard Medical School
Name: Dimitrios Iliopoulos
Position: Research instructor, Harvard Medical School, department of biological chemistry and molecular pharmacology
Background: PhD, Ohio State University, 2006
In a study published online in PLoS One Nov. 17, researchers showed how they integrated microRNA and proteomics data to gain a better understanding of the molecular pathology of osteoarthritis.
The study’s authors wrote that the “absence of large-scale molecular studies limits our ability to understand the molecular pathobiology of osteoarthritis and identify targets for drug development.”
MicroRNAs regulate gene expression and has been implicated in important cellular processes such as lipid metabolism, apoptosis, and organ development, and has also been associated with “well-defined clinicopathological features and disease outcomes, the researchers wrote.
“It is well-known that microRNAs exert their biological functions through suppression of their target genes,” the authors said in their PLoS One article.
While bioinformatics algorithms have been constructed to predict microRNA gene targets, it has been shown microRNA regulates gene targets only at the protein level, not on the mRNA level, and a recent study has suggested that microRNA can repress the production of hundreds of proteins.
“Therefore, it becomes evident that proteomic data are needed in order to accurately detect microRNA gene targets,” the authors said.
Their work led them to identify 16 microRNA osteoarthritis gene signatures. Using reverse-phase protein microarrays, they detected 76 differentially expressed proteins between osteoarthritic and normal chondrocytes. Further, integrating the microRNA and proteomic data with microRNA gene-target prediction algorithms, the researchers created an interactome comprising 11 microRNAs and 58 proteins linked by 414 potential functional associations.
ProteoMonitor recently spoke with the first author of the study, Dimitrios Iliopoulos. Below is an edited version of the conversation.
What is the major take-away message from this study?
There are two take-away messages from our study. The first one is that osteoarthritis is known as a multi-factorial disease; it’s a complicated disease [dealing with] the destruction of the cartilage due to many different factors. And it affects more than 20 million people in the US.
This disease has very high prevalence in the US and other countries [but] there are very few studies concerning the molecular biology of this disease.
There have been many clinical studies that have implicated the causative role of obesity, for example, in osteoarthritic development. So our study is the first that shows on the molecular level that there is a link between obesity, inflammation, and development of osteoarthritis.
The important thing is that we were able to connect all this using high-throughput technologies and we performed microRNA [and] proteomic analyses, and we integrated the data … using different bioinformatics tools.
Is this the first study that looks at osteoarthritis on this microRNA and protein level, and then integrates the data?
We know that … there are non-coding RNAs and they seem to play a very important role in many different diseases. Also, it has been shown that microRNAs can regulate gene expression on the mRNA and protein levels, and in many cases they affect only protein levels.
An important [goal] in the microRNA field is to try to identify gene targets for the microRNAs. … So one could do this [with] different bioinformatic databases that are publicly available. Previous studies have tried to match microRNA data with cDNA data. However, they miss many targets because microRNAs may affect protein levels and in many cases, they don’t affect the mRNA levels, so by matching cDNA and microRNA data, you miss a lot of microRNA gene targets.
However, in our case, we matched microRNA protein data, and actually this is the best thing somebody can do in order to identify microRNA gene targets. We tried to see which differentially expressed microRNAs that we found in osteoarthritis potentially target differentially expressed proteins in the same tissues, in the same samples.
This is important because, of course, we had real data but we don’t have … the bioinformatics predictions. Those predictions can give you proteins that may be in different cell types, and here also, we have quantification. We know how much microRNA is differentially expressed.
For example, microRNAs that are overexpressed target genes that are down-regulated, so by having the data from the microRNA and the proteomic analyses, we were able to find 32 pairs of microRNA regulating proteins.
So we think this is a very good strategy not just for osteoarthritis … but generally in more complicated diseases like cancer or diabetes [where people] can perform both analyses in the same samples and it seems that the integration of microRNA and proteomic data … gives very good results.
Actually, we were able to verify this further analysis. That’s we propose this integration of data seems to be effective [in predicting] microRNA targets.
Were you looking at very specific gaps in gene targets, or were you leaving it open as to what you would find?
The target we took was unbiased in the beginning because we take the data so we see the results from the databases, and the databases, for example for microRNA, may have 150 gene targets.
We took the data from three different databases … and then according to the predictions of these databases, we … see what happens, and we were able to identify the pairs that these databases predicted.
In the paper you describe an interactome network involved in the pathogenesis of osteoarthritis. Can you describe it?
Actually the interactome that we identify is not just, let’s say, the classic interactome where you have protein-protein interactions, but what we call the interactome is any kind of interaction. For example, when a transcription factor maybe regulates the expression of the gene … or [when] microRNAs regulate gene expression [at] the post-transcriptional level, or you have protein-protein interactions, you have protein modifications, so the interactome that we tried to build includes all these modifications that come closer to a real system than the classical system where you have only protein-protein interactions.
We tried to include more than one way of gene regulation … and the important thing is that in cartilage, there was very few data in comparison to other tissues and other diseases.
So actually, it was to characterize the proteome of what proteins are potentially differentially expressed with a normal [control] and cartilage from people with osteoarthritis.
In order to [decide] which proteins to include in our [study] we have looked at the literature. There are some studies that performed mass spectrometry and they have identified some proteins, so we included all the proteins that were potentially cartilage-related or bone-related, but we tried to include as many proteins as we could.
And with these technologies that we used, we were able to have on the array 214 proteins, which is a big [sample] in comparison to previous studies, where usually using reverse-phase protein arrays, [people] have used 40 to 50 proteins.
Why did you choose to use protein microarrays?
There are different protein arrays. There are many different technologies and every day … we have more and more new technologies. We did not use antibody arrays. It’s a good technology … but the problem is that you can have many samples and perform analyses for one protein. In our case, we didn’t know the proteins, so actually we wanted to test a lot of proteins, so it was not appropriate for our study.
Another thing is the amount of proteins that we wanted was very low in comparison to other things, so [the method we chose] is rapid, it’s automated and the analysis performed similarly to the cDNA microarray.
We were able to validate our results. We have performed Western blot analysis, and this was easy because we performed this protein arrays using these antibodies, so we knew already that they are working for these proteins.
One major thing for this analysis is to find good antibodies. Not all antibodies work well.
And in mass spectrometry … the problem is that … it misses a lot of proteins because many proteins are expressed at low abundance, and in our case, in cartilage, you have proteins like that, so with mass spectrometry we would miss many proteins that we have detected now.
Would a high-throughput technology like mass spec have any use for the work you’re doing?
I think if the cost was not so high and was more efficient.
Did you build your own arrays?
Most of the stuff, we have made. … Actually, we’re planning in the future to use other protein array methods, for example, to clone thousands of ORFs in vectors and perform cell-free system protein synthesis in order to have thousands of proteins that we can put on an array.
You found 76 proteins that were differentially expressed. Were any of them of particular interest?
We found several proteins [which] I think are very important, and our study is more like a pilot study … so we found proteins like integrin a5. … We found this is highly regulated in osteoarthritis, and it was also highly correlated with body mass index, so probably this is one potential link on the molecular level, how obese people, they have more pressure and this [protein] is activated and starts to cause destruction in the cartilage.
And we have connected inflammation, obesity, and osteoarthritis. And a surprise to us was that we found a protein called PPAR alpha. It’s a well-known receptor [and has been studied for its role in diabetes] but nobody has studied it in cartilage or arthritis, and we have found that this receptor when it’s inhibited, it causes up-regulation of interleukin … and also it was very interesting that this interleukin finally regulated the whole gene network that consisted of cartilage homeostasis proteins.
So an interleukin and inflammatory molecule was able to control structural proteins, so this is an important finding.