 Hello students, I hope you are enjoying various applications which can be done using different type of protein microarrays, you have got exposure of variety of platforms and different type of applications which are possible by using protein microarrays. I am going to give you a glimpse of another type of clinical application using human proteomers. I will focus mainly on the brain tumors. So, let us talk today about auto antibody screening in brain tumors. But before I come to the workflows, I would like to tell you that when you have variety of samples from which you can do proteomics investigation, especially in the clinical context, then you can choose different path, different technologies to address different type of biological questions. Here I have shown you two different path, one essentially to look for the most abundant antigen screening which one could do either using the gel based platform or using mass spectrometry based proteomics. Alternatively, especially for various type of bio fluids, one could use protein array platform and look for the auto antibody detection. Of course, in the same samples, one could obtain more complementary information and all of that could contribute towards the systems level information. If you recall from my very beginning lecture, when I talked to you about different technology platforms in the field of proteomics and how field is going to progress more towards system biology, I think we have to become very unbiased and appreciate variety of technologies and what kind of new insight they can offer from the same type of clinical samples. In this manner, it is important that we should understand that which technology, which platform can provide what kind of unique insight. So, this work what I am going to describe today essentially proteomics based investigation of brain tumor, focus on the auto antibody screening. This work was done in collaboration with Tata Memorial Hospital in Mumbai, where we obtain variety of clinical samples from the patient, its tissue, serum, plasma or service panel fluid. All of these kind of samples provide some unique and interesting information. And as I mentioned, you can now decide what is the balance question you want to address and accordingly, you can choose different type of platform either mass spectrometry based or gel based proteomics investigations or using protein microarrays, which I have shown in the middle panel here that the discovery phase could be done using variety of platform from the proteomic technologies. But eventually it comes the validation and validation one could do either using antibody based approaches or one could also use the mass spectrometry based new assays which are coming forward, especially based on the selected reaction monitoring where you are not involving antibodies and directly measuring the concentration of a given peptide or protein using mass spectrometers. Again the entire field of mass spectrometry based proteomics and target proteomics in itself is a full content for the course, but here I am kind of trying to give you the balanced view that how you should be applying the same samples to understand different type of questions by applying on different type of technology platforms. So, question is can we identify the auto antibody signatures emerging from the brain tumor patients? This is one of the very interesting hypothesis where many reports and many publications, they have reported that in addition to the auto immune diseases even in different type of cancer, auto antibodies are emerging from the patient samples and especially in the cancer, the auto antibodies are generated against the tumor derived proteins which are either absent or barely detectable in the normal tissue or in response to the post translation modified proteins. These auto antibodies appear in the blood serum or plasma much before the complete establishment of disease in many cases. So, one could actually use them as early detection biomarkers because one could do the testing from the blood with a routine kind of blood based test which people usually do for the health checkup. So, protein microarrays provide one of the strong platform to look into these auto antibody screening from the patients serum or plasma samples. So, as we have seen that protein microarrays they offer variety of applications and different ways one could make the arrays and finally use those protein array platform which is essentially the miniaturized arrays when you can print your proteins or your CDNA and you want to make the proteins from that which could be utilized for many applications. But biomarker screening using auto antibodies is one of the very important application which is we are going to focus today. So, as I mentioned the onset and progression of cancer often results in expression of mutated and apparently express proteins which could act as the self-antigens. Now these self-antigens they evoke an immune response which leads to the production of auto antibodies. Here I have shown you a workflow how to use human proteome arrays for detection of auto antibodies. Just keep in mind that when we are talking about human proteome arrays one could either use the purified proteins or use the CDNA NAPA based approaches. In this case we are using purified protein based arrays which consists of almost 18,000 human proteins printed in duplicate. Idea is if we add the patient serum sample we can detect unique auto antibody signatures against the tumor antigens using protein array platform. Of course while assay looks very straightforward and simple but data analysis and further data interpretation becomes very challenging. Here I have shown you the workflow for auto antibody screening using protein microarrays. You can think about a similar experiment which you have done using western blots. Exactly same concept is here except that you do not have a membrane you are talking about the protein arrays on the different substrate especially on the glass or nitrocellulose membrane. All the proteins are printed on the chip. First you want to block your slide so that all the other region where proteins are not printed they should not give you the background. The blocking can be done either milk dissolved in the PBS or you can use the super block. Once blocking step has been done then now you can use the arrays where you can add the patient serum sample. And think about we are talking about very thin glass slides where you can add even simply 10 to 15 or 20 micro liter of the patient serum sample which is diluted in the buffer and therefore even hardly with 200 micro liter you can cover the entire slide. Idea is that these patients serum sample contain antibody against tumor. And if these antibodies are produced then can they bind to the antigens which are printed on the chip. So we have 18,000 human protein antigens printed on the chip and these antibodies which are part of the serum sample once they come on the chip can they come and bind. And if they are binding then one could try to detect those antibodies using secondary antibodies which is anti-human IgGs linked with the Psi 3 or Psi 5 or various type of alexa fluorophores. So the idea is the primary antibody is coming directly from the serum sample and secondary antibody you are adding either anti-human IgGs or IgM and then further you can detect the signal with any different type of detection strategies. So far I think you got a glimpse that doing the protein array based experiment is not very difficult and you do not need lot of you know big setup and big instrumentation in any standard molecular biology lab these experiments can be done. But when we talk about high support experiments you have to keep in mind that these experiments has to be done very meticulously and lot of attention has to be paid and what kind of data are you going to get high throughput does not mean that you know you can get data in a very high quality in a very very short time high throughput requires lot of attention lot of you know careful thoughts that how well your experiments are proceeding is the data what you are going to generate is of high quality and can be rely on that data. So in this slide first of all if you have the printed slide you would like to make sure that protein is actually printed properly or if you are using NAPA arrays so proteins are expressing properly. So if all the clones which you have printed eventually which from which you made the protein if they contain some tag for instance the GST tag or MIG tag any other tag now if you can use the antibody against that so all the features where the protein has been printed or protein is expressed then one could test out from the antibody. In this case first we want to test on the human proteome arrays which we collaborated from the Johns Hopkins laboratory in US. So these arrays contain the purified protein and in each one of these clones contain the GST tag. So as you can see on the slide that first we want to make sure that wherever the proteins are printed those are all showing us the signal which is coming from anti GST antibody detection. Also we have the you know series of purified proteins printed for the GST protein which shows the gradient different concentration of the GST spots from the lower to higher concentration which also helps us to decide that you know how much protein is expressing on any given feature. The next thing comes that you know you have now the patient serum sample you want to apply that on the human proteome arrays. Idea is what should be my best experiment which I should try to apply on the protein arrays. If we are let us say having the serum with very high concentration that might get very sticky and you see a lot of background a non specific signal. So you have to make the first appropriate dilution of the serum sample. Then further how long I should keep the serum on my you know the slide should I keep it just for an hour 2 hours overnight very long incubation just on the room temperature or doing the 4 degrees or 37 degrees variety of you know optimizations are required right. So before you actually start applying your 100 patient sample you know it is good idea for you to optimize the signal first and that is what we try to optimize the assays. If you look at the left side panel on the slide the top the center image show that you know when we are adding for the overnight incubation of the serum we see a lot of non specific signals coming. So now the first image which I showed you earlier was the GST with the you know green signal and now we are showing you the anti human IGG signal coming as a red signal here. So now we are looking at auto antibody responses and there should be response which we want to detect but it should not be everywhere which is non specific. So the middle top panel shows that there is lot of red signals which is now coming non specific throughout the slide. The bottom panel what you see like that looks pretty much clear that you know everything is not lighting up only few spots show red signal and that is with the less incubation when we have at the room temperature. So these kind of optimizations are very much required. Also what you want to test out when you are using the same patient sample applied on two different ship how much reproducibility you can see slide 1 to slide 2. Also the spot to spot reproducibility across different ships. So I hope you recall from my previous lecture that how much you know the QC checks are important right. So now we are doing the actual project actual experiment here. Aim is to look at the brain tumor patients whether there is auto antibody response and can we identify some biomarkers. But before we start actual patient serum sample screening we would like to make sure our assays are working fine and the reproducibility of first say has been determined. So on the right side you can see we have measured intraceptic reproducibility, intraceptic reproducibility and on the bottom panel you can see we have looked at day-to-day variations across different type of spots and different type of ships. Once we have seen that okay our CVs are tight assays looks fine then only we are ready to proceed for the actual experiment. Additionally we have to also keep an eye that when we are doing the assays our positive controls and negative controls they are behaving the way they should. So for instance in this case we have if we are applying the anti-human IgGs. So IgG spots which are purified immunoglobin proteins printed on the chip they should definitely show as a positive controls. And that is what you see in the in the middle here the positive control with the red spots are seen whereas when we have empty spots or the spots you know when we have only vector we do not have the the genes printed on that. So then those should be negative controls when we are not expect to see a protein expression happening and those are the one which are shown as the totally black spots which are the negative controls. So again now you see that okay how much we have already looked carefully a variety of features before we actually start applying our patient samples and start investigating the questions of biological interest. And further once we have done all the screening then one need to look at different ways of a statistical analysis to look for that is how my signals were appearing before normalization. So if you look at the left hand side variety of you know different grade of patient samples they are showing different level of signals. But we need to normalize all of them so that every signal now start from the same baseline and then only we are able to calculate what should be the protein expression change for the up regulation or down regulation. And then right side shows you the schematic that after doing those kind of normalization then one could proceed for looking at the significant proteins the best classifier using multidimensional scaling analysis and also look at various type of heat maps and find out the best classifiers. I hope now you are clear that you know what one should do before starting a project of you know the clinical interest or biological interest. So now let me kind of walk you through a little fast but the idea is to give you the glimpse how one could now apply this information this knowledge of doing an assay on the protein arrays for the actual patient related problem. So I am going to talk to you about some deadly brain tumors the gliomas, meningomas and how from the serum and CSF cerebrospinal fluid one could use the same protein array platform and look for the auto antibody biomarkers. So I do not want to talk too much in depth about biology of these diseases but just to reiterate that these are one of the most challenging tumors which you know are very aggressive in nature especially the glioblastomas and these are the most commonly found brain tumors especially melangeoma shares the largest fraction of all the brain tumors. So just to start we are going to talk to you about the gliomas first which have the origin from the glial cells and ideally you know we have the patient which are having the low-grade tumors or the high-grade tumors the low-grade gliomas grade 1 and 2 they are slow growing and you know they are definitely mild in nature not as aggressive as you know you will see in the higher grades but as that this progresses to the grade 3 and grade 4 we start seeing that the they are very malignant aggressive and invasive nature and then eventually the glioblastoma or the grade 4 forms which are highly malignant most aggressive and most deadly brain tumors which also shows the rapid angiogenesis necrosis etc. The median survival of these patients are very less sometime a year or 15 months therefore you know detecting these kind of diseases at early stage probably can help better diagnosis and better therapeutic strategies for these patients. So for doing this project we took the patients from different grades of glioma who are suffering from different grades of the brain tumor the grade 2, grade 3 and grade 4 as I said grade 4 are the most deadly glioblastoma multiformi patients and we had this human proteome arrays the first version which we use for this project was having 17,000 full length protein all of having GST tag an idea was to apply the patient serum sample from different grades of these patients one at a time on each slide and then you scan the images and then eventually use the image analysis for looking at any pattern emerging from these images for the different grades of tumors. So as I mentioned that you know while screening is not very difficult to experiment but could do you know 10-20 slides per day but eventually when you want to put together the hundreds of you know patient data images then the data analysis data normalization becomes very challenging and this is what we encountered here the different you know grades of patients and the control we were doing in batches and then before doing the analysis we could see that you know their signals are not same. So we did the data normalization and further after doing all the statistical analysis we found that the grade 2 patients having the less number of autoantibodies only 6 we could detect whereas grade 3 showed the highest number of autoantibody response which is 177 and grade 4 showed the moderate numbers. So why you know we are very surprised that why we could see very few number of autoantibody proteins in grade 2 but when we spoke to our clinicians and our collaborators they are not much surprised reason being that if you think about the brain tumors especially the gliomas they are protected with the blood-brain barrier and once the tumor start growing much further and starts going to the aggressive stage of grade 3 and grade 4 then only blood-brain barrier is going to breach and then many of these proteins are going to infiltrate and reach to the bloody screams. So therefore very few proteins are ideally expected to see in the grade 2 in the blood and grade 3 should have large number of proteins and because the grade 4 everything you know body's immune system is shutting down everything is almost deregulated so very few changes we are seeing at the grade 4 level. So this was actually clinically quite interesting observation and now the idea was which are these proteins which we can use for our further analysis. Additionally if you see these Venn diagrams there are 4 proteins which are showing they are common in grade 2, 3 and 4. So can we find some protein which are showing common you know at least these are commonly present in all the tumor but if there is a trend can we detect them at the very early stage because another idea for this project was to look at early detection biomarkers. So after these analysis then one would like to find out can my proteins which we have identified segregate these you know the brain tumor patient with the healthy controls. So that was the idea and we used you know the multidimensional scaling support vector machine based analysis and try to segregate the control with the gliomas. But of course the success was not much as you could see that the pattern was not very clear that you know the healthy individual versus gliomas we could not segregate very well based on our protein signatures. Then we started looking at can we look at control versus grade 2s, control versus grade 3s, control versus grade 4 and as we started analyzing you know variety of these patterns we could see some trend in grade 3 versus control and some trend in grade 4 versus control. But of course given the nature of the disease which is very complex it was really not expected that we can segregate it you know magically from the healthy individual to the disease individual and within disease so easily that which grade these patients come from. Nevertheless as I mentioned earlier we found those 4 protein which were common in grade 2, 3 and 4 and our idea was can we start investigating these proteins further and what are these proteins. So these proteins as you can see on the screen one is sort in next in one protein which is usually shows over expression which leads to the EGFR degradation lot of interesting biology is already understood for this protein and this protein if you can look at from the control to grade 2 you can start seeing some signal is coming from the grade 2 patients and as this is progressing to grade 3 than 4 we can now see more intense signal in grade 3 than 4. So what we are looking at we are looking at can we find some protein which can be detected at the early stage of the tumor and that is where I think SNX1 looked very interesting that you know even in the grade 2 patient we can start seeing those signal. If you look at IJG1 this is again one of the immunoglobulin family protein which might place an important role in the immune evasion mechanism and this protein again showed a very strong signal even in the grade 2 and that remain consistent in grade 3 than grade 4. Now the other 2 protein which we identified is EYA1 protein and PQBPA1 protein. So EYA1 protein which is the I is absent protein that plays an important role in the innate immune response and also in the DNA damage repair. So it is already known to play an important role in cancer but what we are finding interesting that the level of this protein is high in the control and as the disease is progressing from grade 2, 3 and 4 the signal for these spots are going down. Another protein which was polyglutamine binding protein 1 or PQBP1 protein again if you look at their signal from the control to the grade 4 the signal is going down and this protein is again involved in inhibiting the transcriptional activation of another protein BRN2 which is associated with development of the glial cells. So ideas can be start looking at some of these biomarkers and look at what are their trends across other diseases as well because many proteins might be showing response just generic response for any tumor type. So we started looking at the level of SNX1 across other tumor type as well by looking at Oncoma database and what we can see here that if on the right side if you look at the image that especially in the brain tumors and senus tumor the level of SNX1 is very high whereas other tumor type it is showing the low expression. So definitely this could be good marker, good candidate to take it for further investigation for the clinical utility. Additionally by looking at these tumor associated antigens and we try to map them that which pathways they belong to. So we looked at the grade 2 proteins, grade 3 proteins and grade 4 proteins that these proteins are mapped in which pathways. As you can understand that you know grade 2 we had very less number of proteins so we could not get some interesting pathways for that. But as the disease started progressing from grade 2 and 3 we see that you know many interesting pathways especially the you know the TGF pathway, the wind signaling such as skeletal remodeling, integrin mediated cell activation signals and variety of you know the androgenic receptors all of them started appearing on the interface of grade 2 and 3 and as disease started progressing more then we see there are more new pathways emerging. So again if you look at the bottom panel here we have hyaluronic SA.CD44 based pathways, we have chemotaxis neutrophil migrations and we have impaired inhibitory action of the lipoxin on neutrophil migration in CF patients. So some of these are showing how the disease is progressing towards the higher grades and again if you are now you know look at the disease progression moving from the grade 3 and 4 now we see some new path will emerging like IL4 signaling pathways, HSP60 and HSP70 pathways. So in turn you know while doing this analysis what we realize that some new biology is appearing which we can now see that you know as disease progressing there are new and new pathways which are also emerging based on the different candidates which we see the progressing from the low grade to the high grade of gliomas. Of course once you have the platform and each patient you are screening individually then you can do many different ways of investigation. Now our tradition collaborators they had interest that you know how these you know glioma multifarmy patient which are the most aggressive type some of them survive better some of them do not survive well in some patients the tumor is very aggressive in some tumor is not aggressive. So how best we could look at this kind of profile and try to look at the impact of the location of tumor on the patient's survival. So they had some interesting observation based on the radiological images and in this case as you see in the on the side we have the patients where the tumor is very close to the sub-integral zone region and those are known as SVD positive or tumor is moving for a part from the SVD region and those are known as SVD negative. So SVD positive are you know they are very very aggressive tumors as to the negative are less aggressive tumor. So of course the patient's survival and prognosis will be based on some of these you know features and can we find out some proteins which might be interesting when we are looking at the impact of the location of the tumor based on radiology. So in this case we identified an interesting prognostic biomarker which was lead 9 and lead 9 is one of the neural precursor cell expressed developmentally down liquid protein 9 which is involved in invasin and the cell migration. So what we are talking right now is the same grade of the patient which are the grade 4 GBMs and within the same grade of patients some patients are having the tumor in a different location SVD positive and some in the away from the SVD region which are SVD negative. So it is very challenging you know problem for us to address because patients have the same grade but just one minor change with the tumor location. Based on this we found very few protein which were actually differentially expressed from the SVD positive versus SVD negative. But this protein lead 9 looked interesting and we still see some differences coming when we compared for sizable number of patients within the SVD positive and SVD negative population. Additionally we also looked at are there some prognostic markets of glioma based on the mutation effects. So IDH1 or isosetrate dehydrogenase is known of the one of the very interesting gene which we have shown a lot of mutation in the glioma patients. So now the next investigation we want to do we already have the data from individual patients from each slide. Can we now look at just the effect of these mutations on these patients and look at you know are there some interesting differences we see from the wild type and the mutant of the IDH population. So all these patients were also sequenced for the IDHG and based on that then only few patients where we could get the confirmed sequence data for IDH mutation. Only those we took forward for analysis based on the protein array autotubody screening. And in this case we could identify almost 22 proteins but 2 protein which looks very interesting where YWHAH protein and step 1 protein. So YWHAH protein that is actually known to be involved in the proliferation of glioma cells and it was found to be upregulated in the wild type cohort as compared to the IDH positive cohort. Now it is possible that this kind of you know deregulation might be attributing towards the poor prognosis of wild type patients. So we can conclude this part that based on protein arrays based screening in a very less effort one could actually do the screening of thousands of protein from the patient serum sample. Of course the most challenging part is the data analysis part and doing the further data investigation using different type of pathway analysis and making biological sense of the data. So doing experiments are much simpler as compared to doing the further data processing and data analysis aspect. Based on these analysis we found the set of proteins especially 4 proteins which looks like promising candidate for the early detection of the tumor especially from the serum sample. We also investigated the effect of the tumor location based on radiology whether that could also reflect in the serum sample at the auto antibody level and one protein NET9 definitely showed promise. Additionally we could also see some other protein like hemo pexin and SOX2, HDAC7 those are also interesting protein in to differentiate these type of SVD behavior. We also looked at are there effect of the IDH mutation on these patients and we could see around 22 proteins were different in IDH positive and the wild type population. So in general I hope you got a glimpse that how to perform the auto antibody screening using a patient serum sample and what is the workflow involved how to do the data processing and finally how to make some meaningful insight from the data set. So this work was published in scientific reports which was you know a big collaborative work from the clinicians part of Memorial Hospital from some of the technologies on the protein array platform from the Johns Hopkins University and of course my team and in a variety of PhD student and postdocs from IT Bombay. So again you have to also appreciate that in this kind of research you need to make good teams and you need to bring the interdisciplinary strength to really try to achieve some very interesting information which is otherwise not possible to obtain all kind of you know specialization just from your own way. So good to build the team which are all having the different type of strengths and then try to work on a given problem and try to see what kind of meaningful insight we can obtain. I hope you got some idea that how one could use protein arrays for the biomarker discovery program and more applications are going to follow soon. Thank you.