Van Eyk, Dunn - Proteomic and Genomic Analysis of Cardiovascular Disease - 2003 (522919), страница 36
Текст из файла (страница 36)
We next took the normalized array measurementsfrom three independent hybridizations, averaged the signal intensities for each7.2 Phase 1: The Original Experimental DesignThe gf300 array from Research Genetics. 5,123 individual cDNAs are spottedonto the 5 cm ´ 7 cm nylon membrane. Eachmembrane is cut in the upper right corner fororientation. The cDNA spots are arranged intwo fields, 1 and 2, each field containing 8individual grids laid out right to left, Athrough H. Each grid is made up of individually spotted cDNA clones arranged in 12 col-Fig. 7.1umns of 30 rows. Along with the 1,683 knowngenes and 3,440 ESTs present on the gf300array, a series of controls are present including 228 total genomic DNA spots and 228beta actin spots used for normalization of thesignal and for orientation and alignment ofthe membrane when using the software fromResearch Genetics.
The spacing between eachspot is ~750 microns from center to center.cDNA and calculated the standard error of the mean. Differential expression between the RNA isolated from two cell types was determined by performing an unpaired Student’s T-test for each spot, and p < 0.05 was taken as statistically significant. Each of the three cell types under study was analyzed in a pair-wise mannerwith the others.We identified several cell-type specific cDNAs from our array analysis and published the results as an abstract at the 2000 meeting of the American Heart Association [10].
Many of the genes we identified as being differentially expressed in acell-type specific manner agreed with previously published and well known cardiovascular cell markers. In RNA isolated from cardiomyocytes, we detected specificexpression of transcripts that encode atrial natriuretic factor, desmin, and cardiacisoforms of troponin and myosin [11–14]. In smooth muscle cells, transcripts forthe SM22 protein and osteopontin were identified, again agreeing with well described expression markers for this cell type [15, 16]. These results gave us confidence that the microarray methodology we were using was working as predicted,and that the additional novel transcripts we identified as cell-type specific werealso correct.
Interestingly, we had identified multiple ESTs specifically expressedin cardiomyocytes and aortic smooth muscle, and these transcripts were to be thefocus of further study in our lab.1151167 A Cautionary Tale7.3Phase 2: Validation and TroubleshootingWhile preparing a manuscript for publication, and nearing the end of our arrayexperiments, we undertook a series of experiments aimed at confirming the arraydata with more “traditional” means of northern blotting and RT-PCR [17]. Confirming array cDNA expression data by other independent means is an importantcontrol whenever arrays are being used [18]. We selected six cDNAs that appearedto be highly specific for one or more of the three cell types under study and ordered the corresponding bacterial glycerol stocks from ResGen.
Clones were identified by accession numbers and were sequence-verified upon arrival in our lab.Oligonucleotide primers specific to each cDNA were designed for use in PCR so7.3 Phase 2: Validation and Troubleshootingthat the resulting products would only contain the insert sequences from eachplasmid. These PCR products were gel purified and used to probe northern blots,and later, for the dot blots described below.In three out of six cases, Northern blotting failed to confirm our array data.One of the cDNAs did not hybridize to RNA on a northern blot with the samecell type distribution as indicated on the array. Another two did not hybridize toanything at all on a Northern blot, even though the genes appeared to be highlyexpressed on our arrays (Fig.
7.2). RT-PCR, using the same RNA used for arrayanalysis, was also used in an attempt to confirm the expression results seen withthe array experiments. Again, in the same three of the six cases, the expressionprofile we expected was not observed, calling into question the validity of the array results.Our initial inquiries with the technical support staff at ResGen gave us no indication of how our data analysis could be flawed.
The arrays themselves were saidto be highly reliable because of the quality control measures employed by ResGen. Additionally, there had been no reports from any of the other customersusing these particular arrays that would have suggested any problems with theidentity of the spots on the array. After several attempts to validate the clones inquestion with northern blotting and RT-PCR, we decided to undertake a series of“dot blot” hybridizations to check whether or not the cDNAs that we had selectedwere indeed present on the gf300 arrays in the grid locations indicated by ResGen.Our technique was to pick a single cDNA in question, generate a clonal probeby PCR-amplifying the insert, gel purifying the product, labeling it with 33PdCTP, and hybridizing this product to a fresh gf300 array. Our expectation was3Fig.
7.2 A Individual spots representative ofhybridization signals seen on the gf300 arraycorresponding to the location of the BrainFinger Protein (BFP, Accession numberAA997188), and glyceraldehyde-3-phosphatedehydrogenase (GAPDH, accession numberAA924111). These images represent the signals seen on three individual arrays hybridized to cardiomyocyte, cardiac fibroblast,and smooth muscle cell probes. A strong BFPhybridization signal is evident with the cardiomyocyte but not with the fibroblast or smoothmuscle probes.
B Northern blot with *20 lgof total RNA from the indicated samplesloaded in each lane hybridized to the BFPprobe. GAPDH was included in the hybridization as a control. The *1.3 kb GAPDH transcript is indicated with an arrow, and the expected location of the BFP transcript at*3.7 kb is indicated with an arrowhead. De-spite the intense signal present at the BFPlocation on the array when probed with cardiomyocyte RNA, no significant BFP expressionwas seen by northern blotting.
C Dot blotusing a gf300 test microarray hybridized tothe same BFP probe that was used on thenorthern blot shown in panel A. Two high intensity signals are observed on the array afterhybridization instead of a single spot at location field 1, grid F, column 6, row 14 expected for the BFP clone. The observed positions of hybridization signal are 1, E, 6, 14and 1, F, 6, 13. The same result was obtainedfrom similar arrays from three independentprinting lots.
Upon sequence verification, theclone at position 1, F, 6, 14 was found to bean unidentified EST unrelated to brain fingerprotein. D Close up of the E–F section of thearray showing the location of the two hybridization signals.1171187 A Cautionary Talethat a single clone would produce a single, high intensity hybridization spot onthe gf300 array, and furthermore, that the spot would be present in the predictedlocation on the array grid. Arrays hybridized in this manner were processed according to the standard hybridization and washing protocol for use on a complexprobe.
This approach, which is similar to a cDNA dot blot, was undertaken on sixindividual cDNAs. (As an aside, each gf300 array cost *$ 1000.00 and we had noexpectation that the filters would be re-usable after such a test. The decision topursue this line of inquiry thus was not taken lightly.)One example of our array dot blot results is shown in Fig. 7.2 C. We selectedone of the cDNAs whose expression by northern did not match the array result(GenBank accession number AA997188, Brain Finger Protein [19]) and hybridizedit to a fresh gf300 array. To our surprise, the result gave two discreet hybridizationsignals on the array grid (Fig.
7.2), and both were in the wrong grid location. Thetwo other cDNAs that were problematic by northern and RT-PCR were similarlyproblematic on array dot blots. One of the clones hybridized to a different locationthan expected, and the other clone did not hybridize to any grid location on thearray. The three other cDNA clones whose northern analysis agreed with the original array data appeared in the predicted location as a single intense hybridizationsignals by array dot blot.7.4Phase 3: “Postmortem” Analysis at ResGenAs a result of our dot blot experiments, we feared that a large subset of the cDNAspots on the gf300 arrays were incorrectly labeled, and worse, that there would beno way to decide which spots those were. In order to figure out how widespreadthe misidentification of cDNAs on the gf300 arrays may be, we suggested thatResGen sequence verify all 5147 cDNAs present on the filters and correct the database as needed.
At the time, our suggestion was met with skepticism by ResGen, because they perform routine quality control to ensure that individualcDNAs are spotted at the correct location on the gf300 array. The quality controlmethods used by ResGen were designed to identify large scale misalignment ofcDNA spots on each lot of printed arrays. A sample from every printed batch ofarrays was hybridized to a complex probe that give a predictable pattern of hybridization, and that the filters in question had all passed this quality control test.Additional quality control tests done at ResGen would have revealed if individualplates used by the robotic printer to spot the arrays had been inserted into the robotic printer inappropriately.We sent the data files and images of our dot blot tests to ResGen and askedthat they verify the identity of the clones in question.
Over the course of the nexttwo months, ResGen was persuaded to sequence the three clones we had identified as incorrectly spotted on their arrays. They determined that in fact, the cloneswe questioned were misidentified on the gf300 array. Their interpretation was thatsome kind of rare spotting error had become incorporated into these arrays, but7.4 Phase 3: “Postmortem” Analysis at ResGenthat the error was restricted to the three clones that we had, unluckily, chosen fordetailed study.
At that point, ResGen replaced all of our arrays with fresh arraysfrom a different lot that were presumed to be correct based on their internal quality control tests. Instead of using these fresh arrays for our cell type comparisonexperiments, we repeated the dot blot experiments with the three cDNAs in question and again found they were not at their predicted locations.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
