Matrix-assisted Laser Desorption/ionization Mass Spectrometric Analysis of Dna on Microarrays

Background: Matrix-assisted laser desorption/ioniza-tion (MALDI) mass spectrometry is a powerful tool in biomolecule analysis with a wide range of application possibilities, including genotyping of single-base variations (also known as single-nucleotide polymorphisms, or SNPs) for candidate gene studies and diagnostic typing of DNA markers. We tested a method that does not require stringent purification of the nucleic acids and/or the use of modification chemistry before mass spectrometry analysis. Methods: We used an alternative direct analysis approach that allows MALDI analysis of crude DNA samples printed on microscope slides densely coated with primary amino groups that efficiently bind negatively charged DNA. After simple washing of the slides, we applied MALDI matrix and used a conventional MALDI mass spectrometer to detect DNA products. Results: We analyzed crude oligonucleotide samples and performed automated genotyping of single-base variations in 72 DNA samples with a call rate of 95.8% and reliability of 99.98%. Conclusion: This procedure offers an operational shortcut compared with standard MALDI procedures for preparation of oligonucleotides, including purification , and thus is an efficient tool for genotyping applications, particularly those requiring accurate, flexible, and rapid data generation and medium through-put.

number of marker single-base variations under investigation in many DNA samples (4 ), and MALDI provides rapid, flexible, and accurate detection of analyte molecules by use of an intrinsic physical property, the molecular mass.
Several robust procedures for the analysis of nucleic acids or single-base variations by MALDI-MS have been reported (5,6 ).Most of these methods use similar reaction sequences consisting of PCR amplification of the DNA containing the positions of interest for the genome variation, removal of residual deoxynucleotide triphosphates (dNTPs), and finally, primer extension to generate allelespecific products of DNA markers.These methods are used for genotyping single-base variations, for molecular haplotyping, for epigenotyping, for quantitative and allele-specific expression analysis (7,8 ), and for quantitative analysis of alternative splicing (9 ).
DNA and matrix preparation are important in MALDI technology but can severely limit signal intensity and resolution because the multiple negatively charged DNA polyanions form salt adducts with cations commonly used in reaction buffers.In addition, detergents are not compatible with MALDI matrix crystallization.Therefore, stringent purification of the DNA, which can be timeconsuming and costly, is performed before the samples are transferred to the MALDI target (6,7 ).Spotting of nanoliter volumes of stringently purified DNA samples on silicon chips also improved reproducibility and increased parallelization of detection (10,11 ).The only MALDI-based single-base variation genotyping methods that circumvent sample purification are the GOOD assays (12,13 ) and related procedures that use photocleavage charge-tagging with potentially toxic DNA modification chemistry (14,15 ).
Microarray technology using optical detection based on fluorescence to quickly screen the multiple data points

Molecular Diagnostics and Genetics
on microarrays has enabled miniaturized and highparallel analysis of nucleic acids, peptides, and proteins (16 -18 ).Current microarray methods, however, are not suitable for diagnostic typing of a low number of singlebase variations in potentially varying numbers of DNA samples.More flexible optical and solution-based methods such as the TaqMan TM assay (19 ) can be used but are costly because they require fluorescence dyes and quenchers.Moreover, fluorescence detection does not identify the analyte molecules and can make allele scoring and sample tracking difficult.In contrast, MS directly measures sample molecules, and MALDI-MS-based genotyping is a powerful and cost-efficient tool for candidate gene and diagnostic typing in which low numbers of single-base variations must be analyzed in large numbers of DNA samples.We present a new approach for medium-throughput MALDI analysis of DNA.

Materials and Methods
Unmodified oligonucleotides were obtained from MWG.A methylphosphonate-containing oligonucleotide was synthesized by Eurogentec.The oligonucleotides were checked by MALDI-MS with 3-hydroxypicolinic acid (HPA) as matrix.dNTPs and dideoxynucleotide triphosphates were purchased from Biolog.Taq polymerase was produced in-house (Roman Pawlik, Max-Planck Institute for Molecular Genetics, Berlin, Germany).TermiPol DNA polymerase was purchased from Solis Biodyne, and shrimp alkaline phosphatase (SAP) was obtained from Amersham Buchler.Human DNA samples, internally numbered 10, 14, and 27, were taken from a DNA bank in our laboratory.These DNAs had been genotyped previously for the same single-base variation by the GOOD assay (12,13 ) and the TaqMan assay.Chemical reagents were purchased from Aldrich.The MALDI matrix (HPA), the anchor chip targets, and the GenoTools software were from Bruker Daltonik.The adapter (shown in Fig. 1a of the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol52/issue7/) was built in-house with aluminum nickel MALDI targets from Bruker Daltonik.The thermocycling procedures were carried out in an MJ-Research PTC 200 thermocycler (Biozym).Plasticware was from Abgene and Eppendorf.Gold microscope slides were obtained from Nunc.polyamidoamine (PAMAM) dendrimer coating (amino and carboxyl coatings) of these slides was performed by Chimera Biotec.For spotting of reaction solutions and MALDI matrix, the SciFlexArrayer robot from Scienion was used.

oligonucleotides
For the experiments depicted in Fig. 2, we used 2 oligonucleotides (with sequence 5Ј-AATTGAATGGCTCTAG-GAC-3Ј) containing either natural phosphate [expected relative molecular mass (M r ), 5852] or methylphosphonate linkages (expected M r , 5816).For the experiments shown in Fig. 2 of the online Data Supplement, we used 100 nL of a mixture containing 5 M of oligonucleotide containing methylphosphonate linkages and 5 M of an oligonucleotide containing phosphate linkages (5Ј-TGT-GATGGGTGCTCTAGACAAA-3Ј; expected M r , 6815).
All oligonucleotides were diluted in a solution containing 0.33 L of Taq DNA polymerase, 0.21 L of 50 mM MgCl 2 , 0.25 L of SAP (1 U/L), 1.75 L of Tris-Cl (pH 8.0), 0.32 L of TermiPol (5 U/L), and 0.5 L of primer mixture (10 M each).We brought the 2 mixtures up to a final volume of 6 L with doubly distilled water.
We performed single-base variation genotyping with standard procedures consisting of PCR, SAP digestion, and primer extension.pcr DNA was prepared as described recently (20 ).To amplify a stretch of genomic DNA containing single-base variation rs607759, we used the following protocol: 2 L of genomic DNA (5 ng), 5 pmol of the forward (5Ј-AAGCTCTAAAACATGGAAAGGAAA-3Ј), and 5 pmol of the reverse primer (5Ј-TCATGCAATGAAGGGGTCT-TAT-3Ј) were mixed in 20 mM Tris base, 16 mM (NH 4 ) 2 SO 4 , 25 mM KCl, and 2 mM MgCl 2 (pH 8.8) with 100 M dNTPs and 10 U of Taq DNA polymerase.The PCR was performed in a 10-L volume.The reaction mixture was denatured at 95 °C for 4 min, and then thermocycled for 15 s at 95 °C, 30 s at 56 °C, and 60 s at 72 °C, with the cycling repeated 40 times.We performed 24 PCRs for each of the 3 DNAs used.

sap treatment
We added 0.25 L (1 U/L) of SAP and 1.75 L of 50 mM Tris base (pH 8.0) to 3 L of PCR solution and incubated the mixture for 70 min at 37 °C.The SAP was inactivated by heating for 10 min at 90 °C.

primer extension
We added 5 pmol of primer (5Ј-AATTGAATG-GCTCTAGGAC-3Ј), 1.6 U of TermiPol DNA polymerase, 0.5 mM dideoxy-CTP, 0.7 mM dideoxy-GTP, and 0.6 mM dTTP to the PCR in a final volume of 7 L.The initial denaturation step of 2 min at 94 °C was followed by 35 cycles of 10 s at 94 °C, 30 s at 50 °C, and 15 s at 72 °C.

preparation of maldi targets
We used gold PAMAM dendrimer microscope slides with amino or carboxyl coating.For the genotyping experiments, we spotted 4 nL of the crude primer extension reaction solutions containing (theoretically) 20 fmol of oligonucleotide on the amino-modified microscope slides with the SciFlexArrayer robot.The 72 spots were conservatively printed at distances of ϳ5 mm, covering the upper half of the microscope slide (similar to Fig. 1b in the online Data Supplement).We let the droplets dry at room temperature (usually for ϳ15 min).We dipped the slides-with random orientation-generally twice in a tub containing doubly distilled water and dried the slides under a stream of air.We then spotted 4 nL of 120 mmol/L HPA in 15 mmol/L ammonium citrate in deionized water on the dried spots.In solution, ammonium exists as a positively charged ammonium ion, whereas in the gas phase, charge-neutral ammoniac is readily lost, leading to formation of a reduced alkali ion-DNA adduct.For the initial experiments, we applied 200 nL of sample and matrix solution (as shown in Fig. 1b of the online Data Supplement).Conventional purification of primer extension reactions and MS detection with microZipTips TM and OASIS ® HLB is described in Ref. (21 ).

maldi-ms analysis
Spectra were recorded on a Bruker Biflex III time-of-flight mass spectrometer equipped with a Scout MTP TM ion source with delayed extraction.Spectra were recorded in positive-ion linear time-of-flight mode.Typical acceleration potentials were 18 kV.For delayed extraction, the extraction delay was 200 ns.On average, 50 or 10 laser shots per spectrum were accumulated when 200 or 4 nL of matrix was applied, respectively.Spectra were processed and evaluated with the XMass 5.1.2software.For singlebase variation genotype analysis, we used the GenoTools 1.0 software.A description of this software has been published recently (22 ).The settings of the software were similar to the default parameters of the parameter set termed "hard".The lower limit value was additionally increased to 0.3, and the calibration tolerance was set to 15.We searched for signals in the appropriate mass ranges; for example, for the single-base variation genotyping experiment, we examined the mass range m/z 5000 -7000.

Results
This novel MALDI approach for DNA analysis (Fig. 1) allowed efficient binding of nucleic acids to positively charged surfaces (23 ).Purification of the oligonucleotides on microscope slides modified with gold and coated with PAMAM starburst dendrimers (24 ) allowed subsequent MALDI detection of the DNA compared with standard DNA purification methods in the conventional 200-nL scale (Figs. 2 and 3) and was applicable in miniaturized (4 nL) and automated format for single-base variation genotyping by MALDI-MS.
interaction of oligonucleotides and primary amino groups of dendrimer-coated gold microscope slide surfaces With 2 differently modified oligonucleotides [0.2 L (5 pmol/L); 1 containing negatively charged sugar-phosphate linkages and the other synthesized with a charge-neutral methylphosphonate backbone] having the same sequence, we qualitatively demonstrated that oligonucleotides dissolved in reaction solution interact with the amino-modified surface and that their efficient binding is based primarily on ion-ion interaction (Fig. 2).After washing the slides, we detected only the negatively charged molecules, as expected because of their sufficiently tight binding to the positively charged slide surface.Interestingly, we observed efficient binding of DNA to the primary amines in solutions containing salts in concentrations used for PCR and primer extension.Because of low binding efficiency during purification, the charge-neutral oligonucleotides were not detected.In Amino-coated, positively charged gold microscope slides are used as MALDI targets for easy detection of unpurified oligonucleotides.For target preparation, we spot the crude sample (for single-base variation typing) reaction solution containing oligonucleotides on these slides and let the droplets dry.We then wash the slides by dipping them several times in water.After the slides are washed, we spot MALDI matrix at each sample position and detect the oligonucleotides in a conventional MALDI mass spectrometer.TOF, time-of-flight.
contrast, when we used PAMAM dendrimers with negatively charged carboxyl groups at the outer sphere, we could not detect any negatively charged oligonucleotides after the slides were washed, indicating coulombic repulsion.In addition, we did not detect oligonucleotidecontaining methylphosphonate linkages.Purified oligonucleotides containing methylphosphonates were easily detected on our slides (Fig. 2 of the online Data Supplement).After purification on the slides, however, this oligonucleotide signal completely disappeared.
We further studied sample preparation on the modified microscope slides with regard to sensitivity and other mass spectrometric values.To compare sample preparation on slides with standard procedures using 200 nL of HPA matrix and metal targets, we used 200 nL of a solution containing a set of 3 oligonucleotides (M r 5852-7301).For example, using our microscope slide targets with 2 fmol of oligonucleotides, we obtained signal-tonoise (S/N) ratios of ϳ4.With 1 fmol, we were at the detection limit.Mean S/N ratios obtained with our slides were thus similar to those obtained with microZipTips and OASIS HLB purification, with HPA as matrix, and with anchor-chip MALDI targets (21 ).Examples of the detection ranges for our approach with different amounts of oligonucleotides are shown in Fig. 3.
We focused on oligonucleotide concentrations corresponding to single-base variation genotyping experiments as shown below.For example, the mean (SD) S/N ratio for 200 nL of the mixture containing 100 fmol of oligonucleotides was 33 (15), and the full width at halfmaximum value (Fig. 3) was 22 (6).The signal resolution was not always high enough to resolve potential sodium adducts (Figs. 4 and 5).
To investigate potential cross-contamination, we applied 0.2 L of an oligonucleotide at different concentrations (100, 50, and 20 pmol/L, respectively) on the microscope slide.The samples were spotted in close proximity (1.5 mm), and the observed spot sizes with 0.2 L of sample were ϳ4 mm 2 .Washing and matrix preparation were performed as described above.Although we did not observe spreading during sample deposition, we detected MALDI signal peaks up to 1 cm away from the original position for the 100 pmol/L sample and up to 5 mm away for the 50 pmol/L sample.No signals were detected outside the spot position for oligonucleotide sample concentrations of 20 pmol/L or below.The analyte spreading observed during washing made prediction of oligonucleotide loss difficult; however, assuming tight oligonucleotide binding of the amino-modified surface, we roughly estimated that the microarray capacity for primer binding was in the range of 1 pmol/mm 2 , which corresponds to the supplier's specifications based on fluorescence detection.When we used analyte samples with Ͻ1 pmol/mm 2 , we detected only the expected oligonucleotides at the spot position and no signals elsewhere on the slide.
Similar to well-known dried-droplet procedures with 200 nL of HPA matrix, in some cases regions of high sensitivity, so-called sweet spots, were observed with the modified slides.To improve spot-to-spot reproducibility, which is particularly important for automatic measurements, as well as to increase spot densities on the slides, we miniaturized the sample preparation by applying only 4 nL of matrix and DNA sample to our surfaces.As expected, we observed superior reproducibility compared with application of larger volumes and could routinely The same oligonucleotide containing either natural phosphate linkages or artificial methylphosphonate bridges (sequence: 5Ј-AATTGAATGGCTCTAGGAC-3Ј) was used.We performed on-surface purification and MALDI detection, as described, using either amino-or carboxyl-coated slides.Trace A shows the analysis of the natural oligonucleotide purified on amino-coated slides; trace B shows the analysis of the methylphosphonate oligonucleotide purified on the same slide is shown; trace C shows the analysis of the unmodified oligonucleotide purified on carboxyl-coated microscope slides.The spectra were not smoothed or otherwise manipulated.detect 0.2 fmol of oligonucleotides per spot (4 nL), which is far less than the volume required for the single-base variation application shown below.

single-base variation genotyping using aminomodified microscope slides by maldi-ms
Representative results of experiments using the modified microscope slide surfaces to genotype single-base variations are shown in Figs. 4 and 5.The scaling of our sample preparation is adjustable, ranging from a few samples to several thousand samples per slide.We performed 72 single-base variation genotyping experiments (3 DNAs representing all possible genotypes, 24 replications per DNA).We used primer extension reactions for single-base variation allele discrimination (25)(26)(27) and performed PCR, subsequent digestion of residual dNTPs by SAP digestion, and primer extension with 5 pmol of oligonu-cleotide as described in the Materials and Methods section.Typical spectra obtained by automatic measurement without user interference are shown in Fig. 4. The assay quality is shown in Fig. 5, indicated by color codes.When we applied stringent software criteria, particularly with regard to S/N ratio, in automatic analysis (without user interference), we obtained 26 spectra of "high", 31 of "medium", and 12 of "low" reliability.These 69 MALDI spectra were all of sufficient quality for accurate and easy diagnostic allele calling.In our hands, this yield is similar to conventional approaches using an HPA matrix preparation (21 ).Because of poor spectra quality, 3 spectra could not be determined.The genotype analysis is shown on the right side of Fig. 5.Of 69 spectra, 68 could be analyzed correctly.As expected from the preceding experiments, we did not observe cross-contamination.Only 1 genotype in the first row, which was of low reliability Fig. 3. Detection range on gold slides is similar to standard target preparations.
To determine the detection sensitivity of MALDI shown here, we used the oligonucleotides described in the Materials and Methods section.These 3 oligonucleotides can be easily assigned according to their masses.For the top left trace we used 140 fmol per oligonucleotide (which corresponds to total amounts that are typical in conventional MALDI-based single-base variation genotyping experiments); for the top right trace, we used 20 fmol; for the bottom left trace, 2 fmol; and for the bottom right trace, 0.2 fmol.As a result of the uneven height of HPA preparations, the mass calibration can be unstable as variable heights of the matrix preparation can cause a shift of the starting position, affecting the time-of-flight separation of ions in the mass spectrometer.This usually leads to a small mass variation.The hump in the upper right trace potentially derives from high laser fluence and metastable decay.The spectra were not smoothed or otherwise manipulated.
according to the software used, did not correspond with the other 23 homozygous DNA samples placed in this area.A heterozygous genotype is indicated instead of homozygous C. Looking at the problematic spectrum, we observed that the minor peak for the T allele had a S/N that was 4.35-fold lower than the signal for the C allele.The respective S/N values were 37 and 161; in the case of the real heterozygous samples, the absolute S/N values of both signals deviated only slightly from each other (Fig. 4).Thus, the false genotype result was probably attributable to Taq polymerase error and the use of software that does not (yet) sufficiently consider relative signal intensities.The mean error rate per allele for this DNA was 2.1%, and the mean error rate per locus was 4.2%.For the other 2 DNAs, the error rates were all 0%.
In general, human errors in sample preparation and data handling are the main problems in large-scale genotyping (28 ).False allele scoring by software, as seen here, can be another problem in genotyping.Therefore, use of replicates; automated scoring (e.g., by the Genotools software), which probably will steadily improve for specific demands; and subsequent human visual inspection are currently considered the best approach to producing reliable results (28 ).The false genotype detected here was certainly not attributable to the sample preparation procedure.

Discussion
Our procedure omits cumbersome sample purification of nucleic acids before MALDI target preparation by use of chemically modified microscope slides that allow for the first time, to our knowledge, efficient purification of DNA and MALDI detection on the same device.In initial experiments (data not shown), we observed that purified oligonucleotides analysis was possible after preparation with HPA on several microscope slides containing materials such as glass, plastic, and gold.After purification of DNA samples applied to these surfaces, however, only gold surfaces allowed efficient MALDI detection.Substrates such as PAMAM starburst dendrimers were easily coupled to gold surfaces by use of thiol chemistry and provided a high density of positively charged amino functionalities, which are required for binding of oligonucleotides during the purification process.
The microscope slides used in this study are available to any user and can be combined with home-made MALDI adapters (see Fig. 1a in the online Data Supplement) or commercially available products.With increased For variant rs607759 of the opioid receptor, mu 1 (OPRM1) gene, primer 5Ј-AATTGAATGGCTCTAGGAC-3Ј was used in the extension reaction.Respective products of single-base variations were 5Ј-AATTGAATGGCTCTAGGACC-3Ј and 5Ј-AATTGAATGGCTCTAGGACTG-3Ј.To provide realistic data, we used a standard MALDI parameter file for routine measurements.Furthermore, we did not adjust any detection settings for particular oligonucleotides or spots.Three typical spectra of medium quality are illustrated.All products can be easily resolved because they are separated by more than 300 mass units from each other.The top trace shows the analysis of a DNA homozygous for C; the middle trace shows the analysis of heterozygous DNA; the bottom trace shows the analysis of DNA homozygous for T. RP, residual primer from the extension reaction.All products can be easily assigned and resolved from each other.The spectra were not smoothed or otherwise manipulated.The mean (SD) S/N ratio obtained in this experiment for the residual primer signal was 48 (23) and the mean full-width half maximum value was 23 (6).ddC, dideoxycytosine; ddG, dideoxyguanosine; dT, deoxythymidine.We used GenoTools software [a detailed description can be found in Ref. (22 )] to automatically analyze reliability and genotypes in the 72 experiments.The human patient DNAs (trio), numbered 10, 14, and 27, were assembled according to their genotypes in 4 rows.Left, quality score of the experiment; right, the genotype analysis.Green dots correspond to spectra with high, yellow dots to spectra with medium, and red dots to spectra with low reliability.Gray dots indicate results that could not be determined because of very low spectral quality.The genotypes are coded as dark blue (homozygous for C), pale blue (heterozygous), and purple (homozygous for T).We used stringent software criteria so that the reliability of the spectra was rigorously evaluated.However, as can be deduced from the software-independent mass spectrometric values given in legend for Fig. 4 and in the text, 69 spectra were of sufficient quality for further analysis.
sample density, the cost per MS analysis can be reduced proportionally.The price per slide was approximately €40; commercial alternatives such as the chips of Sequenom (http//:www.sequenom.com)cost approximately €200.The slides used here are prototypes, and we expect that the price will decrease substantially with higher production rates.Alternative surface modifications such as those shown here might also allow on-target purification, provided that the density of molecules interacting with DNA is sufficiently high.For example, other dendrimer moieties or alternative structures allowing high binding capacity and alternative interacting groups such as epoxides might be useful.
Our sample preparation method is tailored to applications based on primer extension and MALDI detection.MALDI analysis of single-base variations is a powerful tool for candidate single-base variation genotyping and related diagnostic applications and might become routine in predictive medicine.We recommend triplicate spotting of DNA samples to reduce drop-outs and allele error rates.For analyzing genome-wide DNA variation, our approach is less efficient than much faster fluorescencebased detection on microarrays.In diagnostic laboratories, however, our method can facilitate the repetitive day-to-day and flexible analysis of a limited number of (prognostic) single-base variations in several thousand DNA samples per day.To efficiently apply our slides, the user should carefully select a rapid and flexible arraying robot.The throughput that can be achieved by our approach partly depends on the rinsing time of tips of the specific spotting robot used.The other time-consuming steps of the single-base variation typing procedure shown here are the PCR and the primer extension reactions.New thermocyclers may be able to complete 30 thermocycles in ϳ20 min.The use of these technical devices should make it possible to execute the whole single-base variation typing procedure, including MALDI analysis, in ϳ2 h.We are currently working on a procedure for single-base variation genotyping that reduces the minimum of 3 reaction steps to generate products by primer extension.
In summary, MALDI mass spectrometers have emerged as efficient and accurate instruments in diagnostics (29,30 ) as well as for detection of genetic variation (4 ).Therefore, the implementation of MS is particularly worthwhile in clinical laboratories that routinely perform analysis of proteins and nucleic acids.This work was supported by the European Union (Grant LSHG-CT-2004-503155, MolTools WP2), the German National Genome Research Network (Grant 01GR0414), and the Max-Planck Society.We thank Regine Schwartz, Lajos Nyarsik, Klaus-Dieter Klo ¨ppel, Magdalena Kliem, and Anett Smyra for assistance and help, and Zolta ´n Konthur for critical reading of this manuscript.