Light-directed synthesis of oligonucleotide arrays by MAS technology currently uses nucleotide monomers that have reactive 3′-phosphoramidite groups and 5′-NPPOC photolabile protecting groups (Fig. ) (9
). During array synthesis, the 3′-phosphoramidite attaches to the array surface or growing DNA strand, resulting in the production of arrays with 3′ ends attached to the surface. A potentially more useful configuration would be to synthesize arrays in the 5′→3′ direction, resulting in 5′ ends attached to the array surface and 3′ ends toward the solution. We have synthesized all four 3′-NPPOC-5′-phosphoramidite monomers (Fig. ) that allow for 5′→3′ synthesis, and optimized their use in complex array synthesis.
Structures of monomers used for 3′→5′ and 5′→3′ array synthesis.
UV light dosing studies indicate that 7.5 J/cm2
is sufficient to remove the 3′-NPPOC group from all four bases (data for guanosine shown in Fig. ), which is identical to that required for complete removal of 5′-NPPOC groups (10
). To determine the stepwise chemical yield of 3′→5′ synthesis, we used a method similar to that described previously (15
). Coupling times of 60 s improved overall yields, as compared with 20 s coupling times normally used with the 3′→5′ synthetic monomers (data not shown). This may be due to the fact that the 3′ hydroxyl is a weaker nucleophile and is slightly more hindered than the 5′ hydroxyl. This is consistent with previous studies using DMT based 5′→3′ synthesis (13
). Stepwise yields are slightly lower on average than 3′→5′ synthesis using NPPOC chemistry (10
), even with longer coupling times (Table ). Model experiments with 3′-NPPOC-nucleosides irradiated in solution under similar conditions as in the MAS synthesizer and analyzed for deprotection yield quantitatively by HPLC have given a 1–2% lower yield of the correct deprotected nucleoside than the corresponding 5′-NPPOC-nucleosides (data to be published elsewhere). It is therefore concluded that the deprotection reaction rather than the coupling step determines the cycle yield in the case of the 5′→3′synthesis.
Figure 2 Time course of 3′-NPPOC removal by UV light. Arrays were synthesized with one 3′-NPPOC protected phosphoramidite base layer coupled to the silanized slide. These arrays were then dosed with increasing amounts of UV light to remove the (more ...)
To determine the sensitivity and dynamic range of complex arrays synthesized in the 5′→3′ direction, arrays were designed with six 24mer probes that are complementary to six control oligonucleotides. These control probe sets were represented 20 times across the surface of each array. Six arrays were synthesized, three in the 5′→3′ direction and three in the 3′→5′ direction, and hybridized with the control oligonucleotides covering a range of concentrations from 300 pM to 1 pM. The intensity of the control probes indicate that both 5′→3′ and 3′→5′ arrays have a similar linear dynamic range of at least 2.5 orders of magnitude (Fig. ). Also, the average measured intensity value of the features hybridized to the control oligonucleotide at 1 pM from both types of arrays is >200, roughly four times above the measured background level (<60 for all arrays), indicating a limit of detection in the fM range. The coefficient of variance in the intensities of these control probes across the array surface averaged 8.8% for the 5′→3′ synthesis and 7.0% for the 3′→5′ synthesis.
Figure 3 Dynamic range and sensitivity. Six 24mer probes sequences were synthesized at 20 locations across the array surface. Two separate arrays were synthesized, one in the 5′→3′ direction, one in the 3′→5′ direction. (more ...)
Intra- and inter-array variability were determined by hybridization of labeled mouse spleen cRNA (see Materials and Methods) to arrays containing twenty 24mer probe pairs (perfect match and mismatch) per gene for 950 mouse genes. These probes were randomly distributed across an array quadrant, and each quadrant was repeated four times per array. The resulting array contained 152 000 total probes and 38 000 unique probes, not counting controls. Three 5′→3′ synthesized arrays and three 3′→5′ synthesized arrays were hybridized, and the average difference values (see Materials and Methods) were calculated for all genes on each array. Image details from a 5′→3′ and a 3′→5′ array are shown in Figure . The average difference values for 330 genes were at or below the array background (<60 for all arrays), and thus were dropped from further analysis. The remaining 620 average difference values were used to determine concordance between arrays. The intra-array variability for 5′→3′ synthesis was determined by comparing the average difference values for every possible intra-array quadrant combination. An example of the log transformed data for one comparison is shown in Figure A. The average R2 value for these comparisons indicates good concordance between array quadrants (Table ). R2 values from both the linear and log transformed data are shown. While the linear R2 values are often reported for array data comparisons, the R2 values from log transformed data are a more sensitive measure of array variability. The inter-array variability was determined by comparing all possible combinations of one quadrant from one array with the corresponding quadrants from the other two arrays, for both the 5′→3′ synthesis and the 3′→5′ synthesis. These comparisons (examples in Figs C and B) indicate that the level of concordance across arrays is high, and is similar in both synthesis directions (Table ). Finally, when data from arrays synthesized in the 5′→3′ direction were directly compared with data from arrays with 3′→5 synthesis, the level of concordance diminished significantly when compared to concordance within one synthesis orientation (Fig. D and Table ). This may be explained by the different synthesis yields of the two chemistries that may slightly favor or disfavor synthesis of specific sequences, and potential alterations in probe secondary structure caused by the direction of attachment.
Figure 4 Detail of 5′→3′ and 3′→5′ array images. Arrays were designed with twenty 24mer probe pairs (perfect match and corresponding mismatch controls) per gene for 950 mouse genes. Probes were randomly distributed (more ...)
Figure 5 Intra-array and inter-array reproducibility. Three arrays were synthesized in the 5′→3′ direction and three were synthesized in the 3′→5′ direction. Arrays were hybridized with labeled mouse spleen cRNA. (more ...)
To demonstrate the ability of enzymes to access the 3′ ends of synthesized oligonucleotides, primer extension and ligation reactions were employed. Arrays were synthesized with two 18mer oligonucleotide sequences that were identical except for their 3′ nucleotide, designated oligo A and B (Figs A and A). For the extension reactions, these arrays were hybridized to a 30mer oligonucleotide designated template A, which was complementary to the 3′ nucleotide of oligo A, but contained a mismatch with the 3′ nucleotide of oligo B. This template extends 12 bases beyond the 3′ end of the array oligonucleotides. After hybridization and stringent washing to eliminate non-specific hybridization, the array primers were extended with the Klenow fragment of DNA polymerase I (3′→5′ exo–). The extension reaction incorporated biotinylated dUTP, which was subsequently stained with streptavidin-Cy3. Two arrays were extended, the first array with template A and the second with no template to determine the efficiency of template-independent extension. Extension from oligo A resulted in a signal that was approximately four times that of oligo B and the no template controls (Fig. B and C). The oligo B and no template results indicate that a low level of template-independent extension is occurring, most likely from interactions of the 3′ end of a probe with either itself or neighboring probes. The use of thermostable polymerases and higher reaction temperatures will likely eliminate these structures and reduce non-specific signal, increasing mismatch discrimination.
Figure 6 Sequence-specific primer extension reactions. Arrays were synthesized with alternating rows of oligonucleotide sequences designated oligo A (5′-AGG TCA TTA CAG CGA GAG-3′) and oligo B (5′-AGG TCA TTA CAG CGA GAC-3′), which (more ...)
Figure 7 Sequence-specific ligation reactions. (A) Ligation scheme showing template A hybridized to the array oligos A and B and ligation of the labeled ligation oligo (5′-phosphate-GATTATAGGTCA-Cy3-3′) to the end of the array oligos with DNA (more ...)
For the ligation reaction, template A was hybridized to the array, along with the ligation oligo (Fig. A). This strategy should result in efficient ligation of the Cy3-labeled ligation oligo to oligo A on the array surface, while ligation to oligo B should be reduced in efficiency, due to the 3′ mismatch with template A. Ligation to the perfect match oligo (oligo A) resulted in approximately five times the signal of ligation to the mismatch oligo (oligo B) (Fig. B and C). An identical reaction performed without ligase produced no signal, indicating that the signal resulting from the ligation reaction was indeed covalently bound, and not due to hybridization. These results indicate that the 3′ ends of the array are available for sequence-specific primer extension and ligation reactions.