Accurate length determination of DNA molecules visualized by atomic force microscopy: evidence for a partial B- to...

Achieving the most correct estimate of the contour length of digitized DNA molecules is a key aspect of the microscopic analysis of nucleic acids by either electron microscopy (EM) or atomic force microscopy (AFM). Six different methods, that are mathematically not too complex and suited for common, practical use, have been tested here using simulated polymers in two dimensions and real DNA molecules (564, 1054, 2049 and 4297 bp long) imaged in air by AFM. The main result is that the frequently used Freeman estimator (L(F) = n(e) + square root 2n(o)) overestimates the real contour length of the polymers by about 4%. More accurate estimates are obtained with the Kulpa estimator (L(K) = 0.948n(e) + 1.343n(o)) or with the corner count estimator (L(C) = 0.980)n(e) + 1.406n(o) - 0.091n(c)). In the range of the DNA sizes and magnifications we have considered, however, the best results are obtained with an ad hoc developed routine that smoothes the DNA trace by a polynomial fitting of degree 3 over a moving window of 5 points. Under these conditions, the difference between the measured and the real contour length of the molecules is less than 0.4%. The accuracy of this procedure allowed us to reveal a discrete, size-dependent, shortening of DNA molecules deposited onto mica under low salt conditions and imaged in air by AFM. Awareness of this structural alteration, that can be attributed to a partial transition from B- to A-form DNA, may lead to a more correct interpretation of DNA molecules or protein-DNA complexes imaged by AFM.