DNA in sperm undergoes an extreme compaction to almost crystalline packing levels. To produce this dense packing, DNA is dramatically reorganized in minutes by protamine proteins. Protamines are positively charged proteins that coat negatively charged DNA and fold it into a series of toroids. The exact mechanism for forming these ∼50-kbp toroids is unknown. Our goal is to study toroid formation by starting at the “bottom” with the folding of short lengths of DNA that form loops and working “up” to more folded structures that occur on longer length scales. We previously measured the folding of 200–300 bp of DNA into a loop. Here, we look at folding of intermediate DNA lengths (L = 639–3003 bp) that are 2–10 loops long. We observe two folded structures besides loops that we hypothesize are early intermediates in the toroid formation pathway. At low protamine concentrations (∼0.2 μM), we see that the DNA folds into flowers (structures with multiple loops that are positioned so they look like the petals of a flower). Folding at these concentrations condenses the DNA to 25% of its original length, takes seconds, and is made up of many small bending steps. At higher protamine concentrations (≥2 μM), we observe a second folded structure—the loop stack—where loops are stacked vertically one on top of another. These results lead us to propose a two-step process for folding at this length scale: 1) protamine binds to DNA, bending it into loops and flowers, and 2) flowers collapse into loop stacks. These results highlight how protamine uses a bind-and-bend mechanism to rapidly fold DNA, which may be why protamine can fold the entire sperm genome in minutes.

DNA construct preparation and protamine

DNA lengths of 639, 1170, and 3003 bp were amplified from Lambda DNA using standard PCR procedures (41,51). We used an LA Taq DNA polymerase (TaKaRa Bio RR002) and custom primers (Integrated DNA Technologies). Forward primers were tagged with biotin, and reverse primers were tagged with digoxigenin so that the final DNA product could bind to the streptavidin-coated particle and an anti-digoxigenin-coated slide, respectively. GC content for each construct was 52% (639 bp), 33% (1170 bp), and 44% (3003 bp). We have previously seen that loops are more likely to form in regions of lower GC content (52). Other studies (53) have seen that the presence of static curvature in the DNA created by phased tracts of poly(A) regions has led to loops within toroids that have reduced diameters.

After amplification, we checked the purity of our product using a gel purification procedure. We first performed gel electrophoresis using orange loading dye (New England Biolabs, B7022S) and standard protocols for short DNA molecules (54). Samples were then extracted from the gel using a commercial kit (Qiagen QIAquick Gel Extraction Kit, no. 28704). Finally, we checked the concentration and purity of the PCR product (Thermo Fisher, NanoDrop Lite) and discarded samples with an A260/A280 ratio below 1.7.

AFM sample preparation

We created AFM samples by affixing mica slides (Ted Pella, grade V1, 10-mm diameter, ruby muscovite) to magnetic disks using quick-dry epoxy. For control samples, we prepared 20 μL of a solution of 1.0 ng/μL DNA and 2.0 mM magnesium acetate in a DNA LoBind microcentrifuge tube (Eppendorf, no. 022431005) to minimize adhesion to the tube. We pipetted this solution onto the surface of the mica, waited up to 30 s, washed with 1 mL of deionized water, and dried using nitrogen. Samples were stored in a desiccator.

The DNA is known to equilibrate on the mica surface (55) and is not kinetically trapped in 3D. To make sure the DNA has equilibrated on the surface, we check measurements of contour length and height for all molecules, even DNA flowers, to make sure they match the nominal values. Measurements of persistence length for molecules not exposed to protamine also match the nominal values (56). In addition, previous studies have shown that DNA attached to a mica surface can still fold into a toroid when protamine is added to the buffer (14).

As a control, we also prepared samples with just protamine (10 μM) and magnesium acetate (Fig. S3). We did not see any clumps of protamine on the surface, indicating that protamine is likely to be monodispersed.

For all samples with protamine, we found that adding protamine (even small concentrations of 0.2 μM) to 1.0 ng/μL of DNA resulted in DNA aggregates. This is because protamine creates DNA-DNA interactions. To combat this, we reduced the concentration of DNA in solution and performed multiple depositions. Specifically, we prepared 20-μL solutions of 0.2 ng/μL DNA, 2.0 mM magnesium acetate, and the appropriate concentration of protamine. Then, we flowed 1 mL of deionized water to rinse the sample, vacuumed off the liquid, and dried the surface with nitrogen. We repeated this process until there were two to five depositions (with five depositions working the best).

AFM data collection

AFM samples were imaged using either a Dimension 3000 AFM (Digital Instruments) with a Nanoscope IIIa controller or an MFP-3D AFM (Asylum), depending on core facility availability. We used the PPP-XYNCSTR-model cantilever (Nanosensors, resonant frequency = 150 kHz, force constant = 7.4 N/m, length = 150 μm, tip radius <7 nm). Images were taken in air. There was some DNA shrinkage due to drying in air, but most molecules (70%) were still within 20% of the expected contour length. DNA strands previously imaged with this system show heights of 0.55 ± 0.07 nm and radii of 3.9 ± 0.6 nm, close to the expected radius of 3 nm (57). We used a scan rate of 1–4 Hz. Most images are 2 × 2 μm and 512 × 512 pixels, but some are 256 × 256 pixels.

AFM analysis

AFM images were processed in Gwyddion (58). We corrected images by 1) aligning rows using a fifth-degree polynomial, 2) using the fast Fourier transform (FFT) filter to remove high-frequency oscillations, and 3) removing scars. Using the corrected images, we identified DNA singlet molecules (numbers of molecules at each DNA length and protamine concentration are listed in Table S1). Singlets had to be lying flat on the surface with at least one pixel of separation between other molecules. We traced a contour onto each DNA molecule using Gwyddion. The contour lengths were required to be within 20% of the nominal length. We cropped square images of valid singlets and saved them as JPEGs.