DNA protection by stress-induced biocrystallization

1 July 1999

Beitar 400, 83 - 85 (1999) © Tel Hay Publishers Ltd.

DNA protection by stress-induced biocrystallization

SHARON G. WOLF*, DAPHNA FRENKIEL*, TALMON ARAD ⁺, STEVEN E. FINKEL^!, ROBERTO KOLTER^! & ABRAHAM MINSKY*

Departments of * Organic Chemistry and ⁺ Structural Biology, The Weizmann Institute of Science, Rehovot 76100, Israel
^! Department of Microbiology and Molecular Genetics, Harvard Medical School, 200 Longwood Avenue, Boston, Massachusetts 02115, USA

The crystalline state is considered to be incompatible with life. However, in living systems exposed to severe environmental assaults, the sequestration of vital macromolecules in intracellular crystalline assemblies may provide an efficient means for protection. Here we report a generic defence strategy found in Escherichia coli, involving co-crystallization of its DNA with the stress-induced protein Dps^1,2. We show that when purified Dps and DNA interact, extremely stable crystals form almost instantaneously, within which DNA is sequestered and effectively protected against varied assaults. Crystalline structures with similar lattice spacings are formed in E. coli in which Dps is slightly over expressed, as well as in starved wild-type bacteria. Hence, DNA-Dps co-crystallization is proposed to represent a binding mode that provides wide-range protection of DNA by sequestration. The rapid induction and large-scale production of Dps in response to stress, as well as the presence of Dps homologues in many distantly related bacteria, indicate that DNA protection by biocrystallization may be crucial and widespread in prokaryotes.

Under conditions of either oxidative or nutritional stress, E. coli produces high levels of the nonspecific DNA-binding protein Dps^2,3,4, which effectively protects DNA against oxidative agents both in vitro and in vivo⁴. Dps spontaneously forms spherical dodecamers that are homologous to assemblies formed by ferritins⁵. It has consequently been proposed that the protective activity of Dps stems from its putative ability to sequester iron ions⁵. While effective in attenuating formation of free radicals by the Fenton reaction, a defence strategy based solely on ion sequestration cannot provide the all-encompassing DNA protection that is required--and observed--during prolonged starvation^1,6. Indeed, Dps protects DNA against various nucleases² and oxidative agents⁴, thus indicating a general Dps-dependent defence mechanism.

Details of this mechanism were derived from electron microscopy studies of Dps and Dps-DNA complexes (Fig. 1). In the absence of DNA, Dps forms spherical dodecameric particles of 90 Å in diameter², which appear ring-like in projection. Overnight incubation of Dps results in two-dimensional crystals; projection images of these crystals reveal hexagonal packing with a spacing of 78 1 Å, corresponding to a particle diameter of 90 Å (Fig. 1b). An identical value was obtained from the Dps X-ray crystal structure⁵. DNA molecules have a marked effect on Dps organization (Fig. 1c-f). Immediately after the addition of DNA, extensive aggregation of Dps dodecamers occurs that is followed, seconds later, by the formation of numerous crystals. This rapid crystallization is produced by closed supercoiled plasmids, linear double-stranded DNA and single-stranded RNA molecules, leading in all cases to multi-layered plate-like crystals of similar morphology. No crystals are formed, however, in the presence of negatively charged polypeptides such as polyglutamate.

Figure 1 Electron microscopy of Dps and Dps-DNA complexes. Full legend

High resolution image and legend (548k)

Although it greatly accelerates crystallization and promotes a three-dimensional morphology, the DNA does not affect the in-plane lattice spacing of the crystals. Irrespective of the DNA structure, this remains at 78

2 Å, similar to that of two-dimensional Dps crystals obtained in the absence of DNA. To assess the effects of the DNA, differential interference contrast (DIC) and fluorescence microscopy studies of DNA-Dps mixtures were carried out. Mixtures of Dps and fluorescently labelled DNA exhibit fluorescent blobs which exclusively coincide with the crystals observed in the DIC mode; crystals obtained with non-labelled DNA are not fluorescent (data not shown). These findings indicate that DNA is incorporated within the crystals and, in conjunction with the electron microscopy results, imply a remarkably efficient nucleic acid-Dps co-crystallization process. The observation that nucleic acids in their various forms do not alter the crystal spacing is consistent with a model where DNA and Dps form stacked alternating layers, within which DNA is effectively sequestered and protected. This model is reminiscent of DNA-liposome assemblies, where DNA induces a lamellar packing of the lipids and is stacked in between these layers⁷.

Complexes between DNA and nonspecific DNA-binding proteins such as RecA⁸, H-NS⁹, the sporal SASP¹⁰, sperm protamine¹¹ and histone H1 (ref. 12) contain only protein-coated DNA molecules or amorphous aggregates. Thus, the fast DNA-Dps co-crystallization represents an unprecedented binding and protection mode. But is this mode physiologically relevant?

Electron microscopy studies were conducted on E. coli that overexpress Dps⁴. Over-production of Dps to levels that are fourfold higher than those accumulated in starved wild-type bacteria results in the appearance of intracellular crystalline structures (Fig. 2a-d). The flat in vitro crystals adopt a favoured orientation on the electron microscope grid and hence exhibit only intra-layer spacings. In contrast, information on both intra- and inter-layer spacings is provided by the in situ crystalline structures, as they are embedded within randomly sliced sections. Indeed, some of these crystals have a layered morphology, whereas others display square lattices that result from views normal to the layer planes. Although they are smaller and less frequent than the crystals formed following Dps over-production, crystalline assemblies are also detected in starved wild-type E. coli (Fig. 2e), thus indicating that ordered assemblies are not merely a consequence of Dps over-expression. The similarity between the in vitro and in vivo DNA-Dps crystals indicates that the ordered intracellular assemblies are Dps-DNA complexes. Several lines of evidence support this idea. First, dps knockout bacteria do not show crystalline assemblies upon starvation (Fig. 2f). Thus, the presence and intracellular amounts of Dps correlate with the appearance, size and frequency of the ordered in vivo assemblies. Second, when DNA is isolated from bacteria over-producing Dps, Dps is found to be the predominant protein bound to the DNA. Finally, in actively growing bacteria, chromatin is detected as amorphous spaces¹³. The absence of such regions in bacteria that contain crystalline assemblies implies a co-localization of DNA and the crystals.

Figure 2 In situ DNA-Dps assemblies. Full legend

High resolution image and legend (452k)

Both wild-type and over-producer strains form ordered structures only when starved, when Dps is one of the few proteins still being expressed². This observation points to a causal link between starvation and the crystallization process. Because after prolonged starvation the cellular content of most proteins is reduced¹⁴, any interference that might be imposed by competing DNA-binding proteins upon the ordered DNA-Dps assembly is substantially attenuated.

The stress-related intracellular assembly and the rapid DNA-Dps co-crystallization observed in vitro imply a drive for ordered packing that is rarely encountered in living systems or revealed by biopolymers. One intriguing example is the crystalline organization of ferritin in the bacterium Helicobacter pylori¹⁵. This organization has been proposed to reduce the risk of iron poisoning by sequestering intracellular iron. The observation that both ferritin and Dps form intracellular crystalline assemblies is provocative in light of their structural homology^5,16,17 and their abundance in diverse prokaryotes^{17,18,19,20,21}. These findings may indicate an evolutionary pressure to promote biocrystallization processes whose protective activity is exerted through a fast sequestration of damaging agents and their targets.

Methods
Dps purification. We purified Dps from a Dps over-producer E. coli strain (SF101) that harbours plasmids in which the dps gene was cloned under the control of the arabinose-inducible P_BAD promoter⁴. Cells were grown at 37 °C in LB medium to an absorbance at 560 nm of 1 and induced with 0.3% arabinose. Purification was performed as described², with several modifications. Specifically, we replaced the Sephadex G-100 and Sepharose 6B columns with a DEAE-cellulose ion-exchange column. The final purification step remained DNA-cellulose affinity chromatography.

DIC and fluorescence microscopy. Dps-DNA complexes (78 µg ml^-1 Dps, linearized pBluescript DNA in 10 mM NaCl, 10 mM Tris) were prepared at a Dps/DNA w/w ratio of 1/5. DNA was labelled with YOYO fluorophore before complex formation at a 1/1 molar ratio of dye to base pair (bp). Images in both DIC and fluorescence modes were taken on an Axiovert 35 mm Zeiss microscope with an enhanced Gatan CCD camera.

In vitro electron microscopy. We incubated samples containing Dps and DNA (closed circular supercoiled or linearized pBluescript, 2,958 bp) at concentrations indicated in the caption to Fig. 1 at room temperature in 20 mM NaCl, 10 mM Tris (pH 7.2) on carbon-coated grids. Specimens were stained with 1% uranyl acetate and probed on a Philips CM12 electron microscope with a Tietz CCD, operating at 100 kV.

Thin-section electron microscopy. Wild-type ZK126 E. coli cells² were grown in LB medium and starved for 48 h. E. coli harbouring the pBAD-dps plasmid that carries an arabinose-inducible dps gene⁴ were treated with 0.02% arabinose (resulting in Dps accumulation to a level that is 3- to 4-fold higher than that obtained in starved wild-type bacteria). Following induction, we incubated the cells in LB medium at 37 °C for 48 h. Bacteria were fixed by ultra-fast freezing in liquid ethane and subsequently cryosubstituted as described¹³. In control experiments, chemical fixation was performed according to the RK-U procedure²². Samples were embedded in epon; thin sections were stained with 1% uranyl acetate and examined on a Philips CM12 electron microscope.

Received 15 March;accepted 5 May 1999.

References

Kolter, R., Siegele, D. A. & Tormo, A. The stationary phase of bacterial life cycle. Annu. Rev. Microbiol. 47, 855-874 (1993). Links