The Role of Capsid Maturation on Adenovirus Priming for Sequential Uncoating*

Background: Adenovirus proteolytic maturation is required for correct uncoating in the cell. Results: Maturation makes the virion metastable and facilitates penton and peripheral core protein release, as well as cooperative genome ejection. Conclusion: Precursor proteins act as scaffolds favoring assembly. Maturation primes adenovirus for uncoating. Significance: Identifying the molecular determinants of virus stability and uncoating is key to understanding the infectious cycle. Adenovirus assembly concludes with proteolytic processing of several capsid and core proteins. Immature virions containing precursor proteins lack infectivity because they cannot properly uncoat, becoming trapped in early endosomes. Structural studies have shown that precursors increase the network of interactions maintaining virion integrity. Using different biophysical techniques to analyze capsid disruption in vitro, we show that immature virions are more stable than the mature ones under a variety of stress conditions and that maturation primes adenovirus for highly cooperative DNA release. Cryoelectron tomography reveals that under mildly acidic conditions mimicking the early endosome, mature virions release pentons and peripheral core contents. At higher stress levels, both mature and immature capsids crack open. The virus core is completely released from cracked capsids in mature virions, but it remains connected to shell fragments in the immature particle. The extra stability of immature adenovirus does not equate with greater rigidity, because in nanoindentation assays immature virions exhibit greater elasticity than the mature particles. Our results have implications for the role of proteolytic maturation in adenovirus assembly and uncoating. Precursor proteins favor assembly by establishing stable interactions with the appropriate curvature and preventing premature ejection of contents by tightly sealing the capsid vertices. Upon maturation, core organization is looser, particularly at the periphery, and interactions preserving capsid curvature are weakened. The capsid becomes brittle, and pentons are more easily released. Based on these results, we hypothesize that changes in core compaction during maturation may increase capsid internal pressure to trigger proper uncoating of adenovirus.

Adenovirus assembly concludes with proteolytic processing of several capsid and core proteins. Immature virions containing precursor proteins lack infectivity because they cannot properly uncoat, becoming trapped in early endosomes. Structural studies have shown that precursors increase the network of interactions maintaining virion integrity. Using different biophysical techniques to analyze capsid disruption in vitro, we show that immature virions are more stable than the mature ones under a variety of stress conditions and that maturation primes adenovirus for highly cooperative DNA release. Cryoelectron tomography reveals that under mildly acidic conditions mimicking the early endosome, mature virions release pentons and peripheral core contents. At higher stress levels, both mature and immature capsids crack open. The virus core is completely released from cracked capsids in mature virions, but it remains connected to shell fragments in the immature particle. The extra stability of immature adenovirus does not equate with greater rigidity, because in nanoindentation assays immature virions exhibit greater elasticity than the mature particles. Our results have implications for the role of proteolytic maturation in adenovirus assembly and uncoating. Precursor proteins favor assembly by establishing stable interactions with the appropri-ate curvature and preventing premature ejection of contents by tightly sealing the capsid vertices. Upon maturation, core organization is looser, particularly at the periphery, and interactions preserving capsid curvature are weakened. The capsid becomes brittle, and pentons are more easily released. Based on these results, we hypothesize that changes in core compaction during maturation may increase capsid internal pressure to trigger proper uncoating of adenovirus.
Adenovirus is a mild pathogen for humans but can become clinically relevant in immunocompromised patients, and it is widely examined as a therapeutic vector (1,2). The large (ϳ950 Å) nonenveloped, pseudo T ϭ 25 icosahedral capsid is assembled from at least 11 different types of proteins. The general icosahedral architecture can be described as two different systems of tiles. Nine trimers of the major coat protein, hexon, form the central plate of each facet, known as the Group of Nine (GON). 5 The five peripentonal hexon trimers, together with the penton base, form the second tile system, known as Group of Six (GOS) (3). Trimeric fibers protrude from each vertex.
Minor coat proteins IIIa, VI, VIII, and IX are required for correct capsid assembly and occupy specific positions in the capsid (3). Polypeptide IX has an extended structure and forms a sort of hairnet on the outer surface of the virion, keeping together the hexon trimers in each GON and binding GONs to GONs across the icosahedral edges. Each GON is further stabilized by copies of polypeptide VIII located around the icosahedral 3-fold symmetry axis inside the capsid. Also on the inner capsid surface, the N-terminal domain of polypeptide IIIa mediates the interaction between penton base and the peripentonal hexons to keep each GOS together. Finally, IIIa and VIII cooperate to bind each GOS to its five surrounding GONs. The remaining minor capsid protein, polypeptide VI, could not be traced in this study, but it has been assigned to density within the internal cavity of each hexon trimer (4,5). The adenovirus core contains a single copy of the dsDNA genome and over 25 MDa of protein, including four DNA-binding viral proteins: polypeptides V, VII, , and the terminal protein. A few copies of the adenovirus protease (AVP) are also packaged (6). The disposition of DNA and proteins in the core is unclear.
Adenovirus cell entry has been extensively studied using group C human adenoviruses; these are most commonly used for vector development, in particular the highly homologous types 5 (HAdV-5) and 2 (HAdV-2) (7,8). Adenovirus uncoating in the cell proceeds in a stepwise manner (9). The virus starts to disassemble at the plasma membrane, where upon binding to its receptor some fibers are released (10), and the penton base undergoes a conformational change that might result in weakening its interactions with the rest of the capsid (11). The virus is then internalized, and disassembly continues in the early endosome with release of some internal components, such as minor coat proteins IIIa, VI, VIII, and core polypeptide V (9,12). This is a crucial step for infection, as polypeptide VI plays the key role of altering the endosomal membrane to facilitate virus escape to the cytosol (13). The partially disassembled virion does not proceed to later stages of the endocytic pathway but escapes the early endosome (14). At this point there must be a certain degree of dsDNA exposure to initiate inflammatory responses (15), but the virion is still stable enough to survive transport along microtubules to the nuclear pore (16,17). There, the final phase of uncoating takes place. Upon complete dismantling of the weakened capsid, the virus DNA, together with major core protein VII, enters the nucleus (18 -21).
Correct adenovirus uncoating is tightly linked to maturation. Cleavage of several structural proteins by AVP is required to yield the mature infectious virion (22). Immature virions contain the precursor versions of coat proteins IIIa, VI, and VIII and of core proteins VII, , and terminal protein. An experimental model for immature adenovirus is the HAdV-2 ts1 mutant (23). When grown at the nonpermissive temperature of 39°C, HAdV-2 ts1 does not package AVP (24) and produces capsids containing the unprocessed protein precursors. Viral genome packaging is unimpaired, but the virus is not infectious. It has been shown that the defect in infectivity is linked to a defect in uncoating. Immature ts1 attaches to the host cell and follows the same internalization process as the wild-type (WT) virus but fails to escape the endosome and is recycled to the membrane or degraded in lysosomes (14,25).
Cryo-EM structural studies showed that in the immature virus, the presence of uncleaved precursors reinforces the network of interactions holding the virion together at three levels. First, the virus core is more compact and stable due to the condensing action of unprocessed polypeptides pVII and pre-; second, a "molecular stitch" formed by pIIIa and pVIII underpins the boundary between GOS and GONs (5,26); and third, there is a more ordered interaction between pVI and the inner cavity of each hexon in the capsid (4,5). To further investigate how the presence of uncleaved precursors translates into a deficient uncoating phenotype, we have compared the in vitro stability and disassembly of ts1 and WT virions under different types of stress: thermal, chemical, or mechanical. We report the differences between mature and immature virions and discuss their implications for assembly and uncoating.

EXPERIMENTAL PROCEDURES
Virus Production and Purification-We used as wild-type control for mature virions the E1-deleted HAdV-5 variant Ad5GL (27), propagated in HEK293 cells. Ad5GL is completely WT for all structural polypeptides. Immature virus was obtained by propagating the HAdV-2 ts1 mutant in HeLa cells at 39.5°C, as described (5). Particles were purified by equilibrium centrifugation in CsCl gradients, desalted on a Bio-Rad 10 DC column, and stored in 20 mM Hepes, pH 7.8, 150 mM NaCl plus 10% glycerol at Ϫ70°C. Virus titers were 5 ϫ 10 12 virus particles (vp)/ml for Ad5GL, and 1 ϫ 10 13 vp/ml for ts1.
Differential Scanning Calorimetry-DSC measurements were performed using a Microcal VP-DSC instrument (Microcal, Inc., Northampton, MA) at a heating rate of 60°C/h and under an extra constant pressure of 2 atm to prevent degassing during the scan. The VP-viewer package was used for data acquisition. Samples of WT and ts1 at ϳ5-7 ϫ 10 12 vp/ml were dialyzed against 8 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , 150 mM NaCl, and 0.1 mM EDTA, pH 7.4. The reference cell was loaded with buffer from the last dialysis, which was also used to run the buffer-buffer baseline. Apparent excess heat capacity curves were obtained after subtraction of the buffer-buffer baseline from the experimental heat capacity profiles and division by hexon molar concentration.
The Origin-DSC software package (7.0 version) was used for data analysis. The excess heat capacity profiles were theoretically analyzed in terms of independent, irreversible two-state transitions (N i 3 D i ) according to Equation 1 (28), where ⌬Cp ex (T) is the excess heat capacity value at temperature T; T mi is the temperature of the maximum for transition i; ⌬H i app and ⌬E i app are the apparent calorimetric enthalpy change and the activation energy for the same transition, and R is the gas constant.
Fluorescence Spectroscopy-Extrinsic fluorescence spectra were obtained from virus preparations subject to different kinds of stress: high temperature, pyridine, or acidic pH. Mature and immature virus samples (5 ϫ 10 10 vp/ml) were incubated in different buffers, depending on the experiment. Thermal disruption experiments were carried out in 8 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , 150 mM NaCl, and 0.1 mM EDTA, pH 7.4. For pH disruption experiments, the sample was diluted and subsequently dialyzed overnight at 4°C against buffers at the appropriate values of pH: HCl/KCl in the 1-2.5 pH range, citric acid/sodium citrate at pH 3-6, and Na 2 HPO 4 /KH 2 PO 4 at pH range 6.5-7.5, always with 150 mM NaCl and 0.1 mM EDTA.
Fluorescence emission spectra were obtained employing a Hitachi Model F-2500 FL spectrophotometer equipped with a Peltier temperature control device. 1 mM propidium iodide (PI) (Molecular Probes) was added to the sample that was then allowed to equilibrate for 5 min before data acquisition. Sample volumes of 0.150 ml were used in sealed quartz cuvettes. The sample was excited at 535 nm, and fluorescence emission was monitored from 580 to 700 nm, using excitation and emission slit widths of 8 nm. Raw spectra were corrected by subtraction of the PI spectrum at each tested condition. The ratio of PI fluorescence to the initial emission (I/I 0 ) at the maximum position ( max ϭ 607 nm) Ϯ S.E. for three independent experiments was plotted as a function of stress condition values. Fluorescence intensity changes were fitted to a sum of sigmoids using the Origin software package, according to Equation 2, where A i and B i are the lower and upper platform values for each sigmoid (notice that starting from the second sigmoid, the lower platform A i is forced to coincide with the previous upper platform B i Ϫ 1 ); C i is the transition midpoint; and D i is the slope. The number of sigmoid curves in the summatory (up to n ϭ 3) was the minimum necessary to fit the experimental curves, based on best R 2 values. The fluorescence thermal-disruption curve was also fitted in terms of the model used to analyze the calorimetric curves using Equation 3, where I F is the final value of I/I 0 , and f i is the relative contribution of transition i to the total variation of PI fluorescence (29).
Negative Staining Electron Microscopy-For imaging of disassembly products, samples treated as described for fluorescence spectroscopy were adsorbed, at a concentration of 5 ϫ 10 11 vp/ml, onto glow-discharged, collodion/carbon-coated EM grids, negatively stained with 2% uranyl acetate, and observed in a Jeol 1200EX-II transmission electron microscope. Percentages of intact, damaged, and collapsed capsids were quantified by counting the number of particles of each type per micrograph. Error bars refer to differences between different micrographs.
Cryoelectron Tomography-Viral samples subject to different kinds of stress were mixed at a concentration of 1 ϫ 10 12 vp/ml with 10-nm colloidal gold particles (AURION, Wageningen, The Netherlands) and vitrified as described (30). A Tecnai-12 electron microscope (FEI, Hillsboro, OR) operating at 120 kV with a LaB6 source and equipped with an energy filter (GIF 2002; Gatan, Pleasanton, CA) was used to collect tilt series at ϳ4-m underfocus, covering the range Ϯ70°in 2°increments. The data were acquired using the SerialEM package (31), for a total dose of ϳ70 -80 electrons/Å 2 per series. Images were recorded on a 2048 ϫ 2048-pixel CCD camera (Gatan, Pleasanton, CA) at ϫ38,500 magnification (7.8 Å/pixel). Data were pre-processed and aligned using the IMOD software package (32), with gold particles as fiducial markers.
The final aligned tilt series were normalized and reconstructed using the simultaneous iterative reconstruction technique implemented in Tomo3D (33). Individual virus particles were extracted from tomograms using IMOD. Tomograms were denoised by 100 iterations of anisotropic nonlinear diffusion (34). The in-plane resolution of the full tomograms was in the 43-55-Å range, as estimated using the Bsoft program TOMORES (35).
Dihedral angles were measured from central slices of single virion tomograms using the angle measurement tool in Adobe Photoshop. A virion was considered to conserve fibers if at least one fiber was clearly visible in the denoised tomogram. The presence or absence of pentons was determined manually by observation of single virion tomogram slices after alignment with an icosahedral three-dimensional map as a reference, and comparison with model maps of full and pentonless virions. Radial profiles were calculated from average virus tomograms calculated after aligning each individual map with an icosahedral reference. Icosahedral symmetry was considered for alignment but not imposed in the final averaged map or in the individual aligned maps, except for supplemental Fig. S6. Alignment was carried out using maximum likelihood procedures for tomography as implemented in XMIPP (36,37).
Nanoindentation Assays-Stocks of WT and ts1 virus in HBS buffer (20 mM Hepes, 150 mM NaCl, pH 7.8) were diluted in a solution of NiCl 2 in HBS to obtain a final solution with 5 mM of Ni 2ϩ , and adsorbed onto freshly cleaved Muscovite mica (V-1 quality). A drop of 20 l of the final solution, containing virus particles at concentrations between 1.5 and 2 ϫ 10 12 vp/ml, was deposited on the substrate and incubated for 30 min at 4°C before washing with 5 mM NiCl 2 in HBS. The AFM tip was prewetted with 30 l of the same buffer. The mica was placed on the AFM sample holder and immersed in 500 l of buffer. The AFM (Nanotec Electrónica S.L., Madrid, Spain) was operated in Jumping Mode Plus (38) in liquid using rectangular cantilevers RC800PSA and Biolevers (BL-AC40TS) (Olympus, Tokyo, Japan) with nominal spring constants of 0.05 and 0.03 N/m, respectively. Cantilever spring constants were routinely calibrated by using the Sader method (39). AFM exploration at forces of ϳ100 pN indicated a random population of virions on the mica surface.
Single virus particles were deformed by the AFM tip by performing individual indentation (force versus z-piezo displacement (FZ)) experiments. To ensure that the indentation is performed at the top of the virus, the virus shell is zoomed in continuously by reducing the x-y scanning size until the bump of the very top is under the whole piezo scan. Then a large force indentation was executed at the top of the particle. Finally, an image of the virus was taken to confirm its disruption. The maximum force applied during each indentation was high enough to ensure that the virus particles are destroyed. The FZ speed is about 60 nm/s (40). Images were processed using the WSxM software (41).

RESULTS
Thermal Capsid Disruption-To directly detect the energy required to disrupt the capsid and to monitor the stability of the intermediates formed along the heat-induced disassembly process, DSC was used. DSC scans of mature (WT) and immature (ts1) adenovirus at physiological ionic strength and neutral pH are shown in Fig. 1A. As expected for a large, complex macromolecular specimen, thermal denaturation of both WT and ts1 adenovirus was irreversible upon heating above 90°C. Both thermograms show three well defined transitions (denoted as T2, T3, and T4 in Table 1) preceded by a broad endotherm between 36 and 65°C in the ts1 mutant and three discrete peaks in WT that extend over a similar range of temperature (Fig. 1A, inset). Indeed, deconvolution of the heat capacity profiles showed that the overall endotherm of both WT and ts1 adenovirus can be described in terms of six irreversible two-state transitions whose apparent transition enthalpies, activation energies, and T m values are summarized in Table 1.
The two last peaks of the thermograms (T m values at 71.7 and 78.4°C for WT and 72.8 and 81.0°C for ts1) have previously been assigned to denaturation of the purified hexon protein (42), which is the most abundant component (Ͼ60%) of the protein shell. The minor transition (T 2 ) appearing at 67.3°C for WT and 68.9°C for ts1 seems to be related to hexon stabilization upon GON formation, as it is absent in polypeptide IXdeleted recombinant viruses (42). The three low temperature transitions (denoted as T 1A , T 1B , and T 1C in Table 1) may correspond to capsid disassembly and reflect changes undergone by adenovirus during uncoating in the cell. However, whether they are directly due to disruption of hexon interactions or reflect denaturation of other virion components cannot at present be ascertained. All transitions occurred at higher temperatures for ts1 than for WT, with differences in the 1-2°C range for all T m values except T 1A and T 1B , where differences were as high as 5°C. These differences reveal an increased thermal stability of hexon and other virion structures in the immature capsid context. To define the conditions in which the particle undergoes partial disassembly and the DNA is exposed, we measured the DNA accessibility to solvent by using the DNA intercalating dye PI. PI fluorescence is quenched until the virus capsid is compromised and enhanced upon binding to DNA when it becomes accessible (43). As we have shown previously (5), similar degrees of DNA exposure occurred at higher temperatures for ts1 than for WT (Fig. 1B). EF measurements were fitted to multiple sigmoidal curves with parameters shown in Table 2. WT presents a fluorescence burst at 47.5°C. The sharp slope of the curve at this temperature indicates a highly cooperative process, also observed for WT forms of HAdV-2 and HAdV-5 at ϳ45°C (5,44,45). In contrast, all transitions showed low cooperativity in ts1.
EF data were also analyzed using the irreversible transition model used for the DSC curves (Fig. 1B), and the best fitting parameters are compared with those derived from the heat capacity profiles in Table 3. The T m values obtained from EF experiments for WT closely agreed with those deduced from DSC (Table 3), indicating that capsid protein denaturation (DSC) and DNA exposure (EF) events are correlated. Transition overlapping in the immature virus impaired a reliable anal-ysis of the EF curve, as around 90% of the total emission variation can be apparently described by a single transition with a T m value of 57.0°C. Introduction of a third transition approached the T m values to the DSC estimates but increased significantly the errors of the thermal parameters without improving the R 2 value.
Structural changes due to thermal disassembly were followed by EM. We had previously observed that WT capsids were completely disrupted and their contents released at 47°C, whereas the ts1 virion only lost vertex structures (pentons and peripentonal hexons, or GOS) and extruded part of its contents as a single nucleoprotein filament (5). Here, we report a quantitative description of capsid disruption patterns at different temperatures.
WT and ts1 preparations were imaged by negative staining EM after heating at the temperatures corresponding to T m values derived for WT. To interpret the EM observation, we grouped the observed particles into three categories (Fig. 1C): intact virions (I); damaged capsids, still retaining a spherical arrangement (D); and collapsed, flattened capsids (F). Fig. 1D shows examples of particles in the three categories (see also supplemental Fig. S1). At the T 1A transition temperature (40°C), most ts1 virions were intact, whereas a considerable proportion (ϳ30%) of WT capsids showed small defects consistent with loss of pentons (Fig. 1D). The fact that the interior of these capsids does not appear dark indicates that they still contain stain-excluding material (protein/DNA), in agreement with the modest increase in DNA accessibility to PI indicated by EF. At the T 1B transition temperature (47°C), only 10% of WT virions retained their structural integrity, with the rest showing different degrees of damage. In particular, about 40% of the WT virions appear as collapsed flat structures, with the icosahe-  Chemical Capsid Disruption-Next, we subjected the adenovirus mature and immature capsids to different sources of chemical stress. Immature ts1 virions showed higher stability than WT both in the presence of denaturant compounds and, most interestingly, upon acidification.
A critical point in adenovirus uncoating in vivo is acidification within the early endosome, where the luminal pH ranges between 6.5 and 6 (46). Upon acidification, the increase of PI fluorescence for WT, indicative of DNA accessibility, proceeds in a biphasic way (Fig. 2A). The first highly cooperative transition is centered at pH 6.1 and starts as the medium acidity approaches the endosomal range of pH (Table 4). In contrast, the immature virus withstood more severe values of pH (Fig.  2A) and the ts1 curve showed, apparently, only a broad transition centered at pH 4.1, which is near the midpoint of the second WT transition. At pH values below 3, both viruses present the same behavior. A highly cooperative transition leading to almost complete DNA exposure was also observed upon WT capsid disruption by the denaturant agent pyridine (supplemental Fig. S3 and supplemental Table S1). EM analysis ( Fig. 2B and supplemental Fig. S2) indicated that at pH 6 over 80% of the ts1 particles were intact but more than 70% WT virions already showed damage. Again, complete collapse and flattening were rarely observed for ts1 and WT. The same tendency regarding flattening was observed for disruption by pyridine (supplemental Fig. S3).
Acidification damage produced capsids with dark centers, indicating penetration of the staining agent. However, no evident gaps indicating loss of pentons were observed, meaning that the capsid defect allowing DNA exposure to PI and uranyl acetate is small and of low occurrence in the icosahedral network. Additionally, new kinds of disassembly intermediates were observed: particles with fissures and particles that had lost capsid fragments of different sizes and shapes ("chunks") but still retaining the polyhedral curvature (Fig. 2C). These kinds of images are consistent with the capsid releasing the stress imposed by acidification via cracks at hexameric positions, instead of by releasing vertex structures (47).
Interestingly, 74% of WT damaged particles at pH 6 (55% over total number of particles) were stain-filled capsids. In ts1, however, disruption via fissures and chunk release was more abundant: 56% (46% of the total) of ts1 damaged particles corresponded to this category at pH 4.5, and 60% (58% of the total) at pH 3.5. This difference in disruption patterns suggests that the defect in the WT capsid that allows stain penetration is the physiological path for partial uncoating in the endosome, rather than the disruption via cracks. However, from the negative stain images, it was not clear exactly what this defect was, as no clear pattern of lost capsomers (e.g. vertices) was observed.
Cryoelectron Tomography of Disassembly Intermediates-DSC and EF provide bulk information on changes undergone by the specimen, but they do not reveal changes in individual virions. Imaging disassembled virions by negative stain provided single virion images; however, interaction with the carbon support and the acidic staining agent, plus drying and flattening, may result in additional changes not directly related to the stress conditions being tested. Furthermore, in negative stain images only a two-dimensional projection of the three-dimensional virion is obtained, and it is not possible to observe the internal structures of the particle. To obtain a better picture of the structural changes occurring during capsid disruption, we used cryoelectron tomography.
First, we investigated the changes in the structure of WT viruses at the stressor conditions where the first transition was observed in DSC or EF assays and where ts1 virions remained mostly unaltered (Figs. 1-3 and supplemental Movies S1-S3): at 40°C or pH 6.0. In both cases, tomograms showed that capsids conserved to a large extent their structural integrity, including attached fibers ( Fig. 3A and supplemental Fig. S5). However, the aspect of the core differed between control and stressed viruses. In control conditions (25°C, pH 7.4), the virion interior presented a smooth, flat density with gray levels similar to those of the capsid shell. In viruses incubated at pH 6, the core density appeared more heterogeneous and slightly weaker than that of the capsid. This same difference, sometimes more pronounced, was observed in the core of viruses heated at 40°C. A radial average profile calculated from aligned individual virus tomograms indicated that, for both types of mild stress, the density of the outer region of the core was lower than in control viruses (Fig. 3B). This change in electron density is compatible with a relaxation of the DNA-protein complex or loss of peripheral core material. No decrease in core density was observed for ts1 virions at pH 6 (supplemental Fig. S4).
Next, we searched for capsid defects that would account for the increase in DNA exposure observed in EF assays. Slice-byslice analysis of individual virion tomograms indicated that both mild heating and acidification resulted in loss of pentons ( Fig. 3C and supplemental Fig. S5). However, the degree of loss was different depending on the kind of stress; acidified viruses lost only one to two pentons, whereas heating produced the release of three to four pentons (Fig. 3D).
Concomitantly with a larger loss of pentons, fewer fibers were observed in heat-disrupted WT virions, although upon disruption at pH 6 more than 70% virions still conserved fibers (supplemental Fig. S5), i.e. mild acidification by itself is not causing massive fiber release. Occasionally, both at 40°C or at pH 6, a short (25-30 nm long and 4 -5 nm wide) stretch of DNA was seen to protrude out of the missing penton position (supplemental Fig. S6). However, this was an extremely rare event (less than 5% of analyzed viruses under mild stress conditions).
Tomograms of WT and ts1 virions subject to high stress conditions (heating at 47°C or acidification at pH 5; also pH 4 for ts1) were also obtained. At 47°C, most WT virions appeared as empty broken shells with missing GOS (Fig. 4A, supplemental Fig. S5 and supplemental Movies S4 and S5). Only in a few cases some loose core structures remained attached to capsid fragments. Most immature ts1 virions had also lost GOS and in some cases were cracked open; however, the core remained attached to cracked capsids and in a compacted state. We have previously reported that ts1 virions disrupted at 47°C extrude a nucleoprotein filament (5). However, no such filaments were observed in cryoelectron tomograms, although they are common occurrences in negative staining (5) or platinum/carbon shadowing preparations (data not shown). We conclude that filament extrusion is induced by the EM preparation conditions, namely a combination of adsorption to the support, staining with low pH uranyl salts or shadowing, and drying. A considerable proportion of ts1 capsids remained intact at low pH values (more than 90% at pH 6 and up to 60% at pH 4; supplemental Fig. S5). Notably, fibers (which are not directly involved in proteolytic processing) remain attached to immature virions even under conditions as harsh as pH 4 (supplemental Fig. S5). Upon high acidification, WT shells had gaps consistent with loss of GOS and were only rarely seen in contact with any core remains. Conversely, ts1 capsids only lost pentons and cracked open, whereas cores remained attached to shell fragments and in a condensed state (Fig. 4B and supplemental Movies S6 and S7).
Unlike for negative staining EM, no flattened capsids were observed in cryoelectron tomography images, confirming that flattening was a preparation artifact. However, measurement of dihedral angles in tomograms indicated that in high stress conditions the curvature of disrupted capsids departed from the icosahedral value (Table 5 and Fig. 4C) in both WT and ts1 but with a more noticeable effect in WT. We conclude that flattened capsids observed by negative staining in WT correspond in solution to empty deformed capsids with reduced curvature reinforcements that become flat upon adsorption to carbon staining and drying. In ts1, loss of curvature is not so pronounced, and capsids are less prone to flattening during the negative staining procedure. Interestingly, curvature characteristics are preserved in ts1 even in conditions where large vertex structures (GOS) are lost (47°C). This observation suggests that the tighter interactions between capsid and compacted core established by precursor proteins play a role in determination and maintenance of capsid curvature during assembly.
For both WT and ts1, a small proportion of particles that appeared as intact capsids was found to lack some pentons after closer inspection (supplemental Fig. S5, class "P"). When these particles were aligned and averaged, radial profiles indicated that the WT core density had decreased not only in the periphery but also in the center of the virion, i.e. core relaxation or component loss increased with increasing stress levels. Conversely, core density in ts1 virions lacking some pentons remained unaltered, even at pH 5 or 47°C (supplemental Fig.  S4).
Mechanical Capsid Disruption-AFM nanoindentation assays allow exploring virus stability under mechanical stress. After locating an individual intact virus on the surface (Fig. 5A), the cantilever pushes the virion at the very top of the particle. Cantilever bending is recorded as a function of z displacement, thus eliciting the virus deformation as a function of the vertical force (48). During the first stages of indentation, the virus particles undergo a linear deformation that is reflected by a sustained climbing in the force indentation curve (Fig. 5, C and D) and provides the spring constant of the virus k v (49). At a certain point, the curve tendency sharply changes, and the force suddenly starts decreasing due to capsid failure. The point where tendency changes determines the breaking force. Finally, an image of the virus is taken to confirm its disruption (Fig. 5B). In the experimental conditions used here, viruses preferentially attached to the mica in a 3-fold orientation.
We recorded data from single nanoindentations on 25 WT (Fig. 5C) and 28 ts1 (Fig. 5D) particles. From the results obtained in conditions of thermal or chemical stress, we would have expected to find the mature capsid more fragile and softer than the immature one. On the contrary, WT virions presented both higher breaking force

TABLE 5 Statistics of dihedral angles in disrupted capsids under high stress
A total of 30 individual virus tomograms were examined for each condition. The number of angle measurements n for each case is indicated.

Condition WT ts1
Perfect icosahedron 138°138°C ontrol 137 Ϯ 3°(n ϭ 180) 136 Ϯ 3°(n ϭ 180) 47°C 125 Ϯ 13°(n ϭ 160) 133 Ϯ 8°(n ϭ 170) pH 5 132 Ϯ 14°(n ϭ 168) 134 Ϯ 11°(n ϭ 169) Fig. 5, C and D, the total indentation found for adenovirus was equal to or greater than 40 nm, which is about half the capsid diameter. Taking into account this depth, and the particle topography after indentation (Fig. 5B), it is clear that the tip penetrates the virus core. Therefore, the quantitative analysis of the abrupt decay of the force beyond capsid breakage (inset in Fig. 5E) provides information about core fraying. Converting back the force (y axes of Fig. 5, C and D) to cantilever deflection (nm), we estimated the cantilever bending during post-failure indentation by calculating the slope from the breakage point to 40 nm of indentation, and we called this dimensionless magnitude breakage-indentation ratio (BIR). The more negative the BIR, the faster the core is penetrated by the tip in free fall once the capsid is broken. The classification of BIR in histograms for both mature and immature virions (Fig. 5E) shows that WT particles yield faster than ts1 ones, which even show BIR-positive values. This finding provides direct physical evidence of the higher resilience of the immature core, while mature cores unravel more easily when pushed by the AFM tip.

DISCUSSION
Adenovirus maturation and uncoating have been extensively explored for several decades. Elegant biophysics, cellular and molecular biology studies have given insight into the maturation process on the one hand (22,26,50) and the sequential uncoating on the other hand (8,9,12,13). The entry defect of ts1 had previously been related with increased stability (5,8,51). However, the structural and physical determinants conferring extra stability to the immature virion had not been characterized in detail nor had the structural changes undergone during uncoating been observed at single virion levels in three-dimensional maps. Therefore, the results presented here help to elucidate the role of proteolytic maturation in adenovirus assembly and uncoating.
Implications for Assembly-Among the dsDNA viruses, maturation energetics are best understood for bacteriophage P22 and HK97 (52,53). In both cases, large conformational rearrangements accompany a progressive stabilization of the virus particle. In contrast, adenovirus maturation is not accompanied by large structural rearrangements (4,5). Now we show that, in contrast to the bacteriophage case, the immature adenovirus particle is more stable than the mature one under a variety of stress conditions. To our knowledge, this is the first case reported of a virus assembly intermediate more stable than the final, infectious product.
What makes the immature virion highly stable? The three thermal transitions detected by DSC during capsid disruption take place at lower temperatures in WT than in ts1 virus, suggesting a role of protein maturation in virus stabilization. None of these transitions was observed in purified hexon denaturation (42). They may therefore arise from disruption of interactions between hexons in the assembled virion or from denaturation of other proteins present at high enough concentration to be detected. If this were the case, pVII and pVI, followed by pIIIa, penton base, IX, pVIII, and fiber protein (in this particular order), would be the most probable candidates, taking into account their respective contributions to the total protein content. Interestingly, four out of the six candidates, and notably the two most abundant, pVII and pVI, are processed by the virus protease and participate in capsid and core interactions that are affected during virus maturation (5). In addition, by sequence modification, proteolytic maturation might directly affect protein intrinsic stability but also indirect stabilization through available intermolecular interactions. Both effects could explain the differences found in the T m values of transitions leading to virion disassembly and would reinforce the role of precursors as scaffold elements, involved in energetically stabilizing the assembly intermediates and increasing the chances of assembly. A similar process has been described in bacteriophage HK97, where the coat protein ⌬-domain acts as a scaffold to promote procapsid assembly and then is processed by the virus protease allowing the transition to the next assembly intermediate (54).
We have previously shown that the presence of capsid (pIIIa, pVI, and pVIII) and core (pVII and pre-) precursor proteins results in additional ordered interactions in the virus shell and core (5). Here, we show the effect of these extra interactions. First, immature particles are resistant to the loss of pentons, under conditions in which mature virions readily release them (e.g. 40°C, pH 6). This difference indicates a role for the precursor proteins in preventing premature vertex release, particularly during the DNA packaging process. The need to reinforce vertex structures during morphogenesis has previously been hypothesized in other icosahedral viruses (55).
A second aspect of immature adenovirus stabilization is that interactions maintaining the capsid curvature are preserved, even after losing complete vertex structures, including penton and peripentonal hexons (GOS). In the third place, the immature adenovirus core is a tightly compacted sphere that remains bound to capsid fragments even in high stress conditions. Notably, precursor protein pVI, located at the inner surface the icosahedral shell, binds both hexon and DNA (56,57). All these observations indicate that the interactions established by pre-cursor proteins, whether in the capsid or between capsid and core (e.g. pVI), may play a role in establishing and reinforcing correct capsid curvature during assembly.
Implications for Uncoating-A second role of the maturation process is preparing the virus particle for genome release. On entering the new host cell, virions have to fall apart in a concerted way so that their genome is uncoated at the appropriate place and time to start a new productive virus cycle. Here, we show how maturation prepares adenovirus for proper uncoating and what structural changes are undergone by the virion during disruption, particularly at moderately acidic conditions mimicking those in the early endosome.
Unlike the bacteriophage capsids, the mature adenovirus virion does not represent a global energy minimum in the assembly pathway. It is rather a metastable particle, primed for sequential disassembly through a series of irreversible events and massive genome exposure under the appropriate conditions. These energetic differences are likely related to the different infection mechanisms used by bacteriophage and adenovirus. dsDNA bacteriophage translocate their genomes into the host cell leaving the protein shell behind (58), whereas adenovirus is engulfed by the cell and disassembles within to expose its genome to the cell nucleus machinery (59). In this sense, adenovirus is similar to other nonenveloped animal viruses, such as poliovirus, where interaction with the receptor is the trigger to start the cascade of structural changes leading to uncoating (60).
The pathway of capsid dismantling is schematically depicted in Fig. 6. Mature adenovirus uncoating starts by releasing pentons. This was expected, because theoretical models of icosahedral virus capsids predict that vertex capsomers experience higher levels of stress simply due to capsid geometry. They have fewer interactions, because of having only five and not six neighbors, which makes them the first elements to be lost upon disturbances in the system (47). Furthermore, penton loss had already been shown by assays following the fate of different virus proteins upon cell entry (9,10). Our cryoelectron tomography images show for the first time how this happens in single virions. In mild stress conditions, some pentons are released but not all. In particular, mild acidification resulted in the release of only one or two pentons per virion. Initial penton release was not accompanied by loss of other capsomers in the GOS; nor does acidification per se result in fiber release. These observations are in agreement with the hypothesis that an interaction with integrin and a concomitant structural change in penton base are needed for fiber release (11,61).
Penton loss is accompanied by loss of density in the peripheral core regions and, in the case of mild acidification, a massive increase in DNA accessibility to intercalating dyes. Neither pentons nor peripheral core components were released from immature ts1 particles under mild stress. Loss of peripheral core material is consistent with previous observations indicating release of some internal components in the early endosome, such as core polypeptide V and, more critically, the membrane disrupting polypeptide VI (12,13). Both V and VI are thought to occupy positions bridging the internal capsid surface to the core (4,5,62). Other candidates to be released from slightly disrupted WT capsids are the small peptides resulting from AVP cleavages. DNA accessibility in these conditions results from penton defects and core reorganization but not from core ejection, as all virions observed still retained electron dense contents, and only very rarely were short dsDNA stretches observed protruding from the virions. These properties correlate with the need for the partially disassembled virion to keep protecting its genome while trafficking in the cytosol until arrival to the nuclear pore, while at the same time allowing access to cellular sensors triggering inflammatory responses (15).
Higher levels of stress in mature virions result in release of peripentonal hexons, deviations from icosahedral curvature, capsids cracking open, and complete core ejection. As described above, maturation allows for larger variations in capsid curvature and is crucial for relaxing the core organization and facilitating its separation from the disrupted capsid. Theoretical models predict icosahedral capsid burst by cracks running across hexameric positions when protein-protein interactions between capsomers are weakened, for example by swelling, an increase in internal pressure, or curvature changes in general (47). We therefore hypothesize that, although adenovirus does not undergo the large conformational swelling transitions observed in other viruses, a small degree of swelling or an increase in internal pressure may be part of the virus uncoating process.
A Role for Internal Pressure in Adenovirus Priming for Uncoating-Nanoindentation assays using AFM in physiological conditions have lately provided insights into the mechanical properties of virus capsids. Usually, mechanical strength and stiffness are considered a signature of virus stability (63)(64)(65), i.e. the harder the virus, the more resilient to aggressive conditions. However, this may not necessarily be the case, because stiffness does not guarantee that a virus can survive in hostile chemical or thermal environments. This is the kind of behavior we find for adenovirus; immature ts1, which withstands higher levels of thermal or chemical stress than WT, is softer and breaks at lower forces than WT when indented with the AFM tip. Conversely, the metastable WT virions appear mechanically stronger (breaking at higher forces) and stiffer. Thus, in adenovirus, mechanical stability does not correlate with thermal or chemical stability.
The adenovirus mature capsid is brittle and stiff, and its pentons are more prone to come off than those in the elastic immature particle. Maturation drastically changes the core organization and reduces interactions that kept the DNA tightly condensed. The relationship between these observations points to the possibility that relaxation of the core upon maturation increases the internal pressure in the virion. The adenovirus capsid is usually thought to be pressurized, by analogy with other dsDNA viruses (66). However, the case of adenovirus is different from that of bacteriophage, because the dsDNA genome is forming a complex with a large quantity of charge-neutralizing proteins. No direct estimations of the internal pressure in adenovirus are available so far.
Based on the results presented here, we hypothesize that an increase in internal pressure resulting from proteolytic maturation of the charge-screening core polypeptides pVII and preis a determinant in priming adenovirus capsid for stepwise uncoating. This increase, together with severing of capsidcore contacts via the precursor proteins (particularly pVI), would facilitate release of pentons and peripheral core material, followed by capsid deformation, cracking, and complete ejection of the genetic material to achieve virus propagation and therefore infectivity.