1. Cell Biology
  2. Structural Biology and Molecular Biophysics
Download icon

Molecular determinants of large cargo transport into the nucleus

  1. Giulia Paci
  2. Tiantian Zheng
  3. Joana Caria
  4. Anton Zilman  Is a corresponding author
  5. Edward A Lemke  Is a corresponding author
  1. Biocentre, Johannes Gutenberg-University Mainz, Germany
  2. Institute of Molecular Biology, Germany
  3. European Molecular Biology Laboratory, Germany
  4. Department of Physics, University of Toronto, Canada
  5. Institute for Biomaterials and Biomedical Engineering (IBBME), University of Toronto, Canada
Research Article
  • Cited 0
  • Views 1,445
  • Annotations
Cite this article as: eLife 2020;9:e55963 doi: 10.7554/eLife.55963

Abstract

Nucleocytoplasmic transport is tightly regulated by the nuclear pore complex (NPC). Among the thousands of molecules that cross the NPC, even very large (>15 nm) cargoes such as pathogens, mRNAs and pre-ribosomes can pass the NPC intact. For these cargoes, there is little quantitative understanding of the requirements for their nuclear import, especially the role of multivalent binding to transport receptors via nuclear localisation sequences (NLSs) and the effect of size on import efficiency. Here, we assayed nuclear import kinetics of 30 large cargo models based on four capsid-like particles in the size range of 17–36 nm, with tuneable numbers of up to 240 NLSs. We show that the requirements for nuclear transport can be recapitulated by a simple two-parameter biophysical model that correlates the import flux with the energetics of large cargo transport through the NPC. Together, our results reveal key molecular determinants of large cargo import in cells.

eLife digest

Eukaryotes, such as animals, plants and fungi, store the genetic material within their cells inside a specific compartment called the nucleus. Surrounding the nucleus is a protective membrane which molecules must pass across in order to reach the cell’s DNA. Straddling the membrane are nuclear pore complexes, or NPCs for short, which act as the gatekeepers to the nucleus, shuttling thousands of different molecules back and forth whilst restricting access to others.

Large cargoes need to have specific markers on their surface called nuclear localization signals in order to be transported by NPCs. Certain transporter proteins help the NPC carry large molecules across the membrane by binding to these signals. This generates the energy needed to overcome the barrier of transporting it across the membrane.

Some viruses have nuclear localization signals of their own, which can exploit this transport system; these signals allow the virus to enter the nucleus and hijack the genetic machinery of the cell. It has been suggested that viruses have multiple copies of these surface signals to improve their chances of reaching the nucleus. However, it remained unclear how the number of nuclear localization signals affects the transport of large molecules into the nucleus.

To answer this question, Paci et al. engineered a range of different sized particles derived from viral structures which had varying numbers of nuclear localization signals on their surface. These particles were inserted into human cell lines grown in the laboratory, and imaged to see how they were transported into the nucleus. The rate of nuclear transport was then measured for each particle, and this data was used to create a mathematical model.

Paci et al. found that the larger the cargo, the more nuclear localization signals it needed to be efficiently transported across the membrane into the nucleus. This is because inserting big cargoes into the NPC requires more energy. Therefore, by increasing the number of surface signals transporter proteins can bind to, larger molecules are able to interact with the NPC and generate the energy required for crossing.

These findings improve our current understanding of how nuclear transport could be hijacked by viruses. It could also help scientists who are developing targeted nanoparticles to deliver therapies for genetic conditions to the nucleus.

Introduction

Cargo transport across the nuclear envelope is a hallmark of eukaryotic cells and is central to cellular viability (Knockenhauer and Schwartz, 2016; Jamali et al., 2011; Fahrenkrog and Aebi, 2003). In a typical HeLa cell, more than 2000 nuclear pore complexes (NPCs) span the nuclear envelope (Ribbeck and Görlich, 2001; Maul et al., 1972). With ≈120 MDa in metazoans (Reichelt et al., 1990) and roughly half that weight in yeast (Rout and Blobel, 1993; Yang et al., 1998), the NPC is among the largest macromolecular complexes found inside the cell. NPCs are the gatekeepers of nucleocytoplasmic transport and restrict access of cargoes larger than the typically reported threshold of 40 kDa (Paine et al., 1975; Keminer and Peters, 1999; Mohr et al., 2009), albeit recent work points to a rather ‘soft’ barrier model and a gradual decrease of passive transport rates with size (Timney et al., 2016). However, cargoes that present a nuclear localisation sequence (NLS) and bind nuclear transport receptors (NTRs) can rapidly enter into the nucleus. Several studies have characterized the NTR-mediated transport process, typically focusing on cargoes with one to five NLSs, and their nuclear import kinetics have been shown to follow a mono-exponential behaviour (Ribbeck and Görlich, 2001; Kopito and Elbaum, 2007; Timney et al., 2006).

NPCs are remarkable in the diversity of sizes of cargoes they can transport, ranging from import of nuclear proteins (including histones and transcription factors), to viral import and nuclear export of pre-ribosomal subunits and mRNA complexes (Panté and Kann, 2002; Grünwald and Singer, 2010; Grünwald et al., 2011; Babcock et al., 2004; Mor et al., 2010; Au and Panté, 2012). How very large cargoes (>15 nm) can be efficiently transported is still an enigma, especially considering the dimensions and structure of the transport conduit itself. The NPC is formed by multiple copies of about 30 proteins, two thirds of which are folded proteins that assemble the NPC scaffold. The recent improvements in electron tomography (ET), paired with X-ray crystallography, have greatly expanded our knowledge on the organisation of these folded components of the NPC (von Appen et al., 2015; Szymborska et al., 2013; Lin et al., 2016; Kosinski et al., 2016). This pore-like scaffold is filled with multiple copies of ≈10 different intrinsically disordered proteins, known as FG nucleoporins (FG Nups), which form the NPC permeability barrier. FG repeats have been estimated to be at concentrations in the mM range inside the NPC (Aramburu and Lemke, 2017; Frey and Görlich, 2007). Our structural knowledge about the actual transport conduit compared to the scaffold is much lower, as its dynamic nature leads to a loss of electron density in the averaging process inherent to ET, leaving a ≈40 nm wide ‘hole’ inside the structural map of the NPC tomogram. As the transport of many large cargoes is believed not to irreversibly alter the structure of the NPC, substantial amounts of FG Nups mass must be displaced in order to facilitate such transport events. In addition to dynamics in the permeability barrier, dilation mechanisms in the scaffold structure itself have also been suggested (Beck and Hurt, 2017).

Despite its high biological relevance, nuclear transport of large cargoes is still poorly understood. In order to address this gap, we designed a set of large model cargoes based on capsid-derived structures. In contrast to using fully physiological large cargoes, such as complete viruses, this strategy enabled us to titrate key features such as size, number of binding sites and surface properties. This reductionist approach opened the possibility to experimentally measure a rigorous set of biophysical parameters. We used a combination of spectroscopy and semi-automated microscopy assays to investigate the kinetics of nuclear import of cargoes ranging from 17 to 36 nm in diameter and with a number of NLSs between 0 and 240 in permeabilised cells. Our results uncovered the quantitative dependence of cargo size and NLS number in an understudied size range. The results are rationalized using a minimal physical model of nuclear transport that takes into account the energy gain from NTR binding to FG motifs, and the free energy cost needed for the insertion of a large particle into a densely filled channel.

Results and discussion

A large cargo toolkit for nuclear transport studies

We first aimed to develop a set of model import cargoes with known size and tuneable number of NLSs (#NLSs) on their surface. Naturally occurring cargoes with multiple NLSs, such as proteasomes, pre-ribosomes, mRNA or RNA-protein complexes do not offer the possibility to control both properties reliably at the same time. Vice versa, for artificial large substrates, like quantum dots or gold nanoparticles, it can be challenging to tune size and #NLSs and extensive functionalisation is typically required. Thus we turned to viral capsids, which are known to self-assemble from one or few proteins into large structures of fixed size. We screened the literature for capsid-like particles obeying the following criteria: i) Large-scale high yielding recombinant expression is possible in an expression host like Escherichia coli. ii) Surface modification via a unique residue is possible. Thus, we focused on systems with existing crystal and/or EM structures and checked for single functional surface exposed cysteines or the possibility of mutating another residue to one with no impact on capsid assembly. iii) Capsid is stable at physiological conditions. iv) Capsid diameter is between 15 nm and 40 nm: this size range focuses on rather uncharted territory, with its upper limit reported to be the largest size of cargoes transported by the NPC (Hepatitis B virus, Panté and Kann, 2002). As a result, the following four icosahedral shaped capsids of different size were selected for this study (Figure 1).

Figure 1 with 1 supplement see all
A large cargo ‘toolkit’ for nuclear import studies.

(A) Schematic representation of the mixed labelling reaction with maleimide reactive NLS peptide and maleimide reactive fluorescent dye. The capsid protein, containing a cysteine mutation (in red), self-assembles into a capsid. The purified capsids are then labelled with a mixture of dye and NLS peptide, in different ratios according to the desired reaction outcome. (B) Capsid structures rendered in Chimera (Pettersen et al., 2004) (top) and EM images of the purified capsids (bottom). The scale bar corresponds to 50 nm. (C) SDS-PAGE gel of MS2S37P samples with increasing number of NLS peptides attached (top band). The lower band corresponds to a capsid protein tagged with dye or no dye, but 0 NLS. The upper band corresponds always to the capsid protein without any dye, but NLS, as evident from the fluorescent scan on the right side. (D) Representative FCS autocorrelation curves for the MS2S37P, I53-47 and MS2 capsids. The curves were fitted with a diffusion model to calculate the capsid brightness and concentration. (E) DLS quantification of capsid diameters (blue bars) compared with reference values from literature and structural information (red bars).

MS2S37P (diameter 17 nm): This capsid is derived from the bacteriophage MS2, formed by a single coat protein with a point mutation S37P. The coat protein assembles into dimers and then into 12 pentamers yielding an icosahedron with a total of 60 copies (Asensio et al., 2016). A cysteine mutation (T15C) that had previously been shown not to interfere with capsid assembly was introduced to allow surface tagging via maleimide labelling (Peabody, 2003).

I53-47 (diameter 23 nm): This artificial capsid is derived from de novo designed capsids developed by the Baker lab (Bale et al., 2016). The I53-47 variant is formed by two different proteins (chain A and chain B), occurring in 60 copies each and organised into 12 pentamers and 20 trimers. A cysteine mutation exposed on the capsid surface was introduced in chain B (D43C), following the recent work where different surface mutations were introduced in a similar capsid variant (Butterfield et al., 2017). We note that another synthetic capsid of a similar type but 27 nm in diameter, I53-50, could not be specifically labelled and thus was not included in this work (Figure 1—figure supplement 1).

MS2 (diameter 27 nm): This capsid is derived from the wild-type bacteriophage MS2 coat protein, which in total of 180 copies assembles into dimers and then into an icosahedron with 12 pentameric and 20 hexameric faces. The same cysteine mutation as in MS2S37P (T15C) enabled tagging via maleimide labelling (Peabody, 2003).

Hepatitis B capsid (diameter 36 or 32 nm depending on isoform): This capsid is based on an assembly-competent truncated version of the HBV core protein (aa 1–149). This truncation leads to higher levels of bacterial expression and to a predominance of the T = 4 capsid (36 nm) with no obvious change in capsid morphology (Zlotnick et al., 1996). The core protein thus assembles mainly into 12 pentameric and 30 hexameric units, for a total of 240 copies. The truncation also removes the C-terminal native NLS (which can be buried inside the capsid), enabling a complete control over the number of exposed NLSs via surface engineering. A cysteine mutation (S81C) was introduced into an exposed loop of the core protein (c/e1 epitope) to allow surface tagging via maleimide labelling. The Hepatitis B capsid is frequently quoted as the largest cargo known to pass the NPC intact (Panté and Kann, 2002), and constitutes the upper limit of the cargoes we investigated.

After successful purification, the next step was to engineer the capsid surface with a fluorescent dye and with NLSs. As detailed in the methods, the use of tangential flow for sample concentration and buffer exchange turned out to be of highest practical relevance to purify preparative amounts of intact capsids for further labelling reactions. We chose maleimide reactive dyes and a synthetic maleimide reactive NLS, with a sequence known to bind tightly to Importinα, which binds to Importinβ via its IBB domain (Hodel et al., 2001). Capsids were labelled with suitable mixtures of dye and NLS peptide simultaneously.

Figure 1 summarises the labelling scheme used for all capsids and its characterisation. Figure 1B shows negative staining EM images of capsids after purification and labelling, visualizing intact capsids with the expected diameter. To guarantee the robustness of the quantitative experiments, it was crucial to determine each capsids’ fluorescence brightness (i.e. how many dyes are attached to one capsid) as well as the #NLSs. The #dyes/capsid was determined via fluorescence correlation spectroscopy (FCS), a widely employed biophysical tool to probe brightness and concentration of a freely diffusing species (Figure 1D, Table 1). FCS can also be used to estimate the size and size distribution (such as substantial contaminations of other species than intact capsids) of the samples, which was found to be in line with the high purity indicated by the EM micrographs. Additional DLS (dynamic light scattering) studies were employed to further validate capsid diameter in solution and presence of intact capsids as the dominant species (Figure 1E). The #NLSs was determined from gel shift assays, as NLS-labelled capsid monomers migrate substantially different than their unlabelled counterparts. In contrast, the dye labelling did not alter capsid monomers mobility on gel (Figure 1C). Estimated #NLSs and #dyes/capsid are listed for all samples in Table 1. The presence of a single cysteine per monomer ensures that each is labelled either with NLS or dye, but not both: in this way, unassembled monomers cannot be fluorescently detectable NLS-dependent import substrates. We note that labelling with a synthetic NLS pre-tagged with a dye was found to be impractical in preliminary experiments, as the unreacted species can contribute to elevated background fluorescence in the nucleus.

Import kinetics of large cargoes are tuned by size and NLS numbers (#NLS)

The different labelled capsids samples (total 30, Table 1) were subjected to nuclear import assays using the widely employed permeabilised cell assay (Adam et al., 1990). Figure 2A outlines the details of the experiments. In brief, mild digitonin treatment was used to permeabilise the plasma membrane of HeLa cells, leaving the nuclear envelope intact. In these conditions, functional nucleocytoplasmic transport can be reconstituted for a few hours by adding the key components of the transport machinery: Importinβ, Importinα, RanGDP, NTF2 (a NTR which allows recycling of RanGDP) and GTP to the cells. Intactness of the nuclear envelope and functional nuclear transport were always validated by a set of control experiments using fluorescently labelled dextran and model cargoes (see Materials and methods). As shown in Figure 2B exemplarily for the MS2S37P and MS2 capsids, cargoes labelled with NLSs showed an increased nuclear accumulation over time, indicative of functional nuclear import.

Figure 2 with 2 supplements see all
Pipeline for import kinetic experiments.

(A) Scheme of the transport assay experiment: HeLa cells were permeabilised and incubated with a transport mix containing the cargo of interest, nuclear transport receptors and energy. Confocal images were acquired in 12 different areas every 2 min, for 80 min in total. (B) Representative time-lapse snapshots of cargo import (MS2S37P and MS2 capsids). The scale bar corresponds to 20 μm. (C) Overview of the image analysis pipeline for import kinetics experiments. Two reference stain images (Hoechst and MitoTracker) were segmented and used to generate three masks corresponding to the regions of interest: nucleus, nuclear envelope and cytoplasm. The masks were then applied to the cargo images to calculate the average intensity in the different regions. (D) Representative raw import kinetics traces for the three cellular compartments of interest. Note that imaging starts after 2 min of adding the transport mix to the cells. Curves depict the average fluorescence measured in the different regions; the shaded areas represent the standard deviation over 12 areas.

Experiments were performed on a semi-automated confocal microscope, recording time-lapse images over several cells and different field of views (error bars correspond to standard deviations between different FOV). Note that for practical reasons, imaging always started ~ 2 min after addition of the transport mix to the cells. This timing offset was accounted for by an offset fitting parameter A in our fit equation (I(t)=A+IMAX(1eτt)).

Besides the nuclear signal, we also recorded the nuclear envelope and cytoplasmic signals using suitable imaging masks (Figure 2C, Materials and methods and Source code 1 for details). We took precautions to distinguish nuclear fluorescence from nuclear envelope fluorescence by eroding the nuclear mask to a region furthest away from the rim. This turned out to be important, as some capsids showed nuclear envelope targeting but no substantial accumulation into the nucleoplasm (for instance, HBV and MS2 capsids with few NLSs). In addition, this method enabled us to discriminate nuclear signal from sticking of capsids to the cytoplasm, which was observed in some cases.

Figure 2D summarises the three kinetic traces that were obtained from a typical experiment. In the representative experiment shown for a MS2S37P capsid sample, the cytoplasmic fluorescence stayed constant, while nuclear envelope signal increased pointing to recruitment and accumulation of capsids at the NPCs. The red curve shows the import kinetics of capsids into the nucleus. Figure 2—figure supplement 1 shows additional control experiments (addition of the Importinα export receptor CAS to transport mix and excess of GTP or Importinα) to establish that the observed saturating nuclear import depends on the substrate size and #NLSs and is not due to any of the components in the transport mix becoming limiting during the course of the experiment.

To further support our findings under fully physiological conditions, we carried out microinjection of representative capsid samples in starfish oocytes to observe their nuclear accumulation in live cells. The results of these experiments are presented in Figure 2—figure supplement 2 and are qualitatively in agreement with the quantitative nuclear import assays in permeabilised cells described in the next paragraph.

Figure 3 shows representative nuclear import data for the three kinetically investigated capsids MS2S37P, I53-47 and MS2 (see Figure 3—figure supplement 1 for full dataset). The results for HBV capsids will be discussed in the next paragraph. Figure 3 panel A displays typical confocal images of cargoes with different #NLSs and panel B shows representative nuclear kinetic traces extracted from semi-automated microscopy. Figure 3—figure supplement 1 shows the full dataset overlaid with the mono-exponential fits. In absence of NLSs (0 NLSs), all capsids localised to the cytoplasm and no targeting to the nuclear envelope or accumulation in the nucleus was observed, in line with an Importin-dependent pathway. With increasing #NLSs present on the capsid surface we observed progressive nuclear envelope targeting, and eventually, efficient accumulation of cargo in the nucleoplasm. Strikingly, the #NLSs required to observe similar behavior with different capsids scaled dramatically with cargo size, as can be seen by comparing for example the I53-47 sample image with 35 NLSs and the MS2 one with 86 NLSs. The observation of robust bulk import for all capsid constructs with sufficiently high #NLSs highlights another benefit of using viral capsids as large cargo models: in a previous study using coated quantum dots (18 nm) no bulk import could be detected but only rare import events were captured by advanced single molecule technologies (Lowe et al., 2010).

Figure 3 with 1 supplement see all
The import kinetics of large cargoes is tuned by the NLS number.

(A) Confocal images of nuclear import of the different large cargoes. Cells were incubated for up to 1.5 hr with capsids tagged with different number of NLS peptides on their surface. All cargoes displayed a distinct NLS-dependent behaviour. The scale bar corresponds to 20 μm. (B) Representative nuclear import traces for the three large cargoes labelled with increasing amount of NLS peptides. The corrected nuclear intensities are obtained by background-subtracting the raw nuclear intensities, scaling them according to capsid brightness (#dyes) estimated from FCS (Table 1) and subtracting the initial offset A determined by the mono-exponential fit, to better compare the import efficiencies. The corrected intensities are proportional to capsid concentration and allow us to compare the import efficiency of the different samples. See Figure 3—figure supplement 1 for the full dataset displayed without offsetting by A and overlaid with mono-exponential fits.

Table 1
Sample properties and parameters from fits of import kinetics.

Here we list all capsid sample properties (estimated #NLSs and #dyes per capsids), as well as all parameters extracted from fitting the import traces with an inverse exponential It=A+IMAX1-e-τ*t. The initial flux is calculated as J=IMAX*τ. When different biological replicates were measured for the same sample, the values indicate the average.

sample#dyes#NLSsAIMAXτJΔG
MS2S37P12300.240.390.0220.016.34
215140.650.940.0540.054.54
330191.1212.440.0410.492.01
434230.6416.630.0530.881.25
525290.3328.910.0220.641.51
638381.4543.590.0371.60−0.02
710542.3349.760.0291.370.32
I53-4782400.181.580.0180.035.27
930150.803.710.0850.323.19
1031183.813.890.0560.223.29
1136222.342.950.0800.233.36
1216222.015.350.0600.323.04
1331251.832.910.0940.273.15
146302.277.220.0630.452.41
153351.1816.820.0480.811.49
168371.216.670.0530.352.28
173372.1613.800.0590.811.52
1810411.1313.920.0570.791.54
198440.3111.780.0380.452.00
MS2205000.070.180.0380.016.88
2138420.470.520.1060.054.94
2258540.191.070.2410.262.95
2344570.071.830.0740.133.49
2452770.471.310.0720.093.96
2561860.552.670.0420.113.72
2657930.382.040.0510.103.88
2754980.273.820.0330.123.49

Modified HBV capsids are targeted to NPCs but do not accumulate into the nucleoplasm

We next used the established pipeline to investigate the transport of HBV capsids, achieving a maximum of 50% capsid monomer labelling (120 NLSs). The capsids were targeted to the nuclear envelope; however, no bulk nuclear import could be detected (Figure 4, first row). As we were not able to further increase the #NLS with our chemical labelling strategies, and we wondered whether 120 NLSs might still be insufficient, we resorted to genetic tools to achieve the full coverage of 240 NLSs per capsid. To do this, we designed a capsid based on the SplitCore construct (Walker et al., 2011), in which a core-GFP fusion protein was split into two halves that self-assemble before forming the capsid. This exposes a free C terminus, which we exploited to introduce an NLS. Also for this capsid, we did not observe any bulk import. However, the slightly increased size due to the GFP could potentially push this capsid over the maximum NPC transport size limit. We thus tested another strategy, and introduced an NLS into an exposed capsid loop (Figure 4, last row). Again, no functional bulk import could be observed. EM showed that the engineered capsids are less homogenous, but still a large number of intact capsid was observed. Hence we conclude that none of the tested HBV capsids constructs can functionally be enriched in the nucleus. As the chances that our careful modifications rendered the HBV capsid transport-incompetent seem rather low, our data is in line with studies that suggest that only the mature infectious virus can translocate through the NPC into the nucleoplasm (Rabe et al., 2003; Kann et al., 1999). Our results are consistent with EM data of intact HBV capsids entering the NPC barrier, (Panté and Kann, 2002) as we also see strong NE accumulation. However, additional mechanisms would be required for cargo release into the nucleoplasm such as the previously reported structural destabilisation of mature capsids (Cui et al., 2013) or other mechanism that can disassemble capsid that are docked at the NPC. Collectively, this suggests that 36 nm capsids might be able to enter the NPC barrier, but are too large to pass the NPC intact into the nucleus (i.e. undock or release). We, thus, focus our global quantitative analysis on the three capsids for which we could experimentally identify conditions of functional import and nuclear enrichment.

The NLS-engineered Hepatitis B capsid is not imported in the nucleus of permeabilised cells.

Following the same labelling approach as described in Figure 1, HBV capsids with up to 120 NLSs were generated (first row). In order to test capsids with a higher number of NLSs exposed on the surface, we designed two additional versions of the HBV core protein with a direct NLS insertion (total of 240 NLSs). The middle row shows a construct based on the SplitCore-SplitGFP (Walker et al., 2011), where the HBV core protein is split via artificial stop and start codons into two halves and fused to a split-GFP (GFPβ1–10 and GFPβ11), to which we further added an NLS. Once co-expressed, the two core-GFP halves self-assemble into capsid-like particles. The last row shows a construct where the NLS is inserted in the c/e1 epitope loop of the core protein (orange loop) and a cysteine mutation is introduced to perform labelling with a dye (red star). All capsids were targeted to the nuclear envelope but did not give rise to bulk nuclear accumulation in import experiments using permeabilised cells. (A) Schematic representations of the different HBV core protein constructs. (B) EM images of the purified capsids. The scale bar corresponds to 100 nm. (C) Confocal images of capsid import experiments after 1.5 hr. The scale bar corresponds to 20 μm.

Quantitative analysis of nuclear import in relation to cargo size and #NLSs

Our results on large cargo import kinetics (Figure 3) highlight the strikingly different #NLS requirements for the nuclear import of differently sized cargoes. We formulate here a biophysical model that considers the translocation of a large ‘spherical object’ through the crowded NPC permeability barrier (scheme in Figure 5A) and enables us to extract key information about the energetics of transport from our kinetic data.

Figure 5 with 5 supplements see all
Effect of cargo size and number of NLSs (#NLSs) on import kinetics and biophysical model.

(A) Cartoon of the determinants for large cargo import: the free energy cost of inserting a large cargo into the dense FG Nup barrier must be compensated by the binding to FG Nups via multiple NTRs (binding sites represented in orange, NTRs omitted for simplicity). The NPC scaffold structure is from EMD-8087. (B) Dependence of ΔG on the capsid size and #NLS for aRan=2. Shaded regions show one standard deviation of FR and ϵ. Fitted values for FR and ϵ are shown in Table 2. (C) Initial flux (corresponding to the slope of the kinetic curve at the initial time point) modelled as J1aRan+eFR-ϵN  overlaid on the (normalised) experimental data (dots). Additional experiments with MS2S37P capsids containing additional charges are overlaid and shown as squares. Whenever independent biological replicates were available, the initial flux is calculated as an average and shown with the error extracted from the technical replicates (12 areas imaged in each experiment). In Figure 5—figure supplement 5 we show that the uncertainty between different cells imaged in a single experiment captures well the variability of independent experiments.

The final steady state accumulation and the late kinetics of the capsid import are affected by a number of factors that are still incompletely understood – such as the competition between Ran and NTRs for the cargo, the back leakage of the cargo into the cytoplasm and potential clogging of the pores by the capsids (Kim and Elbaum, 2013a; Kim and Elbaum, 2013b). For this reason, we focus our quantitative analysis on the initial flux J (slope of the kinetic curve at the initial time point). Unlike the steady state accumulation, the initial flux J of cargoes into the nucleus is independent of the rates of cargo-NTR dissociation kinetics and is less affected by any potential rate-limiting steps in the Ran cycle (Kim and Elbaum, 2013a; Kim and Elbaum, 2013b; Görlich et al., 2003). To this end, all nuclear import curves were fitted with a mono-exponential kinetic model I(t)=A+IMAX(1eτt), with IMAX being the plateau value reached by the fit at infinity, τ the reaction constant with units 1/s and A is the offset parameter. A accounts for any nonzero offset, which could be due to: i) initial recruitment of the cargoes to the cells and nuclear envelope. ii) limiting accuracy in pipetting and sample mixing (there is a 2-min delay in our experiments between the addition of the sample and the start of imaging) and for slightly different background levels due to non-specific adhesion of some samples to cellular structures. A is thus fitted in every experiment and not expected to be a constant. The initial flux can be calculated from the fit parameters as J=IMAX*τ (see Table 1 for values of all fit parameters). We emphasize that the mono-exponential fit is a mathematical tool to estimate the initial flux from the data. Calculating the initial flux from the mono-exponential fits was more robust than the alternative of measuring the initial flux directly from a linear fit of the first few data points, since the timing resolution of the experiment and the accuracy of defining the zero time point when mixing the cargo with the cells was limited. We note that more complex fits, such as bi-exponential fits have been discussed in the literature to include additional effects such as cargo leaking back into the cytoplasm. (Kim and Elbaum, 2013a; Kim and Elbaum, 2013b). In Supplementary file 1, we further compare bi-exponential and the mono-exponential fits. The initial rates for all samples are plotted in Figure 5C (experimental data displayed as dots). We also note that despite the samples having different labelling ratios (see #dyes, Table 1), we confirmed that there were no global correlations between overall #dyes/capsid ratio and extrapolated rate (R2=0.14).

Based on extensive previous theoretical and experimental work on the NPC (Iyer‐Biswas and Zilman, 2016; Zilman, 2009; Zilman et al., 2007; Berezhkovskii et al., 2002; Pagliara et al., 2013), the initial flux J  can be approximated as J=kONcaRan+eΔG, where kON is the rate of cargoes reaching the NPC entrance, c is the concentration of cargoes in the cytoplasm, aRan is a number between 1 and 2 depending on the availability of RanGTP at the nuclear exit (aRan=2 corresponds to the absence of RanGTP, and aRan=1 corresponds to RanGTP always being immediately available at the nuclear exit). ΔG is the effective average non-equilibrium free energy potential of the cargo inside the NPC (expressed in units of kBT0.6 kcal/mol); a conceptually similar expression was used in Frey and Görlich, 2007 to analyse the transport of cargoes through FG Nup 'hydrogels'. This expression mathematically describes the fact that the probability of a particle that impinges on the NPC entrance to actually translocate to the other side is Ptr=1aRan+eΔG due to the random nature of the diffusive motion inside the NPC. For cargoes that are strongly repelled by the FG Nup network, ΔG1, and the flux is exponentially inhibited, as Je-ΔG. By contrast, for cargoes that interact attractively with the FG Nups,  ΔG<0 resulting in significant flux through the pore. However, the flux can be significant and well detectable already for ΔG1kBT. This expression remains valid for the low concentrations studied here and the intermediate values of ΔG appropriate for our capsids. For higher concentrations or higher #NLS, which we did not experimentally assess, additional corrections may need to be introduced (Zilman et al., 2007; Pagliara et al., 2013).

From experimental and theoretical studies (Gu et al., 2017; Ghavami et al., 2016; Vovk et al., 2016; Maguire et al., 2020), in the first approximation the main components of ΔG are: 1) the cost of insertion of the capsid into the FG assembly FR, arising from the entropic cost of FG Nup displacement, osmotic pressure and the effective surface tension penalty, and 2) the effective energetic/enthalpic gain ϵ from the attractive contacts formed between the NTR binding sites and FG repeats, which can partially compensate for the cost of insertion. ΔG can also include non-equilibrium logarithmic corrections arising from the dependence of the diffusion coefficient on the cargo size and the #NLSs (Zilman, 2009; Maguire et al., 2020).

Previous studies indicate that the cost of insertion, FR, increases with the particle size (Gu et al., 2017; Ghavami et al., 2016; Vovk et al., 2016). In simple situations such as the partitioning of a relatively small spherical particle into a polymer brush or a polymer-coated channel representing the NPC, this cost scales as  FRRα with 1 ≤ α ≤ 3, where R is the radius of the particle (Gu et al., 2017; Ghavami et al., 2016) but the exact form of the dependence on R is unknown for very large cargoes studied here. The energetic gain is given by the total average energy of binding between NTRs and FG repeats, which, in the first approximation, for independent binding is expected to be proportional to the #NLSs on the particle surface, which we denote here as N. Combining these terms yields G=FR-ϵN, where ϵ is the effective binding interaction energy of an NTR with the FG environment; ϵ is proportional to the product of the density of the FG motifs in the pore ϕ, and ϵ0, the bare average interaction energy of an NTR with an FG motif, so that ϵ=ϵ0ϕ. When the insertion cost term dominates, ΔG is large and positive, and the initial accumulation rate is low. On the other hand, when the energetic gain term dominates (large N), ΔG decreases and eventually becomes negative, and the initial accumulation rate is high.

In order to gain insight into the transport mechanism, we analysed the experimental data using the minimal model described above. First, we inverted the equation for the initial flux to obtain the ΔG values as a function of the #NLSs N for the three capsids of different sizes. For each capsid size, ΔG  values were fit with a straight line, obtaining the values for FR and ϵ from the y-intercepts and slopes, respectively (Figure 5B). We assumed here aRan=2; results for aRan=1 are very similar, as shown in Figure 5—figure supplement 1. The actual value of the initial flux depends on the kON (see above) and the number of the NPCs in the nuclear envelope – variables that are hard to estimate experimentally. Thus, for the purpose of comparison with the model, the data were normalised to the maximal observed value among all technical replicates that was still within 95% confidence interval for that value. The conclusions of the analysis were robust with respect to the choice of the normalisation constant.

Figure 5C shows the experimentally measured initial flux J data (dots) overlaid with the theoretical equation for J using the values of FR and ϵ obtained from the fit (Figure 5B, parameters values are listed in Table 2). The fits in both figures agree well with our experimental data. Consistent with the theoretical expectations, the cost of insertion FR was the highest for the largest capsid. The differences between the insertion costs for the two smaller capsids were within the error bars. To control for the possibility that the similar values of FR observed for all three capsids are an artefact of the limitations on the experimental accuracy at very low fluxes, we repeated the model fit, excluding the #NLS=0 point, which resulted in essentially the same fitting parameter values (Figure 5—figure supplement 2). Another possibility is that for such large capsids the insertion cost saturates to a plateau value at maximal FG Nup compression.

Table 2
Parameters from free energy fit.

Fitted values for FR and ϵ values, for aRan=2. The error corresponds to the standard deviation.

CapsidDiameter [nm]F(R) [kBT]ϵ [kBT]
MS2S37P175.2 ± 0.90.12 ± 0.03
I53-47234.9 ± 0.30.08 ± 0.01
MS2276.0 ± 0.70.03 ± 0.01

Surprisingly, the ϵ values were different for different capsids, with the ϵ for the MS2 (largest) capsid substantially lower than those for MS2S37P and I53-47 capsids. At first glance, one would expect the main difference in the fluxes of capsids of different size to stem from the difference in the insertion cost FR, while the interaction energy would be relatively unaffected by the particle size. It was also surprising that significant accumulation in or near the nuclear envelope was observed even for the cargo samples whose interaction with the NPC is insufficient to cause substantial nuclear accumulation (Figure 2D).

To further understand the implications of these findings, we extended the model to include a variation in the FG Nup density along the pore. Our model is a variant of previously postulated 'vestibule'/'docking'models (Tagliazucchi et al., 2013; Tu et al., 2013; Lowe et al., 2015), with a central 'barrier' region with high density of FG Nups and correspondingly high insertion cost, and a 'vestibule' outside the NPC (corresponding to a low density cloud of FG Nups extending into the cytoplasm). The capsids weakly bind in the vestibule but experience no insertion cost as FG Nups and capsids are unconstrained by the NPC scaffold in this region. Bridging between the barrier and the vestibule there are narrow transition regions at the NPC peripheries, with a medium density of FG Nups and correspondingly low insertion cost. As shown in Figure 5—figure supplement 3, this simple extension of the model allows us to explain the fluxes of all capsids with the same value of the 'bare' NTR-FG binding energy ϵ0, as well as the accumulation in the cytoplasmic 'vestibule' even at low ϵ. Assuming the average number of FG motifs in the pore ∼3000 (Tu et al., 2013), corresponding to the average volume fraction/density ϕ=0.01, the obtained values of ϵ04-15 kBT are within the range of the common estimates of NTR-FG interaction strength (Aramburu and Lemke, 2017; Tu et al., 2013; Kapinos et al., 2014; Eisele et al., 2010; Milles et al., 2015). This analysis should be viewed with the caveat that this minimal model is likely to be modified in the future with more molecular details; we return to this point in the Discussion.

Surface property effects on large viral import

Surface properties such as charge or hydrophobicity have been frequently indicated to influence the import properties through the nuclear pore complex of smaller cargoes, which in many cases were systematically assessed by creating large data sets in which the cargo properties were carefully studied and/or tuned (Frey et al., 2018; Naim et al., 2009; Colwell et al., 2010).

While our capsid study does not lend itself to similar high throughput screening of surface properties, we speculate on the role of surface properties for large cargoes based on a few observations and experiments. i) We found that our minimal physical model describes our experimental data well. As the capsids all have a different and complex surface properties landscape (see Figure 5—figure supplement 4) this can be seen as an indicator that in the regime studied in this paper, the rules of large cargo transport might be dominated by the size of the capsid sphere and the number of NTRs that it can bind rather than direct interactions between the capsids and FG Nups due to surface effects. A potential exception could be at very low #NLS labelling regime, where the signal-to-noise ratio does not offer a detectable measurement of initial flux. ii) To substantially alter surface charges, we labelled capsids with a longer NLS peptide containing a linker with a negatively charged stretch of amino acids (DEDED). We focused on the MS2S37P capsid with high #NLS labelling, where consequently the largest number of additional charges could be included by this method. As shown in Figure 5C (charged capsid data shown as squares), we did not observe substantially different behaviours in capsids with and without the additional charges. We note that we faced practical hurdles in obtaining capsids with a positively charged linker due to precipitation/aggregation of the peptide during labelling and, thus, were not able to experimentally test this regime. iii) Simple geometrical considerations could also support that for large objects like our capsids the actual surface properties might be less relevant in the regime of large #NLS. If we just focus on Importinβ for simplicity and consider its surface footprint of roughly 20 nm2 for the capsids with highest #NLS (1:1 stoichiometric complex of NLS and Importin), the overall surface shielding by Importins is roughly 100% for MS2S37P, 80% for MS2 and 50% for I53-47. Thus, the substantial cargo decoration with Importins would result in a larger portion of the capsid surface being shielded.

Discussion

Our approach based on modified capsids with tuneable surface properties and quantitative imaging in permeabilised cells enabled us to arrive at a substantially enhanced quantitative understanding of large cargo transport through the NPC. Assaying nuclear import kinetics in an unprecedented cargo size and #NLSs range, we have shown that the requirements for transport scale non-linearly with size and can be recapitulated by a two-parameter biophysical model that correlates the import flux to the energetic requirements for nuclear transport.

For small cargo transport, biochemical or physicochemical properties of the cargo surface have been shown to influence nuclear transport (Frey et al., 2018; Naim et al., 2009). While we do not claim that surface effects play no role in large cargo transport, based on the prediction from our experimental assay we would suggest that the binding of multiple Importin complexes seem to partially mask the cargo surface properties.

Our work significantly expands the range of sizes and #NLSs for which nuclear import has been characterised: Tu et al., 2013 previously reported a single molecule study of a β-galactosidase cargo, which has four NLSs. This approximately cylindrical molecule is 18 nm at its longest axis, similar to MS2S37P, and 9 nm along its shorter axis. If the cargo crosses the NPC in a favourable orientation (through its narrow end), this would result in a lower cost of insertion and explain well why for this substrate 4 NLSs are sufficient for import (Tu et al., 2013). By comparison, our smallest cargo, MS2S37P, which is spherical with a 17 nm diameter, was not substantially imported below 10 NLSs within the timeframe of our assay. For the larger MS2, more than 30 NLSs were required to detect a clear signal. It is important to note that in addition to cargo shape (Mohr et al., 2009), its mechanical stability and rigidity are likely to play a role in nucleocytoplasmic transport: the import rate of proteins is inversely correlated with its mechanical stability (Infante et al., 2019), and flexibility is likely relevant for the transport of large deformable synthetic cargoes, such as polymer vesicles (Zelmer et al., 2020).

While our simple biophysical model can explain the experimental data very well with only two fitting parameters per capsid (FR  and ϵ) it also raises several interesting questions. The model provides quantitative estimates of the free energy cost of capsid insertion into the FG Nup assembly, as well as the effective binding energy needed to compensate for the insertion cost. Notably, despite the fact that a single MS2 capsid already occupies ≈1/3 of the estimated volume of the central NPC channel (Isgro and Schulten, 2005) (as illustrated in the cartoon in Figure 5) the free energy cost of insertion is relatively low (on the order of a few kBT’s), and is similar for the capsids of different sizes. This might indicate that further mechanisms facilitate large cargo transport, such as NPC scaffold dilation, a hypothesis supported also by multiple evidences for tentative hinge elements in the NPC scaffold structures (Bui et al., 2013; Kelley et al., 2015), or bistability in the FG density in the radial direction induced by such extremely large cargo (Osmanović et al., 2013).

In terms of the effective interaction energy ϵ, the largest MS2 capsid required a fit with the lowest effective ϵ. This finding is surprising at first glance, because one would expect that the main difference between the capsids would stem from their size difference, while the interaction energy of an NTR with an FG motif stays relatively constant. One can think of several potential origins for this effect, among those are the lack of independence in the NTR binding of the FG repeats in case of large surface coverage, or the loss of accessibility of the FG motifs due to the high compression of the FG assembly by the largest capsid, which will be explored in future work. Nevertheless, we found that all these features can be explained in a minimal model that incorporates the potential heterogeneity of the FG Nup distribution along the NPC axis, whereby there are at least two different regions of different FG Nup densities, as has been previously suggested in a 'two gate' or 'vestibule' pictures of the NPC (Tu et al., 2013; Lowe et al., 2015; Yamada et al., 2010). In Figure 5—figure supplement 3, we show that such a spatially heterogeneous model would be consistent with the data across all three capsid data sets, without invoking different effective interaction energies for the different capsids.

In our model, the energetic terms represent the binding between FG repeats and NTRs. The microscopic binding mechanism between NTRs and FG repeats during NPC passage is probably similar both for import and export, with a few exceptions - such as the binding of the export factor Crm1 to a specific stretch of Nup214, (which has likely a larger role in undocking than permeability barrier passage) (Port et al., 2015; Tan et al., 2018). We thus anticipate that basic principles of our work could also help in the future to better understand export of large cargoes, such as pre-ribosomal subunits and large RNA export complexes.

The theoretical model used in this paper implicitly assumes that the capsids do not interact with each other during transport through the pore. We cannot exclude multiple capsids colliding with each other in a single pore with absolute certainty - and this has indeed been observed in EM of HBV capsids injected in Xenopus oocytes (Panté and Kann, 2002). However, the hallmark of jamming resulting from multi-particle occupancy is the non-monotonic dependence of the flux on the interaction strength and thus on the #NLS on the capsid (Zilman, 2009; Pagliara et al., 2013) - a trend that is currently not apparent in our data within the experimental accuracy, at least in the initial rate, which is the focus of our analysis.

A more complete picture of nuclear transport and refined model building in the future would require taking into account additional features in more detail, such as docking and undocking from the barrier, more realistic modelling of the capsid cargo passage through the pore, and complex entropic effects of capsid-FG Nup interactions. Future studies using our cargo substrates and time resolved high-resolution measurements could provide further insights into the individual kinetic steps of NPC binding, barrier passage and undocking and how those link to FG Nup and potentially scaffold dynamics in the NPC.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional
information
Strain, strain background (E. coli)BL21Invitrogen/Thermo Fisher ScientificAI strain
Cell Line (Homo-sapiens)HeLa KyotoGift from Martin Beck’s LabRRID:CVCL_1922
Recombinant DNA reagentpBAD_MS2_Coat_Protein–(1–393) (plasmid)This studyProtein expression plasmid for E. coli (MS2)
Recombinant DNA reagentpET29b(+)_I53–47A.1–B.3_D43C (plasmid)This studyProtein expression plasmid for E. coli (I53-47)
Recombinant DNA reagentpBAD_MS2_Coat_Protein–(1–393)_S37P (plasmid)This studyProtein expression plasmid for E. coli (MS2S37P)
Recombinant DNA reagentpET28a2-SCSG-GB1-coreN-GFPβ1–10//NLS-GFPβ11-coreC149H6 (plasmid)This studyProtein expression plasmid for E. coli (HBV SplitCore)
Recombinant DNA reagentpBAD-MCS-CoreN-cys-loop-CoreC-TEV-12His (plasmid)This studyProtein expression plasmid for E. coli (HBV core with cysteine and NLS)
Recombinant DNA reagentpET28a2-HBc14SH6_S81C (plasmid)This studyProtein expression plasmid for E. coli (HBV core with cysteine)
Recombinant DNA reagentpTXB3-12His-Importin beta WT (plasmid)This studyProtein expression plasmid for E. coli (Impβ)
Recombinant DNA reagentpBAD-Importα1-FL-InteinCBD-12His (plasmid)This studyProtein expression plasmid for E. coli (Impα)
Recombinant DNA reagentpTXB3-NTF2-intein-6His (plasmid)This studyProtein expression plasmid for E. coli (NTF2)
Recombinant DNA reagentpTXB3-Ran Human FL-Intein-CBD-12His (plamid)This studyProtein expression plasmid for E. coli (Ran)
Peptide, recombinant proteinNLS peptidePSL GmbHMal-GGGGKTGRLESTPPKKKRKVEDSAS
Peptide, recombinant proteinNLS peptide with additional chargesPSL GmbHMal-DEDED-GGGGKTGRLESTPPKKKRKVEDSAS
Chemical compound, drugHoechstSigmaB2261For nuclei labelling
Chemical compound, drugMitotracker greenInvitrogenM7514For mitochondria
labelling
Chemical compound, drugFITC-Dextran 70 kDaSigma53471Used for checking nuclear envelope integrity
Chemical compound, drugAlexa Fluor 647 maleimideInvitrogenA20347Dye for capsid labelling
Software, algorithmUCSF Chimerahttp://www.rbvi.ucsf.edu/chimera/RRID:SCR_004097
Software, algorithmFijihttps://fiji.sc/#RRID:SCR_002285
Software, algorithmSymphoTimePicoQuantRRID:SCR_016263
Software, algorithmIgor ProWavemetricsRRID:SCR_000325
Software, algorithmAdobe Illustrator CS6AdobeRRID:SCR_010279

Large cargo expression and purification

MS2 and MS2S37P capsids

Request a detailed protocol

A colony of E. coli BL21 AI cells containing the pBAD_MS2_Coat Protein-(1-393) or the pBAD_MS2_Coat Protein-(1-393)_S37P plasmids was inoculated in LB medium containing 50 μg/mL ampicillin. The culture was grown overnight shaking at 37°C (180 rpm) and then used at a 1:100 dilution to inoculate an expression culture in LB medium. Protein expression was induced at OD600 = 0.6–0.7 by adding 0.02% arabinose and carried out at 37°C shaking (180 rpm), for 4 hr. Cells were harvested by centrifugation in a Beckmann centrifuge, rotor JLA 8.100 at 4500 rpm, for 20 min, at 4°C. For purification, pellets were resuspended in an equal volume of lysis buffer (10 mM Tris pH 7.5, 100 mM NaCl, 5 mM DTT, 1 mM MgCl2, 1 mM PMSF) and lysed through 3–4 rounds in a microfluidizer, at 4°C. The lysate was incubated with 0.2% PEI (polyethylenimine) for 1 hr, on ice and then clarified by centrifugation in a Beckmann centrifuge, rotor JA 25.50 at 10,000 rpm, for 30 min. A saturated solution of (NH4)2SO4 was added at 4°C drop-wise to the clear lysate under continuous mild stirring up to 25% of ammonium sulphate. After 1 hr, the lysate was spun down again at 10000 rpm, for 30 min. The supernatant was discarded, and the pellets were gently resuspended with 10–20 mL of lysis buffer on a rotator, at room temperature. The lysate was then centrifuged at 10,000 rpm, for 30 min and the clear supernatant was collected. The supernatant was cleared using the KrosFlo system (SpectrumLabs) with a 0.2 μm cut-off membrane to remove large impurities. The membrane permeate containing the cleared sample was collected on ice. In order to maximise protein recovery, the remaining supernatant was washed with 50 mL of lysis buffer and the permeate was pooled with the previously collected one. The sample was then concentrated using the KrosFlo with a 500 kDa cut-off membrane (for the smaller MS2S37P capsid, a 30 kDa cutoff was used).

I53-47 capsids

Request a detailed protocol

A colony of E. coli BL21 AI cells containing the pET29b(+)_I53-47A.1-B.3_D43C plasmid was inoculated in LB medium containing 50 μg/mL kanamycin. The culture was grown overnight shaking at 37°C (180 rpm) and then used at a 1:100 dilution to inoculate an expression culture in LB medium. Protein expression was induced at OD600 = 0.8 by adding 1 mM IPTG and carried out at 37°C shaking (180 rpm), for 3 hr. Cells were harvested by centrifugation in a Beckmann centrifuge, rotor JLA 8.100 at 4500 rpm, for 20 min, at 4°C. The purification procedure was adapted from Bale et al., 2016. Pellets were resuspended in two pellet volumes of lysis buffer (25 mM Tris pH 8.0, 250 mM NaCl, 20 mM imidazole, 1 mM PMSF, 0.2 mM TCEP), sieved to remove clumps and supplemented with 1 mg/mL lysozyme and DNase. Cells were lysed by sonication on ice, and the lysate was clarified by centrifugation at 24,000 g, for 35 min, at 4°C. The clear lysate was incubated with Ni-beads (1 mL/L expression) for 1–2 hr, at 4°C under gentle rotation. Ni-beads with lysate were poured in a polypropylene (PP) column and the flow through (FT) was collected. Ni-beads were washed three times with 20 mL of lysis buffer followed by elution with 5 mL of elution buffer, containing 500 mM imidazole. The elution was immediately supplemented with 5 mM EDTA to prevent Ni-mediated aggregation of the sample. The buffer of the protein was then exchanged to dialysis buffer (25 mM Tris pH 8.0, 150 mM NaCl, 0.2 mM TCEP), at 4°C. After dialysis, the protein was transferred to a new tube and spun down for 10 min, at 5000 rpm, at 4°C, in order to remove any precipitation. The protein was concentrated using the KrosFlo with a 100 kDa cutoff membrane, which also helps removing any remaining unassembled capsid proteins. After concentrating down to 3–4 mL of volume, the sample was washed with 50 mL of fresh dialysis buffer using the continuous buffer exchange mode of the KrosFlo.

HBV capsids

Request a detailed protocol

A colony of E. coli AI cells containing the desired HBV plasmid was inoculated in TB medium containing 50 μg/mL ampicillin. The culture was grown overnight shaking at 37°C (180 rpm) and then used at a 1:100 dilution to inoculate an expression culture in LB medium. Protein expression was induced at OD600 = 0.8–1 by adding 0.02% arabinose and carried out at 20°C shaking (180 rpm) overnight. Cells were harvested by centrifugation in a Beckmann centrifuge, rotor JLA 8.100 at 4500 rpm, for 20 min, at 4°C. The purification procedure was adapted from Walker et al., 2011. Pellets were resuspended in one pellet volume of lysis buffer (25 mM Tris pH 7.5, 500 mM NaCl, 0.2 mM TCEP, 10 mM CHAPS) and lysed by sonication 3 × 30 s, on ice. The lysate was spun down in a Beckmann centrifuge rotor JA 25.50 at 10,000 rpm, for 10 min. The cleared supernatant was then loaded on a step gradient 10–60% sucrose obtained by mixing lysis and sucrose buffers (25 mM Tris pH 7.5, 500 mM NaCl, 0.2 mM TCEP, 10 mM CHAPS, 60% sucrose) in appropriate ratios and by carefully layering the different percentage buffers into ultracentrifugation tubes. The lysate was then subjected to ultra-centrifugation at 28,000 rpm, for 3.5 hr at 4°C. Fractions of 2 mL were collected by gravity, by puncturing the ultracentrifugation tube from the bottom. Fractions containing the capsids were pooled and concentrated using the KrosFlo with a 500 kDa cutoff membrane.

Large cargo maleimide labelling and characterisation

Request a detailed protocol

Purified capsids were labelled via maleimide chemistry to couple a fluorescent dye and NLS peptide to the exposed cysteines. The dye (AlexaFluor647 maleimide, Invitrogen) and NLS peptide (Maleimide-GGGGKTGRLESTPPKKKRKVEDSA, PSL Peptide Specialty Laboratories) were stored at −80°C and freshly resuspended in anhydrous DMSO. The capsids were incubated with different molar excesses of dye and NLS peptide according to the desired degree of labelling for 1–2 hr, at room temperature. A typical reaction was: 30–50 nmol of protein, 50 nmol of dye and 100–250 nmol of NLS peptide depending on the target #NLSs. The reaction was then quenched by adding 10 mM DTT and the protein was spun down at 10,000 rpm, for 10 min, at 4°C to remove any precipitation. The excess dye was removed by loading the capsid sample on a HiPrep Sephacryl 16/60 size exclusion column (GE Healthcare), using the appropriate buffer (for MS2 and MS2S37P: 10 mM Tris pH 7.5, 100 mM NaCl, 5 mM DTT; for I53-47: 25 mM Tris pH 8.0, 150 mM NaCl, 1 mM DTT and for HBV: 25 mM Tris pH 7.5, 500 mM NaCl, 0.2 mM TCEP, 10 mM CHAPS, 10% sucrose). Relevant fractions containing the labelled capsids were then pooled and concentrated using the KrosFlo. For long-term storage at −80°C, the sample was supplemented with 25% glycerol (30% sucrose for HBV) and either flash-frozen with liquid nitrogen or directly transferred to the −80°C freezer (for I53-47 capsids). The ratio of capsid monomers tagged with NLS peptide was quantified by the gel band ratio on a SDS PAGE gel with Coomassie staining, as the labelled monomers migrate differently due to their increased molecular weight. We note that the quantified #NLSs represents an average of the labelling reaction.

Electron microscopy

Request a detailed protocol

Capsid integrity was confirmed by imaging the samples with an electron microscope using negative staining. Carbon-coated 300 meshes Quantifoil Cu grids were glow-discharged for 10 s in a vacuum chamber. Then, a 3 μL drop of sample was adsorbed on a grid for 2 min, blotted with Whatman’s filter paper and washed three times with sample buffer, then three times with a solution of 2% uranyl acetate. Once the grids were dry, the sample was imaged using a Morgagni 268 microscope (FEI).

Dynamic light scattering

Request a detailed protocol

Dynamic light scattering (DLS) measurements to quantify the hydrodynamic radius of capsids and test for sample aggregation or disassembly were performed on a Zetasizer Nano (Malvern). Samples were diluted to a final concentration of 0.5 μM in filtered TB and spun down for 10 min at 10,000 g prior to each measurement. For each sample, at least 10 measurements were acquired, using a 12 μL quartz cuvette. Count rates per second were typically higher than 200 kcps, and the polydispersity index was below 0.2, indicating a monodisperse solution. Data were analysed using the Malvern software, using the Multiple Narrow Bands fitting algorithm and Refractive Index and Absorption settings for proteins.

Fluorescence correlation spectroscopy

Request a detailed protocol

Fluorescence correlation spectroscopy (FCS) was used to characterise the large cargoes and quantify their concentration and brightness (#dyes/capsid). FCS experiments were carried out on a custom-built multiparameter spectrometer confocal setup, equipped with a 60x water objective (NA = 1.27). The capsid samples were diluted in freshly filtered 1XTB and spun down for 10 min at 10.000g at 4°C prior to the start of the experiment. FCS measurements were carried out in 8-well Lab-Tek, which had been pre-incubated for 30 min with a solution of 1 mg/ml BSA to prevent sample sticking. For each sample, at least 10 FCS curves of 30 s each were acquired. Low power (1–5 μW) was used to avoid bleaching of the samples during their diffusion through the confocal volume. A calibration FCS measurement with a free dye solution was carried out every 2–3 samples to measure the structural parameter and confirm the stability of the setup. Data analysis was performed with SymphoTime software. Autocorrelation curves were computed for lag times between 0.0001 and 1000 ms and fitted with a diffusion model. Capsid brightness was calculated by dividing the measured particle brightness by the measured brightness of a calibration dye solution at the same laser power settings. Due to large aggregates in the absence of Importins, HBV was not probed by FCS.

Nuclear import assays

Request a detailed protocol

HeLa Kyoto cells were cultured at 37°C, 5% CO2 atmosphere in Dulbecco’s modified Eagle’s medium with 1 g/mL glucose (Gibco 31885023) supplemented with 1% penicillin-streptomycin (Sigma P0781), 1% L-Glutamine (Sigma G7513) and 9% FBS (Sigma F7524). The cells were regularly tested for mycoplasma contamination and found to be mycoplasma-negative. The cells were passaged every 2–3 days up to maximum of 15–17 passages. Cells were seeded 1 or 2 days prior the experiment at low density (10,000–12,000 cells per well) in a glass-bottom eight-well Lab-Tek II chambered coverglass (Thermo Scientific Nunc, 155383).

Cells for transport assays were stained with 100 nM MitoTracker green (Invitrogen, M7514) in growth medium for 30 min, at 37°C, 5% CO2. For nuclear staining, cells were rinsed once with PBS and incubated for 10 min, at room temperature with 20 nM Hoechst 33342 (Sigma, B2261).

Cells were then washed once with transport buffer (1XTB: 20 mM Hepes, 110 mM KOAc, 5 mM NaOAc, 2 mM MgOAc, 1 mM EGTA, pH 7.3 adjusted with KOH) and permeabilised by incubation for 10 min, at room temperature with digitonin (40 μg/mL). Cells were then washed 3 times with 1XTB supplied with 5 mg/mL PEG 6000 to avoid osmotic shock. After the final wash, excess buffer was removed and the transport mix was quickly added to the cells to start the experiment. The final transport mix was composed of 1 μM Importinα, 1 μM Importinß, 4 μM RanGDP, 2 μM NTF2, 2 mM GTP and 8 nM capsid cargo. In order to allow the import complex to form, the cargo was first pre-incubated with Importinß and Importinα on ice for at least 10 min, then the rest of the transport mix was added and the solution was spun down for 10 min at 10000 g to remove any aggregates. Each experiment was performed side-by-side with control cells incubated with fluorescently labelled 70 kDa Dextran (Sigma 53471) to confirm nuclear envelope intactness throughout the whole experiment. We note that, in order to exclude possible contaminations of free capsid monomers and/or fragments in our experiments, we applied stringent quality checks to each capsid prep and only used samples that had all of the following: uniform EM images, good quality monodisperse DLS, FCS parameters in line with monodisperse cargo of the right size.

Microinjection in starfish oocytes

Request a detailed protocol

Starfish (Patiria pectinifera) were kindly provided by Kazoyushi Chiba (Ochanomizu University, Tokyo, Japan) and kept at 16°C in seawater aquariums at MPI-BPC’s marine facilities. Oocytes were extracted from the animals fresh for each experiment as described earlier (Lénárt et al., 2003). Fluorescent proteins were injected using microneedles, as described previously (Borrego-Pinto et al., 2016; Jaffe and Terasaki, 2004).

Confocal fluorescence microscopy

Request a detailed protocol

Time-lapse confocal imaging of nuclear import was performed on an Olympus FLUOVIEW FV3000 scanning confocal microscope, using a 40x air objective (NA = 0.95). An automated multi-position acquisition was carried out, where 12 different regions (typically containing 10 cells each) were imaged in two different wells. Three channels were recorded at each time step, using the 405 nm (Hoechst), 488 nm (Mitotracker) and 640 nm (cargo) laser lines for excitation. Images were acquired every 2 min for 80–90 min, using continuous autofocusing with Z-drift compensation to ensure imaging stability.

Image and data analysis

Request a detailed protocol

Results of the time-lapse import experiments were analysed with a custom-written Fiji (Schindelin et al., 2012) script (Source code 1). The Hoechst and Mitotracker channels were used to generate reference masks for the nucleus, nuclear envelope and cytoplasm at each time point. Briefly, the two images were pre-processed with Gaussian blur to aid in area segmentation, and then thresholded. The nuclear mask was eroded three times to remove contributions coming from the nuclear envelope, and the envelope mask was generated by subtracting the eroded mask from the non-eroded one. The final masks were then used to extract the average intensity of cargo signal in the different areas of interest. Final data analysis and plotting was performed in IgorPro (Wavemetrics). Fluorescence intensities were background-corrected, rescaled according to the capsid brightness and fitted to an inverse exponential function It=A+IMAX1-e-τ*t, with IMAX being the plateau value reached at infinity, τ the reaction constant with units 1/s and A an offset parameter.

Mathematical analysis of the data

Request a detailed protocol

The initial flux was estimated from the mono-exponential fit as J=τIMAX. Error bars in the initial flux show sample standard deviations across the 12 imaged regions. For comparison with the theoretical predictions, where the flux saturates to a maximal value Jmax/(kONc)=1aRan  reached as N, we normalised all the flux measurements by the maximal observed value of the flux across all the technical replicate (that was still within 95% confidence interval of the mean value). Changing the normalisation value of the flux does not qualitatively change the conclusions of the model; however, it may cause a slight increase to our FR values and a slight decrease to our ϵ values. FR and ϵ values were obtained using a least-squares fit implemented in Python. Plots of ΔG and overlays of our fits onto the initial flux were also performed using Python.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76

Decision letter

  1. Robert H Singer
    Reviewing Editor; Albert Einstein College of Medicine, United States
  2. Suzanne R Pfeffer
    Senior Editor; Stanford University School of Medicine, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The requirements for nuclear transport can be recapitulated by a simple two-parameter biophysical model that correlates the import flux with the energetics of cargo transport through the nuclear pore complex. Together, the results reveal key molecular determinants of large cargo nuclear import in cells.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Molecular determinants of large cargo transport into the nucleus" for consideration by eLife. Your article has been reviewed by a Reviewing Editor and a Senior Editor, a Reviewing Editor, and two reviewers. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

While the reviewers felt the work had merit, they had several concerns about the interpretation of the data, the limitations of the experiments and the relevance to living cells. Their concerns might be addressed by further experiments but in its present state we feel the manuscript has too many issues to address within a few months. Hopefully their comments will be helpful in improving your manuscript. When you have addressed the concerns, we could consider it for eLife, as a new manuscript.

Reviewer #1:

Compared with typical/average-sized nuclear transport cargos, there are little data on the transport rates of large cargos by nuclear pore complexes (NPCs). The authors examined the suitability and transport rates of a series of 5 viral particle capsids of different sizes and with different numbers of NLSs. They conclude that 10 or more NLSs are required for import for the cargos examined, and that the number of NLSs per cargo volume is a better variable than NLSs per cargo surface area for predicting the amount of nuclear uptake. While these are interesting and potentially useful quantitative results, there are significant issues with the results, interpretation and details provided, which tempers my enthusiasm.

Of the 5 viral particle capsids described, the authors had technical difficulties with the two largest, and therefore these were not included in the analysis. Thus, their abstract is misleading as they do not report kinetics for the size range of 17-36 nm, but rather 17-27 nm. The problems with the two largest capsids should be moved to a supplementary section so as not to distract from the main work. In addition, the MS2 cargo is so poorly imported that it does not make sense to use it to draw major conclusions. For example, the slope for MS2 in Figure 5A is so flat that it is impossible to reliably conclude a minimum number of NLSs. The remaining cargos are not well behaved in terms of transport kinetics, as described in more detail below.

The NLSs are randomly attached to the capsid surfaces, making the resultant populations heterogeneous. The quantified number of NLSs is an average, and gel analysis is semi-quantitative. Some discussion of their expected errors is warranted.

The only biologically relevant capsid (HBV) – i.e., one that is imported into the nucleus – is not included in their analysis. Physiologically, this capsid disassembles in the nuclear basket. They have deleted the authentic NLS for their experiments. Thus, the biological implications are limited.

In Figure 1D, are the curves actual data or fits? FCS and DLS signals will be dominated by the large particles, yet free dye and/or labeled capsid monomers can significantly influence the import curves – are these responsible for the non-zero ordinate intercepts in Figure 3—figure supplement 1? Can the labeled capsids be separated from monomers and free dye by size exclusion chromatography?

The authors discount the importance of surface properties at numerous locations throughout the text. But they have not actually tested this, and surface properties are in fact surprisingly important, as multiple studies have shown – changing a few residues or adding fluorescent dyes can dramatically change the import properties of cargos. In fact, I would not be surprised if varying the number of dyes on their cargos would alter the slopes of the plots in Figure 5, or some of the scatter in these plots arises from the dye:cargo:NLS ratio. Minimally, they should tone down their discussion arguing against a minimal influence of surface properties.

While the authors limit the fitting "to the first 40 minutes to extract more accurate kinetics", the opposite is in fact true. Accurate fitting of exponential kinetics requires knowing the asymptotic limit, which is not the case for numerous curves in Figure 3 -figure supplement 1. Also, initial time points in these curves vary widely – this is not expected or discussed.

For 80 min time points, the authors should really consider including CAS, RanGAP and RanBP1 to maintain complete recycling of transport factors.

"Normalized nuclear intensity" needs some explanation. Relative to what? Do these correspond to the same scale for different plots. What does an intensity of 1 signify? How does this relate to the intensity in Figure 1D? The efficiency of nuclear uptake of the different cargos varies widely, but this is not discussed.

The energetic discussion in the last paragraph has little meaning without an estimate of the entropic cost of displacing the permeability barrier.

Reviewer #2:

The manuscript by Paci and Lemke describes experiments addressing nuclear accumulation of large NLS-labeled cargoes. The effort is commendable and the use of modified viral capsids is admirably clever. However, I have some serious problems with the interpretation.

The experiments are based on permeabilized cell assays. These are standard in the field, for better or worse, but they suffer a generic problem in that the rest of the cell is washed away. In a live cell, the transport substrate of interest has to compete with the rest of the proteome for attentions of the transport receptors. This can have a dramatic effect on the transport kinetics.

Like most studies of nuclear accumulation, the analysis does not distinguish properly between permeability of the nuclear envelope and the saturating level of nuclear concentration. The latter is recognized as "robust nuclear import" but depends, quite obviously, on the RanGTP system. The assumption that monoexponential (first-order) kinetics measure permeability through the nuclear pores is simply not justified. The observed kinetics reflect the rate-limiting step, which may be Ran recharging with GTP or recycling to the cytoplasm. See Kim and Elbaum, 2013, and much earlier Smith et al., 2002.

Quantitative measurements of nuclear accumulation can be affected in addition by binding to structures within the nucleus, as suggested by the images in Figure 3 for MS2 with high NLS count. Each NLS adds a considerable amount of positive charge. This may well affect binding to nucleic acids when present in such high local concentration on the viral capsid, especially if DNA/RNA binding proteins are lost in the permeabilization.

The text deals with the level of nuclear accumulation ("endpoint" in Figure 5), but the graphs presented show the accumulation kinetics rather than the saturation as a function of #NLS. The time for half-saturation, (I(t) – A)/Imax = 1/2, is actually ln2/k, not ln2/Imax as written in the text (subsection “Image and data analysis”). Looking at the table in Supplementary file 1, the values for T_1/2 are listed equal to 1/2 * ln2/k. This has the correct units but I don't understand the factor of 1/2.

If the aim of the exercise is to study the degree of accumulation, i.e., Imax, then the proper parameter to measure is the saturating nuclear to cytoplasmic ratio N:C. The logarithm of this ratio is the chemical potential difference, which is the essential thermodynamic quantity. As presented, the data do not show the cytoplasmic intensity and the background correction that was applied is not described. Figure 2C shows a single example of the cytoplasmic intensity where the nuclear to cytoplasmic ratio saturates at about 10 (700 / 70 units on the graph).

Since the fluorescence external to the cells coming from titrated cargo substrates should equilibrate with the fluorescence in the cytoplasm, I looked to see if this might be included in the fitting parameter A. It was not clear whether A is the background correction itself or a fit after the correction is applied. In any case A cannot represent the fluorescence from free cargo. According to the text these are introduced at a constant 8 nM concentration, but the values listed in the supplementary file vary widely, even for a given class of cargo. Why should they vary so widely? Presumably these values are corrected by the same factor as Imax for the substrate brightness. If they are not corrected, shouldn't the capsids with fewer NLS appear brighter, so with larger A? In some cases A is a very large fraction of Imax, leaving little dynamic range for the measurement itself. (Compare I53-47 with 15, 18, and 22 NLS.) In principle the black level to subtract is that of the confocal microscope with the laser blocked, and the fluorescence in the surrounding medium should match that measured in the permeabilized cytoplasm. If the cells are auto-fluorescent in the measurement channel then some additional correction will be required, but it should be specified clearly.

A few relatively technical points:

Why was the labeling with fluorescent dye and NLS done both on cysteine? The proteins could have been labeled first on lysine and then with NLS on the cysteine. The problem is that the molecular weight of the dye is almost half that of the peptide. Is a control available to show that the dye labeling really has no effect on the gel mobility? Figure 1—figure supplement 1 shows both Coomassie and fluorescence in the "unsuccessful" labeling of I53-50. For clarity, the main figure should also show the fluorescence in the successful case.

I did not understand the toy model in subsection “Global quantitative analysis of nuclear import in relation to cargo size and #NLSs”. The binding energy of NTRs to the cargo does not assist in directional translocation, nor is it transferred to displacing the FG repeats. That depends on interactions of NTRs with FG motifs. Crowding in the nuclear pore as shown in Figure 5 is interesting and might relate to kinetics, but not to the saturating concentration ("endpoint").

Nuclear export is not just the inverse of import. See Kim and Elbaum, 2013. There is a fundamental difference between exchange of RanGTP, a reversible reaction in "import", and physiologically irreversible GTP hydrolysis, which is coupled to translocation in "export".

The manuscript is long for a short report, about 3500 words in the main text alone.

Hoping to end on a constructive note, I have to apologize for being such an ornery reviewer here. I do quite like the experiment and I believe the data hold some new truths to be discovered. Wherever the work is ultimately published, I would like very much to see the nuclear accumulation presented as the nuclear to cytoplasmic ratio. This will normalize inherently for substrate brightness and avoid potential inconsistencies carried in by numbers from other measurements, imprecise dilutions, protein losses in aggregation, etc. Surely the data are available without requiring any further experiments. I am sure they could be reanalysed easily, avoiding confusion between kinetics and saturation. Plotting the ratio will clarify whether the additional number of NLS indeed influence the kinetics and saturation as suggested. There might be surprises in store.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Molecular determinants of large cargo transport into the nucleus" for consideration by eLife. Your article has been reviewed by the original reviewer, and the evaluation has been overseen by a Reviewing Editor and Suzanne Pfeffer as the Senior Editor. The reviewer has opted to remain anonymous.

The Editors have discussed the review with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

One of the original reviewers feels the manuscript has been improved but has some issues with the interpretation of the data, and the model. Specifically, the reviewer states "Particular attention must be made to predictions of the model, and interpretations in the context of this model." This reviewer has been thorough in the evaluation, so we feel the comments may likely be helpful in improving the manuscript further.

Because the concerns can be answered without additional data, but only require revisions to the manuscript, or explanations for the reviewer, we opt to send it back to you to address these comments.

Reviewer 1:

This revised manuscript has been substantially improved by tightening up the discussion and presentation to focus on the main story, and with the addition of a mathematical model. However, I do have some concerns about the revised manuscript, listed below in order of importance. While some of these points address accuracy and a logical consistency, other portions are intended to promote a more nuanced and informative picture. Particular attention must be made to predictions of the model, and interpretations in the context of this model.

1) Figure 5B – The model impressively explains the values in the graph. However, all of the ∆G values are positive, suggesting that binding to the permeability barrier is unfavorable. Nonetheless, nuclear rimming is clearly seen during the import experiments, indicating that interaction with the pore is favorable – more favorable than being in the cytoplasmic compartment. This indicates that the NPC is a thermodynamic sink. The data thus seem incongruent with the model, which only postulates an energy barrier. The model in Figure 5—figure supplement 4 is reminiscent of the vestibule model of Tu et al., 2013), yet here too, none of the ∆G values are negative (which was the case in Tu et al.,). Please discuss.

2) They cite four references for the initial flux equation (2, 33-35). I cannot find the equation they use in these references. In fact, two of them describe flux in terms of a constant multiplied by a concentration difference, which seems inconsistent with their equation. More discussion is necessary to elucidate where the model comes from.

3) If I understand the methods correctly, the NLSs and dyes were simultaneously mixed with the capsids. They discuss tuning the NLS/capsid ratio, and this is ultimately determined via a gel shift assay. But what about the number of dyes per capsid? It seems like they have brightness data from FCS experiments, and this should be reported. Do the number of dyes vary inversely with the number of NLSs? They continue to minimize the role of surface properties, yet a few extra dye molecules were shown by Tu et al., to dramatically affect the permeability properties of the cargo. I do not consider it safe to assume that the number of dye molecules does not influence the particle's interaction strength with the NPC. Moreover, they state that F(R) scales with the radius, yet the values for F(R) that they obtain are all essentially the same, which would be consistent with different surface properties of the different diameter capsids. Stating this does not diminish their results.

4) The epsilon values are surprisingly small. For the cargo of Tu et al., this would predict a very small interaction strength of the fully decorated cargo, and even smaller for a single NTR-bound cargo, which nonetheless still clearly binds to the pore. Note that the size (volume occupied) of β-galactosidase is less than MS2(S37P) by a similar ratio that the MS2(S37P) size is less than I53-47. It would be quite surprising indeed if the substantial behavioral differences of the β-galactosidase and MS2(S37P) cargos can be ascribed to the size and shape differences alone. It seems that surface properties must at least somewhat contribute to the observed differences.

5) Discussion section – I do not understand these surface coverage calculations. For maximum NLSs of 38, 35, and 98 for MS2(S37P), I53-47, and MS2, I get 84%, 42%, and 85% surface coverage assuming 20 nm2/β. This does not include Importin α. How much do the diameters increase assuming a full coat of Importins α and β? This is expected to be significant. How does this increased diameter compare with the size of the channel? Is there any experimental evidence that all NLSs on the capsids are bound to NTRs? Taking into account that concentrations and the Kd (~40 nM, α for NLS) are similar, the NLSs on the MS2 capsid are only about 90% occupied, implying 77% surface coverage. While these changes may not materially change their interpretation, a more detailed discussion is necessary to build an accurate picture and to build confidence in the conclusions. Other potential complications: (1) is it possible geometrically for all NTRs on a capsid to be bound to FG repeats? Figure 5A suggests that this may not be possible; and (2) can multiple capsids simultaneously bind to a single pore? Excess cargo, slow import and nuclear rimming suggest this possibility. Would this affect interpretation?

6) It is unclear whether there is any meaning behind the A values. These are highly variable, and I don't know what to make of them. In principle, A could reflect the accumulation of the cargos on the nuclear envelope, but as this arises from an extrapolation to zero time, it seems like this should in fact be zero, or at least some reasonably explained value. One possibility is that import rate could be dependent on the amount of accumulated cargo at the pores, i.e., a release rate, as entrance into the NPCs appears really fast.

7) The data on negatively charged linkers is inconclusive at best, as they are highly scattered. Their conclusions should be toned down.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Molecular determinants of large cargo transport into the nucleus" for further consideration by eLife. Your revised article has been evaluated by Suzanne Pfeffer (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

One reviewer feels the manuscript is substantially improved but there remains an outstanding issue that has not been corrected in the revision. The reviewer feels that Figure 5—figure supplement 4 needs to be clarified as described below. Additional minor comments directed at improving the manuscript are included as well. Please send a revised manuscript that addresses these comments sufficiently that it may not need to go back to this reviewer.

This revised manuscript has been substantially improved, with a much more balanced and informed discussion. All of my major concerns have been adequately addressed, with the exception of one item, the model in Figure 5—figure supplement 4. The figure itself is confusing/unclear, and I do not understand the basis behind building the model the way they did. Specific concerns for this figure are as follows:

1) What is the y-axis in the top panel of 'A'? This should be marked. My guess is that this is some measure of 'FG-Nup density' – are there any relevant units?

2) The dimensions of L1 and L0 do not reflect the values in the caption. Consequently, the diagram is misleading. The Greek letter is inconsistent with the caption. The vestibule region is not marked.

3) It is unclear why a transition region (L1) is included between the vestibule and L0. Comparing the top and bottom panels in A, it appears that the vestibule is equivalent to the cytoplasm. This does not make sense.

4) For L1 = 30 nm and L0 = 5 nm, the first impression is that the barrier gate is biased toward the nucleoplasmic side. Is this the intention? Such a model would be consistent with the nucleoplasmic gate hypothesized by the Weis group, and, if so, should be mentioned. Alternatively, are both the cytoplasmic and nucleoplasmic L1 regions both 30 nm? This would place the barrier in the center, but very narrow. It doesn't make much sense for a 'transition region' to be 6 times the width of the main barrier, so some discussion is needed here.

5) It is unclear why the ΔG for the L1 region changes substantially for the different viral particles, yet the ΔG for the L0 region changes minimally. It seems that the ΔG for the more dense FG nup environment would be more sensitive to particle size. An older hypothesis suggested dense clouds on the nucleoplasmic and cytoplasmic sides, but significantly lower density within the center. Is this being considered here?

6) The authors are correct in their rebuttal that only a portion of the NPC needs to contain a region where the interaction free energy is negative, in order to be consistent with the experimental observation of rimming. However, none of the regions illustrated in Figure 5—figure supplement 4 have negative ΔG. There is a dashed region that is apparently of negative free energy, but what this is remains unclear (point 3), and it is not clear if this energy is included in any way in their fit to the data.

7) In the lower panel of B, the green curve fit approximates the data very poorly, but does much better in the upper panel. Something seems amiss here.

https://doi.org/10.7554/eLife.55963.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Major revisions summary:

We thank both reviewers for their feedback on our work. Following up on all reviewers’ suggestions, we have substantially reworked the manuscript and expanded several points with new supporting data (both experimental and theoretical). We summarize below the main aspects before delving into a more specific point-by-point discussion of the individual reviewer comments:

1) We have completely changed the way we analyze the data, focusing on a more robust analysis of the measured kinetic parameters (completely revised Figure 5, and Supplementary file 2 and Supplementary file 3). This analysis has been performed in a new collaboration with the theoretical group of Dr. Anton Zilman at the University of Toronto. We now present a minimal biophysical model of large cargo transport through the NPC that only incorporates the most salient features of the energetics and the kinetics of transport. The model describes our experimental data very well and enables us to make inference regarding the major determinants of capsid transport in terms of their size and NLS number. We are indebted to the reviewers for encouraging us to look for a deeper interpretation of our data, resulting in a much improved paper that is now a well rounded blend of experiment and theory.

2) We have collaborated with the group of Dr. Peter Lenart at the MPI for Biophysical Chemistry Goettingen to perform injection experiments into live starfish oocytes (new Figure 2—figure supplement 2). All results agreed qualitatively with our permeabilised cell data. In addition, we want to note the merits of permeabilised cell work, which become evident when considering both experiments together: it enables us to put our results into a much larger context of available literature, as this is still the most frequent reported transport assay. Notably, the permeabilized cell assays also provided better access to the quantification of the biophysical parameters.

3) In addition, we reinforce our previous discussion of the role of the surface charges of the capsids with new, direct experimental data using capsids with modified surface charges (revised Figure 5C).

Reviewer #1:

Compared with typical/average-sized nuclear transport cargos, there are little data on the transport rates of large cargos by nuclear pore complexes (NPCs). The authors examined the suitability and transport rates of a series of 5 viral particle capsids of different sizes and with different numbers of NLSs. They conclude that 10 or more NLSs are required for import for the cargos examined, and that the number of NLSs per cargo volume is a better variable than NLSs per cargo surface area for predicting the amount of nuclear uptake. While these are interesting and potentially useful quantitative results, there are significant issues with the results, interpretation and details provided, which tempers my enthusiasm.

Of the 5 viral particle capsids described, the authors had technical difficulties with the two largest, and therefore these were not included in the analysis. Thus, their abstract is misleading as they do not report kinetics for the size range of 17-36 nm, but rather 17-27 nm. The problems with the two largest capsids should be moved to a supplementary section so as not to distract from the main work. In addition, the MS2 cargo is so poorly imported that it does not make sense to use it to draw major conclusions. For example, the slope for MS2 in Figure 5A is so flat that it is impossible to reliably conclude a minimum number of NLSs. The remaining cargos are not well behaved in terms of transport kinetics, as described in more detail below.

We thank the reviewer for the feedback but respectfully disagree with a few points: While the MS2 capsid shows much lower nuclear import total intensity when compared to the smaller MS2_S37P and I53-47 capsids, the signal we measure is still very well quantifiable, significant and robust (as can be seen from the confocal pictures in Figure 2 (the second image panel in Figure 2B was added to directly enable visualization of import for both the small and large MS2 capsids. While total brightness is different, the import signal significantly increases over time for both capsids) and Figure 3 our data show a consistent and robust trend of NLS-dependent import (see kinetic curves in Figure 3B, bottom right). Only because the data for the different capsids are all plotted together on one plot, this appears very low, relative to the much more efficient smaller capsids. We would argue that this rather unexpected large difference between cargoes efficiencies adds to the significance of our work. The new main text Figure 5B and C shows the data normalized. Here it is now clearly visible that import occurs also for the large MS2, which is also explained very well by our new simple biophysical model.

In respect to the “misleading abstract” comment, we argue that the results of our HBV experiments are not related to the technical difficulties but rather represent valid experimental data of a large cargo model (HBV with engineered NLS signals). “No import” is also a result, in particular if one puts this in the context of the widely used statement that the largest physiological import cargo known is HBV. We do agree, that for the “No import” conditions we only show cumulative indirect evidence, as we observed this for three completely different designed HBV capsids. However, as frequently with experimental studies, this refers to the tested conditions, and we discuss this now more rigorously.

The NLSs are randomly attached to the capsid surfaces, making the resultant populations heterogeneous. The quantified number of NLSs is an average, and gel analysis is semi-quantitative. Some discussion of their expected errors is warranted.

We agree with the reviewer that the gel quantification of the number of NLSs represents an average of the labelling reaction (that is expected to be Gaussian distributed). In the revised manuscript we have made sure to clarify this. It is a bulk experiment indeed, and we have no single molecule resolution to look beyond the ensemble average, which however does not make the result less quantitative. We see a very clear band shift when adding the NLS to the protein. As the NLS itself is small, the Coomassie band intensities where determined after scanning the Coomassie stained gel in a rigorous area analysis of the two different bands. Figure 1C shows clearly two nicely separated bands for NLS vs no NLS labelled capsids. The gels show a high purity, and the analysis is very robust and thus yields directly a clear well reproducible measurements of the labelling ratio. While Coomassie analysis might appear as a basic, not very sophisticated tool, its result are robust, and well quantifiable. Following the reviewer’s suggestion, we have expanded our discussion on the number of NLS determination in the Materials and methods section to explain ourselves better.

The only biologically relevant capsid (HBV) – i.e., one that is imported into the nucleus – is not included in their analysis. Physiologically, this capsid disassembles in the nuclear basket. They have deleted the authentic NLS for their experiments. Thus, the biological implications are limited.

Our goal in this work was to study general nuclear import properties using large model cargoes. Very little is known about the biophysics of large cargo transport and we think this is mainly due because most physiological cargoes do no enable to study large cargo import reliably. Taking HBV an example, the fact that the NLS is phosphorylation-dependent and potentially also packing-dependent, and import might actually require capsid disassembly, has left us with much ambiguity in the literature. In contrast, we feel that our reductionist approach provides a very good route to understanding fundamental principles. Additionally, we note that transport of HBV has been shown to depend on the same importinα/β that binds the NLS we use, so our simplified and more controllable system can still be insightful. We hope that reviewer 1 might also get persuaded by the view of reviewer 2, who specifically comments favourably on the choice of system by writing,” The effort is commendable and the use of modified viral capsids is admirably clever.”

In Figure 1D, are the curves actual data or fits? FCS and DLS signals will be dominated by the large particles, yet free dye and/or labeled capsid monomers can significantly influence the import curves – are these responsible for the non-zero ordinate intercepts in Figure 3—figure supplement 1? Can the labeled capsids be separated from monomers and free dye by size exclusion chromatography?

Figure 1D shows representative FCS data traces for the measured capsids, which we then fitted in Symphotime with a diffusion model. As common in FCS traces, very little noise is indeed visible at short correlation times. We agree with the reviewer concern about potential contaminations, and we have made a substantial effort to eliminate this concern. In short, we have performed the following:

1) Our labelling strategy makes it impossible that monomer import could be detected. As we label with dye and NLS separately, only larger structures can be both fluorescently labelled and NLS labelled. The larger the assembly gets, the more likely it would be to pick up contaminations by our analytical pipeline (FCS, DLS and EM).

2) Experimentally, the capsids are separated from the excess of unreacted dye by size exclusion chromatography and an additional step of tangential flow filtration. The sample undergoes extensive washing during this process, that should efficiently separate few kDa monomers and MDa-sized capsids and we also point out that we do not observe any capsid fragments in the final EM images. Further extensive characterization of each sample with EM, FCS and DLS lead us to conclude that the bulk of our preparation is formed by intact capsids. Please note, that FCS and DLS are sensitive to pick up substantial contaminations. As stated above (Figure 1D), also our FCS data are fit well with a single species.

3) For comparison, import of very small cargoes follows completely different kinetics (rather seconds, see e.g. Timney et el., 2006 and Ribbeck and Gorlich, 2001). This eliminates concerns of contamination of our kinetics on the relevant timescale. However, very large assemblies (like a half capsid) could easily contaminate our result. However, those we would easily detect using EM, and we were very stringent on our EM data quality, proceeding only when the EM showed very uniform nearly only perfect round structures.

We now discuss this better in the Materials and methods section and thank the reviewer for pointing out that this part was not made sufficiently clear.

The authors discount the importance of surface properties at numerous locations throughout the text. But they have not actually tested this, and surface properties are in fact surprisingly important, as multiple studies have shown – changing a few residues or adding fluorescent dyes can dramatically change the import properties of cargos. In fact, I would not be surprised if varying the number of dyes on their cargos would alter the slopes of the plots in Figure 5, or some of the scatter in these plots arises from the dye:cargo:NLS ratio. Minimally, they should tone down their discussion arguing against a minimal influence of surface properties.

While we agree with the reviewer that for most (especially passive) cargoes surface properties are key to nuclear import rates, our data point to a minor role for import of large cargoes bearing multiple NLSs. We have substantially extended our discussion about this point (see Discussion section in the main text), where we now elaborate on how much the Importin complex ends up covering (shielding) the capsid surface, even from a plain geometrical point of view.

More importantly, to address the reviewers points, we have performed new experiments, where we directly compare the MS2_S37P capsids labelled with NLS peptide and with a modified peptide bringing additional 5 charges per peptide: as can be seen in Figure 5C, the samples with/without additional charges behaved very similarly in the import assays (compare red dots and squares). This is consistent with our view that the highly decorated cargo surface is mostly shielded by bound impA and impB.

While the authors limit the fitting "to the first 40 minutes to extract more accurate kinetics", the opposite is in fact true. Accurate fitting of exponential kinetics requires knowing the asymptotic limit, which is not the case for numerous curves in Figure 3 —figure supplement 1. Also, initial time points in these curves vary widely – this is not expected or discussed.

We thank the referee for this point, and we apologize for causing this confusion. As we explain in our “major revisions summary box” at the top of this point-by-point, we now address this topic much more rigorously. In the revised manuscript, we focus our analysis on the initial influx rate, which (unlike the long term steady state) is determined by the kinetics of translocation of capsids through the pore. We find that this rate is more reliably extracted by fitting the data with a mono-exponential function (up to 80min) than a linear fit to the first few data points. We do this with an understanding that the long term steady state can be influenced by many additional factors, and make no conclusions from parameters of our monoexponential fits regarding the final steady state values – please also see the response to reviewer 2. To confirm that the mono-exponential function is a good approximation at short times, we have explored the option of a double exponential fit, and found that this does not increase the quality of the fit. Moreover, the fit is “sloppy” – whereby many combinations of the parameters produce fits of comparable quality (as reflected in the extremely high uncertainties in fit double exponential parameter values in Supplementary file 2). We further point out that using a double-exponential function to extract the initial rate gives essentially the same values as our mono-exponential fits (Supplementary file 2, Table 2). We hope these points are clear in the revised version and address the referee’s concerns.

Due to technical reasons, our first measurement point is 2 min delayed after the addition of the capsids, and therefore capsids with different properties accumulate to slightly different levels by that time – this could be a source of scatter of the “initial value”, together with minimal different levels of unspecific background signal. Thus, in the initial minutes, a potentially complex combination of capsid diffusing into the cells, non-specifically adhering to the coverslip or cell membranes, and being recruited to the nuclear envelope could give rise to different initial fluorescence. We have revised the text to clarify this point.

We also want to note, that we tested our new analysis pipeline for robustness of fitting 80 min vs 40 min, and gratifyingly the results are consistent across both fitting ranges.

For 80 min time points, the authors should really consider including CAS, RanGAP and RanBP1 to maintain complete recycling of transport factors.

Control experiments including CAS, excess GTP, excess Importinα and absence vs presence of an energy regenerating system in Figure 2—figure supplement 1 all show that the overall saturation behaviour is conserved also in those conditions. Most notably, we address this concern by a new experiment: we have now performed in vivo experiments where we microinjected live oocytes from starfish with the same representative capsid samples. The in vivo experiments agree qualitatively well with our permeabilised cell data in support with all our conclusions (see new Figure 2—figure supplement 2). Note, that the in vivo experiments areextremely low throughput and simply much harder to quantify (oil droplets, sticking, spatial heterogeneity etc., see Figure 2—figure supplement 2 for more details). The permeabilised cell assays (while still not easy when working with very large capsid cargoes) suffer much less from those problems.

"Normalized nuclear intensity" needs some explanation. Relative to what? Do these correspond to the same scale for different plots. What does an intensity of 1 signify? How does this relate to the intensity in Figure 1D? The efficiency of nuclear uptake of the different cargos varies widely, but this is not discussed.

We have expanded the discussion on the data analysis to better explain how we derive the corrected nuclear intensities to compare them between different capsids. Briefly, the raw intensity values (such as the traces in Figure 2D) are background-subtracted and corrected for the individual capsid brightness (estimated from FCS). These traces are then fitted and analysed. For display purposes, we have plotted the kinetic curved subtracted by offset parameter A in Figure 2.

The energetic discussion in the last paragraph has little meaning without an estimate of the entropic cost of displacing the permeability barrier.

In our revised manuscript we have now completely revised the biophysical analysis of the implications of our results. In collaboration with the Zilman lab (now included as co-authors on the paper), we have analysed the data through the lens of a minimal theoretical model based on previous experimental and theoretical results that incorporates basic physical insights into the kinetics and the energetics of capsid translocation. In particular, the model enables us to estimate the insertion free energy of a capsid into the NPC. The model captures our experimentally measured rates of import very well, as can be seen in Figure 5C.

Reviewer #2:

The manuscript by Paci and Lemke describes experiments addressing nuclear accumulation of large NLS-labeled cargoes. The effort is commendable and the use of modified viral capsids is admirably clever. However, I have some serious problems with the interpretation.

We thank the reviewer for his/her feedback on our work. Following up on all reviewers’ suggestions, we have substantially reworked the manuscript and expanded several points with new supporting data (both experimental and theoretical). We have summarized the main changes in our “major revisions summary” at the beginning of this point-by-point, and kindly ask this reviewer to read this one before delving into the detailed point-by-point discussion below.

The experiments are based on permeabilized cell assays. These are standard in the field, for better or worse, but they suffer a generic problem in that the rest of the cell is washed away. In a live cell, the transport substrate of interest has to compete with the rest of the proteome for attentions of the transport receptors. This can have a dramatic effect on the transport kinetics.

We agree with the reviewer that the permeabilized cell assay might not recapitulate all aspects of a living cell. To further support our results, we have now added an experimental dataset looking at the import of our large cargoes into microinjected living starfish oocytes (Figure 2—figure supplement 2). These experiments are extremely labour-intensive and low-throughput compared to the permeabilized cell assays, therefore we could not aim for the same level of characterization across tens of sample. On top of that, technical difficulties make them less quantitative then the very established permeabilized cell assay, which also serves here to compare our results to the standard in the field. However, we were able to test representative samples from the three main cargoes analysed in the paper and we were able confirm that their behaviour in intact live cells matches what we observe in permeabilized cell assays (see new Figure 2—figure supplement 2).

Like most studies of nuclear accumulation, the analysis does not distinguish properly between permeability of the nuclear envelope and the saturating level of nuclear concentration. The latter is recognized as "robust nuclear import" but depends, quite obviously, on the RanGTP system. The assumption that monoexponential (first-order) kinetics measure permeability through the nuclear pores is simply not justified. The observed kinetics reflect the rate-limiting step, which may be Ran recharging with GTP or recycling to the cytoplasm. See Kim and Elbaum, 2013, and much earlier Smith et al., 2002.

We thank the referee for this excellent point and we regret that it was not clear in the previous version. Indeed, the eventual steady state nuclear accumulation levels are affected by many factors such the availability of NTRs, Ran, Rcc1 and other factors. In particular, the theoretical work by Elbaum and Kim shows the possibility of non-monoexponential approach to the steady state, where the influx of cargoes is balanced by the outflux.

To address the reviewer’s concerns, our analysis in the revised manuscript is now focused on the initial rate of nuclear import (see new Figure 5), which is independent of the long time shape of the intensity curve; we stress that we rely on the mono-exponential fit only to extract the values of the initial import rates (these fits are shown in Figure 3—figure supplement 1). In this initial stage, the increase in nuclear fluorescence is expected to be solely due to the nuclear import of the capsids, while the kinetics of cargo dissociation from transport receptors and further steps in the Ran cycle should only have a minimal effect.

Furthermore, we have re-investigated the appropriateness of the mono-exponential fit for these short time kinetics, and find that it is an excellent approximation in the regime of interest. Double exponential fits do not provide an appreciable improvement in the fit quality; on the other hand, they are “sloppy” – whereby many combinations of the parameters produce fit of comparable quality (as reflected in the extremely high uncertainties in the parameter values for the double exponential fit (new Supplementary file 2), and do not produce very different estimates of the initial import rates (new Supplementary file 2, Table 2). Finally, we also note that empirically the mono-exponential import kinetics is a very common observation in the literature including in the experimental work of Kopito and Elbaum, 2007.

Quantitative measurements of nuclear accumulation can be affected in addition by binding to structures within the nucleus, as suggested by the images in Figure 3 for MS2 with high NLS count. Each NLS adds a considerable amount of positive charge. This may well affect binding to nucleic acids when present in such high local concentration on the viral capsid, especially if DNA/RNA binding proteins are lost in the permeabilization.

We agree with the reviewer that cargo sequestration by binding to nuclear structures may affect nuclear import rates; however, the new experimental dataset we have added employing charged NLS peptides (Figure 5) supports a picture where this contribution is not substantial. In addition, our improved analysis focuses on the initial import rate, which will be minimally affected by charge accumulation.

The text deals with the level of nuclear accumulation ("endpoint" in Figure 5), but the graphs presented show the accumulation kinetics rather than the saturation as a function of #NLS. The time for half-saturation, (I(t) – A)/Imax = 1/2, is actually ln2/k, not ln2/Imax as written in the text (subsection “Image and data analysis”). Looking at the table in Supplementary file 1, the values for T_1/2 are listed equal to 1/2 * ln2/k. This has the correct units but I don't understand the factor of 1/2.

The reviewer is correct, and we indeed did extract the half-saturation as ln2/k (see supplementary file 1). We have expanded our discussion to better clarify the information extracted from the data and, also following the recommendations from all reviewers, we have now the quantitative analysis and the combination with the biophysical model on the initial transport rate, which can be extrapolated from the fitting parameters and is much more robust, as discussed above.

If the aim of the exercise is to study the degree of accumulation, i.e., Imax, then the proper parameter to measure is the saturating nuclear to cytoplasmic ratio N:C. The logarithm of this ratio is the chemical potential difference, which is the essential thermodynamic quantity. As presented, the data do not show the cytoplasmic intensity and the background correction that was applied is not described. Figure 2C shows a single example of the cytoplasmic intensity where the nuclear to cytoplasmic ratio saturates at about 10 (700 / 70 units on the graph).

As mentioned above and following the reviewer suggestion, we now instead focus our analysis on the initial import rate, which isolates the step of translocation through the NPC from the efflux that can be affected by multiple factors. We agree with the reviewer that the N:C ratio could potentially capture the transport efficiency of the different cargoes but we have to note that this quantity is unfortunately not robust enough in the case of our sample to enable a thorough comparison of all samples. This is because some of the capsids have unspecific interactions with cytoplasmic structures and membranes (see for example Figure 3A, I53-47 in absence of NLS) that prevent us from extracting a clean and robust cytoplasmic signal from all samples.

Since the fluorescence external to the cells coming from titrated cargo substrates should equilibrate with the fluorescence in the cytoplasm, I looked to see if this might be included in the fitting parameter A. It was not clear whether A is the background correction itself or a fit after the correction is applied. In any case A cannot represent the fluorescence from free cargo. According to the text these are introduced at a constant 8 nM concentration, but the values listed in the supplementary file vary widely, even for a given class of cargo. Why should they vary so widely? Presumably these values are corrected by the same factor as Imax for the substrate brightness. If they are not corrected, shouldn't the capsids with fewer NLS appear brighter, so with larger A? In some cases, A is a very large fraction of Imax, leaving little dynamic range for the measurement itself. (Compare I53-47 with 15, 18, and 22 NLS.) In principle the black level to subtract is that of the confocal microscope with the laser blocked, and the fluorescence in the surrounding medium should match that measured in the permeabilized cytoplasm. If the cells are auto-fluorescent in the measurement channel then some additional correction will be required, but it should be specified clearly.

A is the offset parameter used in the fitting of the corrected kinetic traces. All traces have been corrected for cargo brightness (estimated from FCS) and PMT background of the microscope prior to fitting. Due to technical reasons, our first measurement point is 2 min delayed after the addition of the capsids, and therefore capsids with different properties accumulate to slightly different levels by that time. In those initial minutes, a potentially complex combination of capsid diffusing into the cells, non-specifically adhering to the coverslip or cellular membranes, and being recruited to the nuclear envelope could give rise to different initial fluorescence. We have revised the text to clarify this point.

A few relatively technical points:

Why was the labeling with fluorescent dye and NLS done both on cysteine? The proteins could have been labeled first on lysine and then with NLS on the cysteine. The problem is that the molecular weight of the dye is almost half that of the peptide. Is a control available to show that the dye labeling really has no effect on the gel mobility? Figure 1—figure supplement 1 shows both Coomassie and fluorescence in the "unsuccessful" labeling of I53-50. For clarity, the main figure should also show the fluorescence in the successful case.

Following the reviewer’s suggestion, we have added an image of the fluorescence signal in the labelled samples in Figure 1C. In control samples without NLS we have verified that there is no substantial effect on gel mobility (a clear single band is consistently observed). Furthermore, the NLS includes many Ks, and many of those are of functional relevance: the cysteine labelling strategy ensures that we do not impair functionality, and that at most we add one label per monomer subunit. Having two labelling sites on one monomer could introduce problems with contamination of kinetic signals due t very small capsids, (see also our reply to reviewer 1 regarding the point of contamination due to broken capsids).

I did not understand the toy model in subsection “Global quantitative analysis of nuclear import in relation to cargo size and #NLSs”. The binding energy of NTRs to the cargo does not assist in directional translocation, nor is it transferred to displacing the FG repeats. That depends on interactions of NTRs with FG motifs. Crowding in the nuclear pore as shown in Figure 5 is interesting and might relate to kinetics, but not to the saturating concentration ("endpoint").

In our revised manuscript we have completely reworked (see box at the beginning of this pbyp) this part by analysing the data using a minimal biophysical model based on previous examinations of the in vitro experiments of the thermodynamic permeability of the FG-Nup assemblies, and kinetics of transport through NPC mimicking nanopores and NPCs in permeabilized cells. The model allows us to examine the major contributions to the transport kinetics: free energy cost of insertion of a capsid the NPC and the energetic gain from interactions between NTRs and FG motifs. The model and the corresponding analysis are now described in subsection “Quantitative analysis of the nuclear import in relation to cargo size and #NLSs” and new Figure 5C.

Nuclear export is not just the inverse of import. See Kim and Elbaum, 2013. There is a fundamental difference between exchange of RanGTP, a reversible reaction in "import", and physiologically irreversible GTP hydrolysis, which is coupled to translocation in "export".

We have now revised the according part of the text to better reflect the differences between the import and the export mechanisms.

The manuscript is long for a short report, about 3500 words in the main text alone.

Following a further extension of our work with new data and the theoretical model, we have now changed the format to a research article.

Hoping to end on a constructive note, I have to apologize for being such an ornery reviewer here. I do quite like the experiment and I believe the data hold some new truths to be discovered. Wherever the work is ultimately published, I would like very much to see the nuclear accumulation presented as the nuclear to cytoplasmic ratio. This will normalize inherently for substrate brightness and avoid potential inconsistencies carried in by numbers from other measurements, imprecise dilutions, protein losses in aggregation, etc. Surely the data are available without requiring any further experiments. I am sure they could be reanalysed easily, avoiding confusion between kinetics and saturation. Plotting the ratio will clarify whether the additional number of NLS indeed influence the kinetics and saturation as suggested. There might be surprises in store.

We thank the reviewer for his constructive criticism and for adding this encouraging and motivating note. We agree with the reviewer, that a nuclear to cytoplasmic ratio analysis would have been nice, and this is what we originally tried. However, due to the different stickiness of the large capsid structures to cellular structures, the nuclear to cytoplasmic ratio does not yield robust results. In fact, as written in the main text, these capsid structures are already the best we identified from a much larger set of “big structures” we tried to get workable for these experiments.

We are convinced though, that our much improved analysis pipeline and completely revised fitting strategy deals well with all issues pointed out the by the reviewers, so that we hope the paper can now proceed to publication.

[Editors’ note: what follows is the authors’ response to the second round of review.]

One of the original reviewers feels the manuscript has been improved but has some issues with the interpretation of the data, and the model. Specifically, the reviewer states "Particular attention must be made to predictions of the model, and interpretations in the context of this model." This reviewer has been thorough in the evaluation, so we feel the comments may likely be helpful in improving the manuscript further.

Because the concerns can be answered without additional data, but only require revisions to the manuscript, or explanations for the reviewer, we opt to send it back to you to address these comments.

Reviewer 1:

This revised manuscript has been substantially improved by tightening up the discussion and presentation to focus on the main story, and with the addition of a mathematical model. However, I do have some concerns about the revised manuscript, listed below in order of importance. While some of these points address accuracy and a logical consistency, other portions are intended to promote a more nuanced and informative picture. Particular attention must be made to predictions of the model, and interpretations in the context of this model.

We thank the reviewer for his constructive feedback on our work.

1) Figure 5B – The model impressively explains the values in the graph. However, all of the ∆G values are positive, suggesting that binding to the permeability barrier is unfavorable. Nonetheless, nuclear rimming is clearly seen during the import experiments, indicating that interaction with the pore is favorable – more favorable than being in the cytoplasmic compartment. This indicates that the NPC is a thermodynamic sink. The data thus seem incongruent with the model, which only postulates an energy barrier. The model in Figure 5—figure supplement 4 is reminiscent of the vestibule model of Tu et al., 2013), yet here too, none of the ∆G values are negative (which was the case in Tu et al.). Please discuss.

This is an excellent question, and we regret this wasn’t explained clearly in the previous version. Nuclear rimming without substantial nuclear accumulation suggests that capsids interact favourably with some regions of the NPC, however not necessarily all regions, as otherwise nuclear accumulation would be visible.

In this respect, a variant of the vestibule picture provides one potential resolution of this apparent contradiction. In a spatially non-uniform effective potential inside the NPC, the averaged ΔG is dominated by regions of the potential that have a higher (more positive) value, while the cargoes still accumulate in the regions with low local ΔG. It may be a thermodynamic sink with locally negative free energy even while the global ΔG is positive.

In the case of the NPC, the most likely explanation for the “rimming” is a region of relatively low density – a diffuse “cloud” of FG nup density at the cytoplasmic side with very low penetration cost. This region has been noticed both in Tu et al., (as a “cytoplasmic vestibule”) and in Lowe et al., 2010 as cytoplasmic “docking” region.

To address the referee’s concern, we have now explicitly incorporated the cytoplasmic “docking/vestibule” area into the free energy profile, as shown in Figure 5—figure supplement 4. Note that the presence of such vestibule affects the flux through the pore minimally because the pore selectivity properties are mostly dictated by the barrier region.

We agree with the referee that the detailed comparison between different models is important. However, we emphasize that it is impossible to make unambiguous resolution of these question based on the bulk measurements only. The current minimal model that is warranted by the data is offering guidelines for the data interpretation, and we relegate more detailed models to the future work with single molecule tracking.

2) They cite four references for the initial flux equation (2, 33-35). I cannot find the equation they use in these references. In fact, two of them describe flux in terms of a constant multiplied by a concentration difference, which seems inconsistent with their equation. More discussion is necessary to elucidate where the model comes from.

We apologise for confusing references. We have updated these, and expanded the explanation of the formula in the main text. We also note that in the case of (idealized) non-interacting cargoes, the flux from the cytoplasm to the nucleus is J=kONcCa+eΔG, the reverse flux is J=kONcNa+eΔG (without taking into account Ran complications) and the overall flux through the pore is JJ=kON(cCcN)a+eΔG, so all these formulations are consistent with each other (in the regime of the approximation validity).

3) If I understand the methods correctly, the NLSs and dyes were simultaneously mixed with the capsids. They discuss tuning the NLS/capsid ratio, and this is ultimately determined via a gel shift assay. But what about the number of dyes per capsid? It seems like they have brightness data from FCS experiments, and this should be reported. Do the number of dyes vary inversely with the number of NLSs? They continue to minimize the role of surface properties, yet a few extra dye molecules were shown by Tu et al., to dramatically affect the permeability properties of the cargo. I do not consider it safe to assume that the number of dye molecules does not influence the particle's interaction strength with the NPC.

In our experimental pipeline, we quantify the number of dyes per capsid via FCS: following the reviewer suggestion we have now included the capsid brightness values to Table 1. Throughout our whole dataset we don’t find a strong correlation between NLS and dye number (R2=0.22). Similarly, we don’t see a correlation of the labelling ratio with the initial import flux (R2=0.14). In the case of the cargo used in the Tu et al. study, the ΒGal enzymatic activity (disaccharide hydrolase, potentially binding disaccharides in the NPC) also contributed to the unspecific import, requiring a total of 4 mutations to obtain a specific import cargo. It is possible that the drastic effect of adding 8 vs 16 dyes is somewhat specific to the ΒGal cargo and its surface, as the same group previously reported that the presence of 4 Alexa 647 dyes did not affect importinα-2XGFP-NLS interaction and cargo residence time (Sun et al., 2008). Our cargoes don’t present complications as the enzymatic activity of ΒGal and for the 0 NLS case we clearly don’t see any import even in presence of high dye numbers (up to 23 for the MS2S37P cargo, most comparable to the ΒGal) and after waiting much longer time (1.5-2 hours).

Furthermore, the shape of the ΒGal is very different compared to our spherical capsids (cylindrical with only 9 nm in diameter for the smaller axis), which in relative comparison to our study, render the ΒGal a small(er) cargo. Following the recommendation of the reviewer, see also below, we discuss now much more carefully in the discussion the surface shielding effect we speculate about for our large spherical structures due to Importin coverage. Those might not be relevant to ΒGal with only for Importins. Also, the topic of dye labelling is now discussed more carefully.

Moreover, they state that F(R) scales with the radius, yet the values for F(R) that they obtain are all essentially the same, which would be consistent with different surface properties of the different diameter capsids. Stating this does not diminish their results.

This is an excellent point, and we regret again that it was not clear in the previous version. The theoretical (and sensible) expectations are that the insertion cost should increase with the capsid size. However, the specific predictions have been obtained only for particle sizes much smaller than the ones studied in this paper, and only serve to guide the assumptions of the model.

One potential technical reason for the similarity of the observed values of F(R) is that the flux of the capsids without NTRs is so low that it is beyond the noise threshold of experimental detection, resulting in similar values of F(R). To control for this possibility, we re-did the analysis, now excluding the zero NLS point from the model fit – and the results of the analysis have not changed significantly. This is now included in the Figure 5—figure supplement 5.

Other potential reasons are that for such large capsids resulting in extreme compression of the FG nups, the insertion cost essentially saturates to its maximal value. We have modified the text accordingly to reflect this point.

Finally, we agree with the referee that at the moment we cannot exclude the possibility that at least at very low NTR coverage (where we basically cannot reliably detect any flux experimentally), surface effects might play a role. We have also added this to the revised discussion.

4) The epsilon values are surprisingly small. For the cargo of Tu et al., this would predict a very small interaction strength of the fully decorated cargo, and even smaller for a single NTR-bound cargo, which nonetheless still clearly binds to the pore. Note that the size (volume occupied) of β-galactosidase is less than MS2(S37P) by a similar ratio that the MS2(S37P) size is less than I53-47. It would be quite surprising indeed if the substantial behavioral differences of the β-galactosidase and MS2(S37P) cargos can be ascribed to the size and shape differences alone. It seems that surface properties must at least somewhat contribute to the observed differences.

We thank the referee for this question. We note that there are two separate issues here: (1) the model-independent experimental fact that even the smallest MS2(S37P) capsid requires significantly more bound NTRs than theΒGal cargo used in Tu et al., and (2) the small (5-6 times smaller than in Tu et al.,) values of the effective binding energy ϵ per NTR in the model that reflect the experimental fact 1.

It is hard to make detailed quantitative comparisons between different experimental platforms, given inherent uncertainties in experimental conditions. However, we feel that the apparent discrepancies between this work and Tu et al., can stem from the following factors.

The ΒGal construct of Tu et al., is a roughly cylinder shaped molecule of 18 nm at its longest axis, similar to the MS2S37P. However, its smaller axis is 9 nm with an accordingly lower cost of insertion, which can explain well why for this substrate 4 NLSs might be enough for import.

We would like to clarify that the ϵ values in Table 1 are physically realistic. The value of the effective energy ϵ per NTR is a product of the bare energy ϵ0 and the average volume fraction ϕ of FG motifs inside the pore. Using common estimates of the FG motif numbers in the pore of the order of ϕ0.01 , the bare interaction energy between one NTR and one FG motif is on the order of ϵ0312kBT (and up to 15kBT in the vestibule model), in accord with Tu et al. and other measurements NTR-FG interactions (for instance, by the Lim group). We have updated the manuscript to clarify this point further.

We cannot exclude at this point that the differences between the surface properties of our capsids and the ΒGal construct might contribute to the observed differences at low NTR coverage – and we now modified the text accordingly. It is also possible that at low NTR our values of ϵ could be closer to those of Tu et al. However, they are unlikely to be the dominant explanation at high NTR coverage, which is the predominant focus of this work. At high NTR coverage, potential sources for the discrepancy could be lower availability of the FG motifs due to extreme compression of the FG nups and/or competition between the FG motifs for the NTRs on the capsid.

We have clarified these issues in the text, and hope in the current form the manuscript addresses the reviewer’s concerns.

Finally, cargo shape has been observed to impact transport properties, with passive elongated cargoes being transported faster than spherical one of the same size (Mohr et al., 2009) and this is likely even stronger in the case of large cargoes, for example considering baculoviruses that need to specifically orient along their long axis to enter through the NPC.

5) Discussion section – I do not understand these surface coverage calculations. For maximum NLSs of 38, 35, and 98 for MS2(S37P), I53-47, and MS2, I get 84%, 42%, and 85% surface coverage assuming 20 nm2/β.

For the surface calculations we have taken the samples with highest NLS numbers, respectively 54, 44 and 98 for the three capsids (see Table 1). This explains the difference in the calculated values, and we have now clarified this better in the text.

This does not include Importin α. How much do the diameters increase assuming a full coat of Importins α and β? This is expected to be significant. How does this increased diameter compare with the size of the channel?

As the reviewer correctly points out, for simplicity we haven’t considered importin α explicitly in our back-of-the-envelope estimates of surface coverage. Taking the globular conformation of bound importin β, we could expect a radial increase of 9 nm in case of 100% cargo surface decoration by NTRs. The import complexes would then have an overall diameter of 34, 40 and 44 reaching a size comparable to the scaffold channel itself (approximately 40 nm).

However, we would like to stress that these size increase estimates are in our view not definitely meaningful. If one considers the increasing evidence that importins are an integral part of the permeability barrier (as for example reviewed in Lim et al., 2015) and FG-Nups/NTR complexes are highly dynamic (see Milles et al., 2015 and Hough et al., 2015) so FG-Nups can easily penetrate the cargo-importin complex. In such cases the additional layer due to Importin coverage does not effectively increase the volume of the transported cargo, as this layer is part of the transport machinery and readily penetrated by Nups which do interact with the Importins.

We therefore find that it is more appropriate to focus on the bare cargo size (which we could consider the FG-Nup excluded volume), and compare results accordingly.

Is there any experimental evidence that all NLSs on the capsids are bound to NTRs? Taking into account that concentrations and the Kd (~40 nM, α for NLS) are similar, the NLSs on the MS2 capsid are only about 90% occupied, implying 77% surface coverage. While these changes may not materially change their interpretation, a more detailed discussion is necessary to build an accurate picture and to build confidence in the conclusions. Other potential complications: (1) is it possible geometrically for all NTRs on a capsid to be bound to FG repeats? Figure 5A suggests that this may not be possible;

To ensure maximum binding of cargo and NTRs, we have always used excess of importins and pre-incubated them at high concentration with the cargo prior to start of the experiment. In DLS measurements of cargo + transport mix we detect two peaks corresponding to the capsids and small proteins in the mix, if we incubate capsid and importins we only detect the capsid peak, pointing to most importins being bound. Indeed, this does not necessarily ensure complete saturation of the NLSs, but as the reviewer him/herself points out, it still does not change the results, are we are always comparing the cargoes to each other in similar conditions. As for binding of NTRs to FG repeats, we do not see how the (perhaps oversimplified) scheme in Figure 5A suggests that it isn’t possible: FG-Nups are highly flexible and can extend easily several tens of nm, so they could easily enwrap the import cargo complex. For the sake of simplicity, we have represented this in a rather simplified way in the scheme.

Of course, it is indeed possible that, at a given moment, a given NTR is not bound to an FG motif at any of the bindings sites on the NTR. However, when we did a toy model calculation in our 2015 paper for only two binding sites, we saw that the probability of being unbound drops dramatically relative to a single site binding model. With Importins and Nups being highly multivalent, this becomes very unlikely. In that respect, ϵ is the average binding energy of an NTR to the FG domains, which already includes this possibility.

and (2) can multiple capsids simultaneously bind to a single pore? Excess cargo, slow import and nuclear rimming suggest this possibility. Would this affect interpretation?

Although we cannot exclude multiple capsids binding to a single pore with absolute certainty – and this has indeed been reported in EM of HBV capsids injected in Xenopus oocytes (Pante and Kann, 2002) – simple theoretical estimates show that these are likely lower probability events under our experimental conditions, at least during the initial period important for the calculation of the initial flux, which is the focus of our analysis.

6) It is unclear whether there is any meaning behind the A values. These are highly variable, and I don't know what to make of them. In principle, A could reflect the accumulation of the cargos on the nuclear envelope, but as this arises from an extrapolation to zero time, it seems like this should in fact be zero, or at least some reasonably explained value. One possibility is that import rate could be dependent on the amount of accumulated cargo at the pores, i.e., a release rate, as entrance into the NPCs appears really fast.

We attribute the differences in A values to limitations in the experimental design: (i) The start of the experiment is given by manual pipetting of the capsid solution to the well. We always image two wells (each 8 regions) in parallel using an automated data acquisition imaging procedure. The different capsids have to diffuse into the cells, and could be recruited to the nuclear envelope at a slightly different initial rate. Slightly different cargo accumulation in the first 2 minutes of the experiment prior to the start of the experiment is intrinsic to the resolution of the manual pipetting defining the start. (ii) In addition, we have observed different amounts of non-specific adhesion of the capsid samples to cellular membranes, as can be seen for example by comparing the confocal images of MS2S37P and I53-47 in Figure 3. Note that we do not extract any interpretations from this offset parameter, but simply use it to account for this complex mixture of effects giving rise to slightly different initial values. Finally, this parameter is quite sensitive to imperfections in the kinetic fit, which could be a further contributing factor to its different values.

7) The data on negatively charged linkers is inconclusive at best, as they are highly scattered. Their conclusions should be toned down.

Following the reviewer’s feedback, we have toned down our discussion of the experiments with charged linkers.

[Editors’ note: what follows is the authors’ response to the second round of review.]

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

One reviewer feels the manuscript is substantially improved but there remains an outstanding issue that has not been corrected in the revision. The reviewer feels that Figure 5—figure supplement 4 needs to be clarified as described below. Additional minor comments directed at improving the manuscript are included as well. Please send a revised manuscript that addresses these comments sufficiently that it may not need to go back to this reviewer.

This revised manuscript has been substantially improved, with a much more balanced and informed discussion. All of my major concerns have been adequately addressed, with the exception of one item, the model in Figure 5—figure supplement 4. The figure itself is confusing/unclear, and I do not understand the basis behind building the model the way they did. Specific concerns for this figure are as follows:

1) What is the y-axis in the top panel of 'A'? This should be marked. My guess is that this is some measure of 'FG-Nup density' – are there any relevant units?

We apologize that this had been overlooked. The relevant units are the volume fraction. We have re-drawn the diagram in Figure 5—figure supplement 4 and included this information.

2) The dimensions of L1 and L0 do not reflect the values in the caption. Consequently, the diagram is misleading. The Greek letter is inconsistent with the caption. The vestibule region is not marked.

We sincerely apologize for the confusion, and made sure that in the current version the labeling and the values of l1 and l0 are consistent throughout the text and the captions. We have also re-drawn the diagram in Figure 5—figure supplement 4 to match the fonts of ϕ everywhere, and clearly marked the vestibule region.

3) It is unclear why a transition region (L1) is included between the vestibule and L0. Comparing the top and bottom panels in A, it appears that the vestibule is equivalent to the cytoplasm. This does not make sense.

We apologize for the confusion. The “vestibule” indeed represent the low density cloud of the FG nups that extends into the cytoplasm beyond the NPC scaffold region. In that sense it can indeed be seen as a variant of the “vestibule” region of Tu et al., (2013) and/or “capture” region of Lowe et al., (2015). The region of the intermediate density l1 is simply the transition region between the vestibule and the high density barrier region within the NPC. We have edited the figure and the main text to explain these issues more clearly. Please also see the reply to the next comment.

4) For L1 = 30 nm and L0 = 5 nm, the first impression is that the barrier gate is biased toward the nucleoplasmic side. Is this the intention? Such a model would be consistent with the nucleoplasmic gate hypothesized by the Weis group, and, if so, should be mentioned. Alternatively, are both the cytoplasmic and nucleoplasmic L1 regions both 30 nm? This would place the barrier in the center, but very narrow. It doesn't make much sense for a 'transition region' to be 6 times the width of the main barrier, so some discussion is needed here.

We apologize again for the confusion in the values for l1 and l0 in the figure, which have now been corrected. l0=30nm is the central high density “barrier” region, while l1=5nm is the intermediate density transition region. There are two “peripheral” l1 regions on either side of the central barrier ( l0 region), therefore the barrier is not biased and is located at the center of the NPC.

5) It is unclear why the ΔG for the L1 region changes substantially for the different viral particles, yet the ΔG for the L0 region changes minimally. It seems that the ΔG for the more dense FG nup environment would be more sensitive to particle size. An older hypothesis suggested dense clouds on the nucleoplasmic and cytoplasmic sides, but significantly lower density within the center. Is this being considered here?

This is a great question. According to the model analysis, the costs of insertion into the central “barrier” region are relatively similar for all capsid sizes. By contrast, they are more variable in the l1 region. This is the main reason why the values of ΔG in the l0 region are more similar between the capsids of different sizes than in the l1 region.

One potential mechanistic explanation is that the insertion cost into the dense l0 region is probably close to saturation already even for the smallest capsid, while in the less dense l1 region, the difference between the capsids is more visible. However, we emphasize that the model depicted in the Figure 5—figure supplement 4 is just one possible model of the NPC informed by and consistent with the previously suggested ones. Full examination of different models and the inference of the actual density distribution of the FG nups inside the NPC is beyond the scope of the present paper, and will be studied in future work.

6) The authors are correct in their rebuttal that only a portion of the NPC needs to contain a region where the interaction free energy is negative, in order to be consistent with the experimental observation of rimming. However, none of the regions illustrated in Figure 5—figure supplement 4 have negative ΔG. There is a dashed region that is apparently of negative free energy, but what this is remains unclear (point 3), and it is not clear if this energy is included in any way in their fit to the data.

We thank the reviewer for this comment. Indeed, the free energy in the “vestibule” shown by the dashed line is negative and is responsible for the observed “rimming”. We have modified our schematic and the descriptions in the main text and figure caption to clarify this issue.

7) In the lower panel of B, the green curve fit approximates the data very poorly, but does much better in the upper panel. Something seems amiss here.

We thank the reviewer for this observation. This was indeed an oversight on our side, where the values of the parameter aRan were mixed up between the lower andthe upper panels. We sincerely apologize for the mistake and have corrected the figure; the quality of fit in both panels is now consistent.

https://doi.org/10.7554/eLife.55963.sa2

Article and author information

Author details

  1. Giulia Paci

    1. Biocentre, Johannes Gutenberg-University Mainz, Mainz, Germany
    2. Institute of Molecular Biology, Mainz, Germany
    3. European Molecular Biology Laboratory, Heidelberg, Germany
    Contribution
    Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0565-4356
  2. Tiantian Zheng

    Department of Physics, University of Toronto, Toronto, Canada
    Contribution
    Software, Formal analysis, Validation, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Joana Caria

    1. Biocentre, Johannes Gutenberg-University Mainz, Mainz, Germany
    2. Institute of Molecular Biology, Mainz, Germany
    3. European Molecular Biology Laboratory, Heidelberg, Germany
    Contribution
    Validation, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Anton Zilman

    1. Department of Physics, University of Toronto, Toronto, Canada
    2. Institute for Biomaterials and Biomedical Engineering (IBBME), University of Toronto, Toronto, Canada
    Contribution
    Resources, Supervision, Funding acquisition, Investigation, Methodology, Writing - review and editing
    For correspondence
    zilmana@physics.utoronto.ca
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8523-6703
  5. Edward A Lemke

    1. Biocentre, Johannes Gutenberg-University Mainz, Mainz, Germany
    2. Institute of Molecular Biology, Mainz, Germany
    3. European Molecular Biology Laboratory, Heidelberg, Germany
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Investigation, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    edlemke@uni-mainz.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0634-0503

Funding

Deutsche Forschungsgemeinschaft (240245660)

  • Giulia Paci
  • Joana Caria
  • Edward A Lemke

Natural Sciences and Engineering Research Council of Canada

  • Tiantian Zheng
  • Anton Zilman

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are particularly grateful to Niccolò Banterle for help with initial HBV experiments. We are very grateful to Peter Lenart and the light microscopy facility at the MPI for Biophysical Chemistry for their significant help with the oocyte work. We also thank Gemma Estrada Girona, Miao Yu and Sofya Mikhaleva for their contributions to this work. We thank SFB 1129 (Project number 240245660 funded by DFG, German Research Foundation) and the National Science and Engineering Research Council of Canada (NSERC) for generous funding. We thank the Baker lab for providing plasmids for their artificial capsid structure and the Nassal lab for HBV constructs. We thank Ulrich Schwarz for insightful discussion. We also thank the ALMF and EM facilities as well as the mechanical workshop at the EMBL.

Senior Editor

  1. Suzanne R Pfeffer, Stanford University School of Medicine, United States

Reviewing Editor

  1. Robert H Singer, Albert Einstein College of Medicine, United States

Publication history

  1. Received: February 12, 2020
  2. Accepted: June 18, 2020
  3. Version of Record published: July 21, 2020 (version 1)

Copyright

© 2020, Paci et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,445
    Page views
  • 195
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cell Biology
    2. Medicine
    LouJin Song et al.
    Research Article Updated

    The lymphatic vasculature is involved in the pathogenesis of acute cardiac injuries, but little is known about its role in chronic cardiac dysfunction. Here, we demonstrate that angiotensin II infusion induced cardiac inflammation and fibrosis at 1 week and caused cardiac dysfunction and impaired lymphatic transport at 6 weeks in mice, while co-administration of VEGFCc156s improved these parameters. To identify novel mechanisms underlying this protection, RNA sequencing analysis in distinct cell populations revealed that VEGFCc156s specifically modulated angiotensin II-induced inflammatory responses in cardiac and peripheral lymphatic endothelial cells. Furthermore, telemetry studies showed that while angiotensin II increased blood pressure acutely in all animals, VEGFCc156s-treated animals displayed a delayed systemic reduction in blood pressure independent of alterations in angiotensin II-mediated aortic stiffness. Overall, these results demonstrate that VEGFCc156s had a multifaceted therapeutic effect to prevent angiotensin II-induced cardiac dysfunction by improving cardiac lymphatic function, alleviating fibrosis and inflammation, and ameliorating hypertension.

    1. Cell Biology
    2. Stem Cells and Regenerative Medicine
    Na Li et al.
    Research Article

    Brain natriuretic peptide (BNP) treatment increases heart function and decreases heart dilation after myocardial infarction (MI). Here, we investigated whether part of the cardioprotective effect of BNP in infarcted hearts related to improved neovascularisation. Infarcted mice were treated with saline or BNP for 10 days. BNP treatment increased vascularisation and the number of endothelial cells in all areas of infarcted hearts. Endothelial cell lineage tracing showed that BNP directly stimulated the proliferation of resident endothelial cells via NPR-A binding and p38 MAP kinase activation. BNP also stimulated the proliferation of WT1+ epicardium-derived cells but only in the hypoxic area of infarcted hearts. Our results demonstrated that these immature cells have a natural capacity to differentiate into endothelial cells in infarcted hearts. BNP treatment increased their proliferation but not their differentiation capacity. We identified new roles for BNP that hold potential for new therapeutic strategies to improve recovery and clinical outcome after MI.