The HCMV Assembly Compartment is a dynamic Golgi-derived MTOC that controls nuclear rotation and virus spread

The AC is a dynamic structure that controls host cell behavior

To image the AC in living cells we fluorescently tagged two HCMV proteins; UL99 and UL32. UL99 localizes to trans-Golgi vesicles and labels the AC, and also binds virus particles (Moorman et al., 2010; Sanchez et al., 2000a; Sanchez et al., 2000b). UL32 localizes to the nucleus very early (1–2d.p.i.) but becomes predominantly cytoplasmic thereafter, associating with clathrin-containing compartments recruited to the AC. At later stages (~5–6d.p.i.) UL32 is detected in the nucleus again, incorporating into new virus particles that then bud into the AC (Moorman et al., 2010; Sampaio et al., 2005; Sanchez et al., 2000a). We engineered the HCMV clinical strain, TB40/E to express either UL99-eGFP or UL32-mCherry. To validate these viruses, we first infected primary normal human dermal fibroblasts (NHDFs) with TB40/E-UL99-eGFP. Cultures were fixed at different days post-infection (d.p.i.) and stained for the trans-Golgi marker, TGN46. In line with fixed imaging of laboratory adapted strains (Moorman et al., 2010), UL99-eGFP labeled the Golgi that reorganized into an AC between 2–5 d.p.i. (Figure S1A). Notably, 15% of infected cells imaged (65/431) contained two distinct or weakly inter-connected UL99-eGFP and TGN46-positive structures (Figure 1A). Staining for the viral glycoprotein gB confirmed that both structures contained other viral proteins and were likely nascent ACs. However, live cell imaging of UL99-eGFP-labeled structures revealed that in 100% of cases when two nascent ACs were present in a cell, they merged into one (Figure 1B; Movie S1A). As such, multiple ACs observed in a single cell in fixed images, here and evident in other studies (Buchkovich et al., 2010; Das et al., 2007), represent intermediate stages of merging. This seemed to ensure that only one AC resided in cells as infection progressed but also bore a striking resemblance to Golgi reorientation in migrating fibroblasts (Wu et al., 2016), discussed later.

We next co-infected NHDFs with TB40/E-UL99-eGFP and TB40/E-UL32-mCherry followed by two-color live cell imaging. In line with fixed images (Sanchez et al., 2000a), between 3–5 d.p.i. UL32-mCherry labeled cytoplasmic structures that gradually accumulated around the UL99-eGFP-labeled Golgi (Figure 1C; Movie S1B). As infection progressed, UL32-mCherry-positive domains also formed in the nucleus, followed by the appearance of small uniform structures in the AC and cytoplasm. These co-labeled with UL99-eGFP and their size was consistent with HCMV virions (Figure 1D–E; Movie S1B). This order series of events occurred in all cells imaged and is in line with predictions from prior studies using fixed samples. This established UL99-eGFP as an appropriate marker of the AC from its earliest stages of formation.

While imaging the AC we noted that infected cell nuclei rotated (Movie S1B). To visualize nuclei, we generated NHDFs stably expressing mCardinal fused to a Nuclear Localization Signal (mCardinal-NLS) and either mock-infected or infected them with TB40/E-UL99-eGFP. Live cell imaging over 3–5 d.p.i. confirmed nuclear rotation in infected cells (Movie S2). Notably, mCherry-NLS signal in infected cells became progressively weaker over time. This did not occur in uninfected cells, suggesting it was not due to photobleaching but instead reflected increased nuclear porosity in HCMV-infected cells (Buchkovich et al., 2010). To measure the extent of nuclear rotation we used nucleoli and the “nuclear pinch” as reference points. Results showed that ~80% (53/66) of infected cell nuclei exhibited persistent rotation >180° (Figure 1F). This net range underestimates the extent of nuclear rotation, as often nuclei reversed directionality or rotated ~720° (see below). By contrast, uninfected-cell nuclei exhibited no discernible rotation and translocated in the direction of cell motility.

Nuclear rotation in infected cells ceased prior to cell migration (Figure 1G; Movie S3A–B). Before migration, in 95% (57/60) cells imaged the AC was either already positioned at or moved toward the leading edge (between the nucleus and direction of migration) of cells. This demonstrated that the AC does not always remain at the nuclear “pinch”, and suggested it might also regulate cell migration later in infection. In these later stages the AC continued to exhibit merging behaviors. If the AC lost integrity it rapidly “regrouped” while if two ACs encountered one another, presumably upon cell fusion, they merged (Movie S3C–D). Although these later events arose from diverse, uncontrollable circumstances making them difficult to quantify, they illustrated the continued dynamic behavior of the AC throughout infection. Overall, AC merging, nuclear rotation and AC positioning at the leading edge all suggested the AC might function as an MTOC.

The HCMV AC acts as a Golgi-derived MTOC

Although MTs emanate from the AC (Sanchez et al., 2000a), the true nature of their nucleation and function remains poorly understood. We therefore stained fixed samples of mock or HCMV-infected NHDFs for two key MT nucleation factors. Imaging detected bright γ-tubulin and pericentrin puncta at centrosomes in uninfected cells and within ACs (Figure 2A). However, centrosomes in infected cells appeared “split” as shown in prior reports (Hertel and Mocarski, 2004). Moreover, γ-tubulin and pericentrin were broadly distributed throughout the AC in regions positive for TGN46 (Figure 2A and S1B). This was not observed in uninfected cells since the Golgi does not normally contain high levels of nucleating material (Sanders and Kaverina, 2015). Western Blot (WB) analysis of mock or infected cell lysates revealed that HCMV also increased γ-tubulin abundance around the time of AC formation (Figure 2B). This suggested HCMV increased the availability of nucleating material to potentially form new MTOCs within the AC.

To test this directly, we performed nocodazole washout assays to identify sites of MT nucleation. Nocodazole disrupts the Golgi in uninfected cells and the AC in infected cells (Chabin-Brion et al., 2001; Sanchez et al., 2000a). While this prevented us from imaging MT growth from an intact AC, this effect was advantageous as it spatially separated the centrosome from surrounding Golgi sites and avoided uncertainty over the origin of new MTs. Staining for TGN46, pericentrin and tyrosinated tubulin, which detects dynamic MTs, confirmed that nocodazole depolymerized MTs and dispersed Golgi and AC structures (Figure 2C). Quantitation revealed that new MTs almost exclusively originated from the centrosome in uninfected cells (Figure 2C–D). In infected cells, new MTs originated from Golgi-based sites distributed throughout the cell due to AC disruption. Notably, uninfected or infected NHDFs contained no detectable detyrosinated MTs beyond centrosome asters (Figure S1C). As such, detyrosination does not affect the use of tyrosinated tubulin antibody to image MTs in NHDFs. Confocal imaging and quantification of new MTs further demonstrated that centrosomes in uninfected cells nucleated MTs more efficiently than in infected cells (Figure 2E–F). This suggested that HCMV overturned the normal dominance of the centrosome to favor Golgi-based nucleation.

A characteristic of Golgi-derived MTs is their rapid acetylation (Chabin-Brion et al., 2001; Efimov et al., 2007; Rios et al., 2004; Sanders and Kaverina, 2015). We therefore stained uninfected or infected samples for acetylated tubulin and TGN46. Imaging and quantification revealed that beyond singular asters, uninfected NHDFs contained very few acetylated MTs (Figure S1C–D). However, HCMV induced the formation of acetylated MTs that co-localized with TGN46-positive regions of the AC (Figure 2G and Movie S4), suggesting Golgi domains nucleated these MT subsets. To test this, we again performed nocodazole washout assays staining for acetylated tubulin and TGN46. Although centrosomes stained brightly for acetylated tubulin they were associated with relatively few acetylated filaments, while abundant arrays of acetylated MTs emanated from TGN46-positive structures dispersed throughout the cell (Figure 2H). Cumulatively, these findings demonstrated that Golgi regions of the AC nucleated new MTs that were quickly acetylated.

Differential regulation and localization of EB proteins during HCMV infection

Formation of MTs at Golgi sites was recently shown to require EB proteins (Yang et al., 2017). Therefore, we tested whether HCMV influenced their abundance or localization. WB analysis over a timecourse of infection (Figure 3A–B) revealed that the abundance of EB1 and EB2 gradually declined in mock-infected NHDFs, likely as cells became quiescent, while levels of other proteins remained unchanged. However, in infected NHDFs all three EB proteins increased in abundance by 2–3 d.p.i. Similar increases in EB proteins were observed with independent HCMV strains, the timing of which correlated with differences in each strain’s replication kinetics (Figure S2A). Alongside protein increases, qRT-PCR revealed that HCMV increased the transcript abundance of each EB relative to mock-infected cells (Figure 3C). The relative increases in transcript and protein levels for EB1 and EB2 largely corresponded but increases in EB3 protein were greater than those of its transcript. This suggested that EB3 might be additionally regulated at post-transcriptional levels.

Increases in EB levels coincided with expression of the viral transcriptional activator, Immediate Early protein 2 (IE2) in all three HCMV strains tested (Figure 3A and S2A). To test if IE2 was involved in regulating EBs we suppressed IE2 expression in infected cells using siRNAs. HCMV IE1 and IE2 proteins are encoded by alternatively spliced transcripts of the major immediate early (MIE) gene (Mocarski et al., 2007). NHDFs were treated with control non-targeting or MIE (IE1/2) siRNAs prior to infection with HCMV. WB analysis showed that IE1/2 depletion prevented increases in EB1 and EB2 by HCMV (Figure 3D). IE1/2 siRNAs did not reduce EB abundance in uninfected cells (Figure S2B), demonstrating that these changes were not off-target effects. Moreover, IE1/2 depletion did not block EB3 increases (Figure 3D), supporting the idea that EB3 was regulated at additional levels. To test if IE2 was sufficient to increase EB levels, we transduced NHDFs with lentiviruses encoding GFP control or IE proteins and found that IE2, but not IE1, was sufficient to increase EB1 and EB2 abundance (Figure 3E). However, EB3 expression was not detectably affected. As such, independent siRNA and expression approaches supported the notion IE2 enhanced EB1 and EB2 accumulation, likely at the transcriptional level although changes in mRNA stability cannot be ruled out. Both approaches also suggested additional modes of regulation for EB3 were involved, beyond simply increasing EB3 transcript levels.

EB3 stability is controlled by Cyclin Dependent Kinase 1 (CDK1)-mediated phosphorylation (Ban et al., 2009), and HCMV activates CDK1 (Hertel et al., 2007; Sanchez et al., 2003). To test the contribution of CDKs to EB3 accumulation, NHDFs were mock infected or infected in the presence of CDK1 inhibitor, JNJ-7706621. WB analysis of time-courses showed that in cells infected with AD169, where EB3 increases begin 1–2d.p.i. (Figure S2A), JNJ-7706621 suppressed HCMV-induced CDK activation as determined by pan-CDK1 substrate antibody, and prevented increases in EB3 at 1 and 2 d.p.i. (Figure 3F). Although JNJ-7706621 modestly affected IE2 expression, siRNA approaches above showed that reduced IE2 was not sufficient to affect EB3 increases. As such, JNJ-7706621 demonstrated that CDK, or the combined action of IE2 and CDK were required to increase EB3. However, by 3d.p.i. both CDK substrate phosphorylation and increases in EB3 abundance became insensitive to JNJ-7706621, despite daily inhibitor replenishment. Notably, HCMV encodes a CDK homolog, pUL97 that is resistant to host CDK inhibitors (Hertel et al., 2007). As such, our findings establish that CDK activity regulates EB3 abundance during infection, but that the viral CDK homolog likely takes over from host kinases once infection is established. Overall, these findings established that HCMV uses multiple strategies to increase the abundance of EB proteins.

We also tested HCMV’s effects on EB protein localization. While EB1 and EB3 track MT plus-ends, the more divergent EB2 exhibits weaker plus-end-specific localization (Komarova et al., 2009). In line with this, EB1 and EB3 formed “comets” indicative of plus-end tracking on MTs in both uninfected and infected NHDFs, while EB2 was more broadly distributed along the MT lattice (Figure S2C–E). Notably, while EB1 comets localized throughout infected cells and near the AC, EB3 comets intensely concentrated around γ-tubulin-positive regions of the AC (Figure S2C–F). As such, EB1 and EB3 exhibited differences in both regulation and localization in infected cells, suggesting they might perform discrete functions.

EB1 and EB3 play distinct roles in HCMV infection

We next examined the role of EB proteins in HCMV infection. EB1 is required for some viruses to reach the nucleus and establish infection (Jovasevic et al., 2015; Sabo et al., 2013). Similar effects on early HCMV infection would confound our understanding of their roles in later processes. To avoid this, we exploited HCMV’s protracted lifecycle by reducing siRNA pre-treatment time such that the establishment of infection was unaffected. Using this approach, viral IE protein production was not affected despite efficient depletion of EBs, or by secondary dosing of siRNAs at 3d.p.i. (Figure 4A). The fact this approach did not affect the establishment of infection was further evidenced by the formation of ACs and stable MTs, shown later. Using this approach, we assessed effects of EB depletion on HCMV replication using independent siRNAs targeting EB1 or EB3. In cultures infected at low multiplicity of infection (MOI) for 12d, WB analysis of viral protein accumulation (Figure 4B) or imaging and measurement of plaque sizes (Figure 4C–D) revealed that both EB1 and EB3 were required for efficient HCMV spread. We also assayed effects of EB1 or EB3 depletion on the production of virus in single round high MOI infections. Titration of infectious virus produced at 7d.p.i. revealed a 100-fold and 14-fold reduction in supernatant virus, and 10-fold and 4-fold reduction in cell-associated virus in EB1- or EB3-depleted cells, respectively (Figure 4E). These findings suggested EB1 and EB3 might regulate HCMV replication by distinct mechanisms.

To determine whether EBs had different roles in infection, we tested how EB1 or EB3 depletion affected MT behavior. NHDFs expressing CLIP170-eGFP, a +TIP widely used to visualize MT growth, were treated with control or EB-targeting siRNAs followed by infection with HCMV. Imaging at 4.d.p.i. revealed that while EB3 depletion had no detectable effect on the number or behavior of CLIP170-eGFP comets, EB1 depletion caused a re-distribution of CLIP170 along the MT lattice (Figure 4F; Movies S5A). Line-scan analysis of CLIP170-eGFP signal intensity and distribution further confirmed its broader distribution along MTs in EB1-depleted cells (Figure 4G), suggesting the accumulation of CLIP170 at growing MTs was specifically regulated by EB1.

We next tested whether EBs were required for the formation of acetylated MTs by HCMV. Staining showed that EB1 depletion increased acetylated MTs, while EB3 depletion suppressed their formation compared to controls (Figure 5A). These effects were quantified scoring cells as 1) lacking acetylated MTs (none), 2) having acetylated MTs extending from the AC (medium), 3) having extensive acetylated arrays filling the cytoplasm (high). This quantification approach confirmed that EB1 depletion increased, while EB3 depletion decreased the extent of acetylated MT formation by HCMV (Figure 5B). EB1 and EB3 compete for binding to MT plus-ends (Komarova et al., 2009), which potentially explained their differential effects on the formation of acetylated MTs. In line with this, the staining intensity of EB3 comets at the AC increased in EB1-depleted cells compared with controls, while EB1 did not localize to the AC in EB3-depleted cells (Figure 5C). This suggested EB1 could not functionally substitute for EB3 in infected cells. Moreover, line-scan analysis showed that in controls, EB1 formed bright tracks ahead of EB3 (Figure 5D). EB3 formed longer and brighter comets upon EB1 depletion, while EB3 depletion resulted in brighter and longer EB1 tracks. This demonstrated that EB1 and EB3 compete for binding to MT plus-ends in HCMV-infected cells, and that EB1 depletion enhanced EB3 activity and localization the AC. The levels of acetylated MTs in each condition correlated with the extent of EB3 localization to the AC and MT plus-ends. Overall, this suggested that both EB1 and EB3 contributed to efficient infection, but did so through effects on distinct MT subpopulations.

Acetylated MTs control nuclear rotation and AC structure

We examined whether EB1 or EB3 influenced nuclear rotation or AC formation. NHDFs were treated with control or EB-targeting siRNAs followed by infection with TB40/E-UL99-eGFP. Live cell imaging over 3–5d.p.i together with nuclear displacement measurements revealed that in control or EB1 siRNA-treated samples ~86% (37/43) nuclei rotated >180° (Figure 5E; Movie S5B). These involved cases of >360° unidirectional rotations or 180° rotations with reversed directionality. By contrast, other than limited general displacement similar to mock-infected cells (Figure 1G), nuclear rotation could not be detected in EB3-depleted cells (Figure 5E). Using faster frame rate imaging, in control siRNA-treated cells UL99-eGFP-labeled ACs consisted of a dense core while smaller structures and uniform-sized particles, likely virions (Figure 1E), moved bi-directionally in the cytoplasm (Figure 5F; Movie S6). By contrast, in EB1-depleted cells the AC was diffuse and fewer motile organelles and particles could be detected in the cytoplasm. In EB3-depleted cells the AC also appeared disorganized but the cytoplasm contained large structures that resembled multi-vesicular bodies (MVBs). These results demonstrated that depletion of EB1 or EB3 resulted in distinct alterations to AC structure, while EB3 was also required for HCMV-induced nuclear rotation.

A small peptide targeting EB3 suppresses HCMV replication

Our data suggested EB3 might be a “druggable” target to suppress HCMV replication. Peptide aptamers containing SxIP motifs specifically bind EB1 or EB3 due to variations in their flanking amino acid sequence, and displace their interacting +TIPs (Lesniewska et al., 2014). We designed a myristoylated peptide (Myr-RSMKRSLIPRWIGNKR, dubbed HEBTRON) to target EB3 and first characterized it in vitro. Using Drug Affinity Responsive Target Stability (DARTS) assays wherein drug binding protects target proteins from proteolytic degradation by thermolysin, we found that HEBTRON protected purified recombinant EB3 (Figure S3A). This was specific to HEBTRON as an EB1-specific peptide did not prevent EB3 proteolysis (Figure S3A). Using isothermal titration calorimetry (ITC) to measure binding affinity further revealed an exothermic binding with the dissociation constant (K_D) of 537 ± 94.7 nM and enthalpy (ΔH) of −2.138 ± 0.1592 cal/mol (Figure 6A and S3B–C), indicating a strong and stable interaction of HEBTRON with EB3. The measured binding stoichiometry (N) of 0.47 ± 0.03 suggested formation of a 1:2 complex, in line with HEBTRON binding both EB3 subunits of dimers.

We next tested the effects of HEBTRON on EB3 dimer stability using nano Differential Scanning Fluorimetry (nanoDSF). In this assay, thermal unfolding of proteins is measured to identify melting temperature (Tm). Compared with EB3 alone, a 0.2°C peak shift was detected with HEBTRON present (Figure 6B). This suggested HEBTRON increased EB3 dimer stability. To independently test this, we used a FRET-based assay in which preassembled YFP-EB3 dimers were mixed with CFP-EB3 dimers. In solution, input homodimers dissociate and form mixed species of dimers over time, resulting in FRET signal when YFP/CFP tags are proximal within heterodimers (De Groot et al., 2010). In the absence of HEBTRON increasing FRET was observed over time with YFP/CFP-EB3 dimer formation (Figure 6C). However, HEBTRON blocked FRET signal indicating that it stabilized input EB3 dimers.

We next determined uptake of myristoylated HEBTRON by cultured NHDFs. Confocal time lapse imaging showed that 5’6-fluorescein(FITC)-conjugated HEBTRON initially labeled the plasma membrane but was observed uniformly distributed throughout the cytosol within 7–8 minutes, with t_1/2 uptake of 235.5 ± 1.4s (Figure 6D–E). We next tested whether HEBTRON affected MT behavior during infection. NHDFs expressing CLIP170-eGFP were treated with vehicle (DMSO) or 25µM HEBTRON followed by infection with HCMV. Time lapse imaging combined with line-scan analysis of CLIP170-eGFP intensity and distribution showed that HEBTRON had no detectable effect on CLIP170-eGFP behavior (Figure S3D). However, when fixed samples were stained for acetylated tubulin, imaging and quantification using approaches described above demonstrated that HEBTRON suppressed acetylated MT formation by HCMV (Figure 6F–G).

To understand how it functioned, we examined the effects of HEBTRON and two HEBTRON mutants containing amino acid substitutions in either one (MutN) or both (MutN-MutC) SLIP-flanking regions (Figure 6H). We first assessed their effects on plaque formation by TB40/E-GFP or AD169-GFP in low MOI spreading assays. Phase and fluorescence imaging at 12d.p.i. along with plaque size measurements showed that HEBTRON suppressed HCMV spread compared with DMSO control, while both HEBTRON mutants had no effect (Figure 6I–J). To further test its antiviral effects, NHDFs were treated with DMSO or 25µM HEBTRON before low MOI infection with an unrelated poxvirus, Vaccinia Virus (VacV). Imaging and measurements of plaque sizes formed by VacV expressing B5-GFP (Figure S3E–F) or WB analysis of VacV protein accumulation (Figure S4G) revealed HEBTRON had no significant effect on VacV spread. As such, HEBTRON exhibited specific antiviral effects toward HCMV that were dependent on guiding amino acids.

As published aptamers displace +TIPs from EB proteins (Lesniewska et al., 2014), we tested whether HEBTRON affected the localization of representative proteins known to utilize an SxIP motif to bind EB3; Tastin and CDK5RAP2 (Jiang et al., 2012). IF imaging of infected cells showed that HEBTRON did not affect EB3 localization to the AC compared with DMSO controls (Figure S3H). However, staining DMSO-treated cells revealed that Tastin localized in the same area as EB3, behind EB1 on MT plus-ends, and that HEBTRON displaced Tastin from MT plus-ends (Figure S4A–B). Notably, the second SxIP-interacting protein, CDK5RAP2 not only binds EB proteins but also stimulates the nucleating activity of γ-TuRC (Choi et al., 2010; Fong et al., 2009; Wang et al., 2010; Wu et al., 2016). CDK5RAP2 was present at centrosomes in uninfected NHDFs, but upon infection it localized throughout Golgi regions of the AC (Figure 7A). Imaging and measurements threshold brightness revealed that HEBTRON, but not the MutN-MutC form of HEBTRON, significantly reduced CDK5RAP2 recruitment to the AC (Figure 7B). HEBTRON did not decrease CDK5RAP2 or EB3 abundance (Figure 7C), demonstrating that HEBTRON specifically affected CDK5RAP2 enrichment in Golgi regions. Cumulatively, these data supported the notion that HEBTRON suppressed infection by interfering with EB3-mediated enrichment of factors such as CDK5RAP2 at non-centrosomal sites within the AC. This, combined with recent reports that EB proteins function in MT minus-end organization and nucleation at the Golgi (Yang et al., 2017), suggested that EB3 likely regulated the nucleation of MTs at the AC. To test this, we performed nocodazole washout assays on EB3-depleted or HEBTRON-treated cells infected with HCMV. Staining samples for α-tubulin, acetylated tubulin and TGN46 revealed that in control siRNA or DMSO-treated cells new MTs formed at Golgi regions dispersed throughout the cytoplasm, and these MTs were acetylated (Figure S4C–D). By contrast, formation of new MTs at Golgi fragments was suppressed in either EB3 depleted or HEBTRON-treated cells.

To further determine how HEBTRON suppressed HCMV replication we tested whether it affected nuclear rotation or AC structure. NHDFs were treated with DMSO or 25µM HEBTRON followed by infection with TB40/E-UL99-eGFP. While time lapse imaging over 3–5d.p.i. detected nuclear rotation >180° in ~71% (36/51) DMSO-treated cells, rotations could not be detected in cultures treated with HEBTRON (Figure 7D; Movie S7A). Furthermore, consistent with our results in siRNA-treated cells (Figure 5F), imaging of UL99-labeled AC structures at 5d.p.i. revealed that DMSO-treated cells contained a dense AC core while the cytoplasm was filled with UL99-positive structures, many of which were uniform in size and exhibited bi-directional motility suggestive of HCMV particles (Figure 7E; Movie S7B). By contrast, HEBTRON treatment resulted in aberrant ACs and large structures resembling MVBs in the cytoplasm, similar to effects of EB3 depletion.

Finally, we measured HEBTRON’s effects on infectious virus production. In low MOI spreading assays, 25µM HEBTRON suppressed the release of infectious HCMV approximately ~100-fold compared with DMSO controls (Figure 7F). In single-round high MOI infections virus release was suppressed ~13-fold (Figure 7G), similar to effects of EB3 depletion above. Overall, HEBTRON not only validated siRNA-based observations but provided additional insights into the underlying mechanism by which EB3 regulated MT formation at the AC to facilitate HCMV replication.

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Derek Walsh (derek.walsh@northwestern.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

METHOD DETAILS

KEY RESOURCES TABLE

See key resources table