Previously we had proposed three possible models with one

Previously we had proposed three possible models, with one being that hCrm1 fundamentally interacted more favorably or strongly with Rev-RRE complex, compared to mCrm1. We decided to test that model, which was the simplest and most straightforward, by both biochemical and genetic means. By using bacterially-produced Rev-GST and Ran soaked in GTP, along with in vitro transcribed HIV RRE, we were able to demonstrate that in vitro transcribed-translated hCrm1, but not mCrm1, bound to HIV Rev, dependent upon the presence of HIV RRE. As expected, RevM10-GST did not bind either hCrm1 or mCrm1, due to mutations in the NES domain, in the presence or absence of RRE. This result largely confirms the differential binding between mouse and human Crm1 and HIV-1 Rev, as shown by Frankel and colleagues (Booth et al., 2014). To further these results, we tested a number of genetic complementation systems, and the one that gave the most robust and reproducible results was the mammalian two hybrid system. Using that genetic system, in human TCS 359 we were able to reproducibly demonstrate a stronger interaction between hCrm1, compared to mCrm1, and HIV Rev, independent of adding HIV RRE. The fold-effect was modest but significant and repeatedly and reproducibly observed, and consistent with fold-effects that we and others have seen on increases in unspliced HIV RNA (Coskun et al., 2006, Elinav et al., 2012, Sherer et al., 2011, Swanson et al., 2010) and the roughly 2-fold higher binding affinity of purified hCrm1 compared to mCrm1 to Rev (Booth et al., 2014). Of note, in no genetic system did we detect full-length hCrm1 interacting with itself or mCrm1 or mCrm1 interacting with itself with or without Rev-RRE, observations that have been made by single particle electron microscopy with purified Crm1s in the presence of Rev-RRE and Ran-GTP (Booth et al., 2014). Our negative results, however, may be due to the moderately large size of the Crm1s or relative positioning of the fusion proteins. Because these results suggest but do not directly demonstrate that HIV Rev interacts with both the NES domain and HEAT repeat 9A of hCrm1, it will be important to perform additional binding studies and biochemical experiments. Previously it had been shown that Crm1 binds Snurportin-1 50-fold greater than Rev (Paraskeva et al., 1999), but it is likely that in those experiments Rev was in monomeric, not multimeric form. Additionally, multiple hCrm1 sites of interaction may be one way that Rev out-competes cellular cargo, such as Snurportin-1, that bind to the NES domain and an adjacent acidic patch of Crm1 (Dong et al., 2009). Enhanced HIV production from murine cells observed with the Rev ×2NES mCherry fusion protein are consistent with these results, suggesting that the additional NES sequence increases the avidity of the Rev fusion for mCrm1, allowing more efficient nuclear export of viral, intron-containing mRNAs. Although tighter binding to mCrm1 may not allow the Rev ×2NES mCherry to recycle back to the nucleus, this may not be relevant in a transient over-expression system in immortalized murine fibroblasts. Our attempts to test the genetic interaction of Rev ×2NES mCherry fusion with various Crm1s failed since the VP16 fusion was poorly expressed for uncertain reasons and did not trans-complement our Rev minus HIV packaging vector system. Taking into account the modeling study of Booth et al. (2014) suggesting that Crm1 dimerization in the presence of Rev-RRE in vitro depends upon the three key amino acid residues in HEAT Repeat 9A of 411, 412, and 414, and possibly others, a unifying hypothesis is that the interaction of Rev with hCrm1's NES domain and the dimerization interface is stronger than that of mCrm1. Higher resolution cryoelectron microscopy imaging of the Rev-hCrm1 vs. Rev-mCrm1 complex may be informative in this regard. Using the mammalian two-hybrid system, in human cells we observed similar fold-effects for hCrm1 but not mCrm1 interaction with FIV and HIV-2 Rev, but not EIAV Rev. Although all of the Rev fusion proteins were at the expected molecular size, we did not perform cell-based complementation assays to determine their functionality. These results are largely consistent with what we observed when transfecting in various lentiviral constructs and over-expressing either hCrm1 or mCrm1 in murine cells (Elinav et al., 2012). One possibility regarding EIAV Rev is that both hCrm1 and mCrm1 are marginally but equally functional and that equine (E. caballus) (e)Crm1 is required for optimal virus production, which we did not test. eCrm1 differs from hCrm1 at aa positions 412 and 414 whereas eCrm1 differs from mCrm1 at aa position 411, the three key residues of HEAT repeat 9A for hCrm1 binding to HIV Rev (Table 1). It should be noted, however, that eCrm1 differs from both hCrm1 and mCrm1 at several other aa positions as well. A prediction would be that provision of eCrm1 to either human or mouse cells would boost production of infectious EIAV. The aa sequences of these three critical aa residues of Crm1 in HEAT repeat 9A, along with three more C-terminal residues in repeat 10A, of the various mammalian species that have exogenous lentiviruses are summarized in Table 1. That EIAV and FIV behave differently in human and mouse cells suggest that other determinants within Crm1 or other, unknown host factors may be important for EIAV production.