Production of infectious HCV in cell culture so far is only possible with the genotype 2a isolate JFH-1 which replicates to very high levels without requiring REMs. In contrast, all genotype 1 isolates described until now replicate very poorly and need enhancing mutations. As shown in this report, at least in the context of Con1, but probably also for other genotype 1 isolates, with the exception of the K1846T substitution in NS4B REMs interfere with virus production. This is an important finding for two reasons: First, these results support earlier assumptions that REMs augment RNA-replication via different mechanisms 
. While most of the mutations we analyzed more or less completely abolished virus production, the K1846T mutation in NS4B elevated RNA replication but interfered with virus formation only to a minor extent. These results clearly point to qualitative differences in the mode by which REMs modulate RNA replication and (in)-directly virus production.
Second, these data clarify why Huh-7 cells transfected with genomes containing REMs 
failed to produce virus particles. It is not due to unfavourable host cell conditions like the lack of assembly factors, but rather a consequence of these mutations especially those in NS3 that interfere with virus production, most likely particle assembly. These data therefore explain the attenuation of adapted Con1 genomes in vivo 
. In fact, in an earlier study we had shown that a Con1 genome containing the three adaptive mutations of the Con1/NS3+S2197P construct (E1202G and T1280I in the helicase and S2197P in NS5A) was unable to establish an infection upon intrahepatic inoculation of a chimpanzee. A Con1 genome with only the NS5A mutation (S2197P) was attenuated and rapidly reverted to wild type. Taking the data from the present study into account, we can assume that these genomes replicated in RNA-inoculated hepatocytes of the chimpanzee, but due to impaired assembly progeny virus was not produced by Con1/NS3+S2197P and therefore the infection was abortive. Since the Con1/S2197P mutant still releases core protein (virus), albeit to very low levels, initial virus spread in the animal most likely was very limited until the mutant had reverted to wild type.
We note that REMs have also been described extensively for the genotype 1a isolate H77 
. These mutations reside primarily in the center of NS5A, but cooperative mutations have also been found in the helicase 
. Although initial attempts to produce infectious H77 virus in cell culture failed with genomes carrying these mutations 
, a highly adapted genome containing 5 REMs has been described recently that replicates to levels comparable to JFH-1 
. Most notably, cells transfected with this H77-S genome release infectious virus particles, but the amounts are very low. Moreover, the specific infectivity calculated as the ratio of HCV RNA molecules (genomes) per infectious unit was about 400-fold lower as compared to JFH-1 (5.4×10e4 vs. 1.4×10e2, respectively) 
. The reason why H77-S transfected cells release such high amounts of HCV RNA is unclear. However, the low buoyant density of the RNA in density gradients and the presence of NS3 and NS5B in these fractions suggest that replication complexes possibly released from dying cells due to cytotoxicity of the efficiently replicating H77-S genome may in part account for these high RNA copy numbers in culture supernatants.
Owing to poor replication, infectivity assays of Con1/wt and Con1/K1846T viruses were extremely difficult. Although replication of the wild type genome could be stimulated with the kinase inhibitor H479, only a very low number of NS5A positive cells became detectable. In case of the Con1/K1846T genome, intrinsic replication efficiency of this genome was still too low for unambiguous detection of viral RNA or proteins. Inclusion of additional REMs either reduced RNA replication (in case of REMs residing in NS5A) 
or blocked core release (in case of REMs residing in NS3). Furthermore, treatment of Con1/K1846T transfected cells with the kinase inhibitor H479 reduced rather than enhanced RNA replication, comparable to what has been described for REMs residing in NS5A 
. Finally, attempts to adapt Con1/wt or Con1/K1846T genomes to continuous Huh-7 cell culture failed, because the genomes could not be maintained in passaged cells or culture supernatants, due to insufficient replication capacity (V.L. and R.B., unpublished).
Although the underlying mechanism interfering with virus assembly is unclear, our result argues for a cross-talk between structural and non-structural proteins during the assembly process. In fact, several TEMs have recently been described in the context of JFH-1 and various JFH-1-based infectious chimeras. These mutations reside in the region encoding core to NS2, but very often in the NS3 helicase domain and the RNA binding replicase factor NS5A 
. The mutations stimulate production of infectious virus particles without major effects on RNA replication arguing that the viral NS proteins modulate the efficiency of virus production 
. In fact, we and others have recently shown that NS5A plays a very critical role in the assembly process, which occurs in close proximity of lipid droplets 
. Core protein accumulates on the surface of these organelles and appears to recruit NS5A or the replicase complex to these sites to trigger virus assembly 
. It was also found that alterations of NS5A phosphorylation, for which casein kinase I appears to play a major role 
, have a strong impact on NS5A – core interaction and virus assembly 
and that most, if not all, REMs reduce NS5A hyperphosphorylation 
. Finally, pharmacological inhibition of NS5A hyperphosphorylation enhances RNA replication as is the case with REMs 
. The current model of HCV assembly that emerges from these observations assumes that via its domain 2 core protein efficiently localizes to lipid droplets 
whereas NS5A is primarily a component of the replicase complex. We speculate that depending on its phosphorylation status, NS5A is recruited to lipid droplets to interact with the core protein in a way that the viral RNA genome is transferred to core, thus triggering virus assembly. In this respect REMs described here may interfere with the interaction between NS5A and core or recruitment of NS5A to lipid droplets or the RNA transfer from the replicase (helicase, NS5A) to the core protein thereby attenuating virus production. The fact that most REMs enhance RNA replication could therefore be due to retention of the viral RNA within the replication complex at the expense of RNA transfer to lipid droplets and/or RNA delivery to the core protein. In this context it is important to note that enhanced replication itself is not responsible for the interference with virus assembly since an adapted Con1 genome with an inactive NS5B polymerase still does not support virus production (data not shown) whereas the analogous replication deficient genome lacking REMs does (). Moreover, JFH-1/wt supports assembly in spite of highly efficient RNA replication. Therefore, we hypothesize that REMs may arrest the viral RNA in a state that prevents the assembly process. The low-level release of virus from H77-S transfected cells may be due to an alternative assembly/release pathway that predominates under these experimental conditions. Clarification of these hypotheses requires more insights into the mechanisms of HCV particle assembly and release.
Although extensive tests with other HCV isolates have not been performed we hypothesize that non-adapted consensus genomes, at least those with proven in vivo
infectivity, will also support production of infectious virus particles in transfected Huh-7 cells. However, owing to the very low replication levels of these genomes, demonstration of infection of cell cultures will be very difficult, even when stimulating replication e.g. by kinase inhibitors. As shown here, inoculation of xenografted mice with cell culture grown HCV particles is an alternative that is more robust and reliable. In fact, infection of uPA-SCID mice with supernatants of Con1/wt or Con1/K1846T transfected cells resulted in a well detectable viremia. In contrast, supernatants of Con1/NS3+K1846T transfected cells turned out to be non-infectious although these supernatants also contained viral RNA and low amounts of core protein. The nature of these RNA/core structures is not known but due to their low abundance they are not amenable to a biophysical characterization. They may correspond to lipid-containing replication complexes that were released from dying cells, similar to what we and others described earlier 
For several positive strand RNA viruses it has been shown that RNA translation, replication and assembly are tightly coupled 
. This coupling may act as a proof-reading mechanism to exclude from progeny particles those viral genomes that have a defect in either translation or RNA replication. As shown here a HCV genome that is unable to replicate (Con1/D318N; ) still releases core protein to an amount comparable to the wild type. Although formal proof is missing that this core protein indeed corresponds to virus particles, our data suggest that HCV particle assembly may occur even in the absence of RNA replication.
In summary we demonstrate the production of infectious HCV particles in the Huh-7 cell line upon transfection with the genotype 1b isolate Con1. The interference of REMs with the assembly process provides an explanation why earlier attempts to produce infectious HCV in cell culture were of very limited success. Although this hurdle has in principle been overcome with the identification of the JFH-1 isolate, more replication and assembly competent HCV isolates are urgently needed to cover the full spectrum of genotypes, especially those that are poorly accessible to antiviral therapy. The observation that infectious HCV particles can be produced in Huh-7 cells by the genotype 1b isolate Con1 may provide a new starting point that likely can be extrapolated to other isolates with proven in vivo infectivity.