The single most important postulate of using a LAIV is that LAIV mimics natural infection and thereby presents the most effective vaccination strategy. The LAIV mounts all the phases of immune responses and thus provide the most solid protection against infection with a wild type virus. In contrast to inactivated vaccine, the LAIV is administered via the intranasal route, which is the typical entry route of influenza virus, and thus is able to stimulate mucosal immune responses, including substantial levels of secretory IgA antibodies at the respiratory tracts. Furthermore, in addition to humoral immunity mediated by antibody response specific to surface proteins, the LAIV, undergoing restricted replication in infected host, delivers a variety of intracellular viral peptides destined to be bound to MHC class I or II molecules and stimulates specific T cell repertoires. Potential advantages and disadvantages of LAIV are described in Table 1.

The biggest challenge to the development of a LAIV is to render a candidate virus strain as avirulent as possible to address the safety issue of a using live virus, while maintaining a substantial level of viral growth in production hosts to elicit sufficiently high protective immune responses. Once such attenuated virus strain with desired level of safety and efficacy is successfully established, it can be reproducibly used as a donor strain for a variety of reassortant vaccines carrying suitable surface antigens, HA and NA from wild type viruses. The two major attenuation principles adopted for LAIV include the adaptation of influenza virus to nonhumans and at suboptimal temperatures, each of which is achieved through a stepwise or serial passage of the virus in either condition. The host-range vaccines generated in nonhuman cells were produced by serial passages of human influenza virus in nonhuman cells or embryonated chicken eggs [15,16]. More than 30 years ago, A/Puerto Rico/8/34 (H1N1) and A/Okuda/57 (H2N2), both of which have undergone a series of passages in nonhuman cells in the laboratory, were developed as LAIV donor strains. The reassortants based on these attenuated strains were shown to be attenuated and immunogenic in humans [16,17]. However, the reassortants in the genetic background of the A/Puerto Rico/8/34 (H1N1) showed variable degrees of virulence in some clinical trials, thus precluding their further use as LAIVs due to safety issues [18].

Avian influenza viruses as LAIV donor strains for humans also seemed to be suitable as a host-range vaccine since the avian influenza virus was genetically distinct from human strains [19,20]. Such avian-human reassortant vaccines, like the earlier host-range vaccines, showed attenuated phenotype and were immunogenic in humans [19,21], and their particular genetic constellations required for the satisfactory level of attenuation were described [22-24]. However, the avian-human reassortant vaccines also documented residual virulence, especially in seronegative infants and young children, and in some cases the generation of desired reassortant was inefficient [25], indicating genetic incompatibility between avian and human influenza viruses.

Although the two classical attenuation strategies gave way to the cold-adaptation process to establish safe, stable, and effective LAIV donor strains, they obviously provided the conceptual frameworks for subsequent advances in rational vaccine design with parallel advances in the understanding of viral pathogenesis.

Several animal viruses including twelve RNA viruses and two DNA viruses have been successfully adapted to grow at lower temperatures for the development of live attenuated vaccines [26]. Consequently, cold-adaptation has been considered a practical approach for the maintenance of genetic stability of live attenuated vaccines [27,28], unlike other previous strategies. Influenza viruses have also been cold-adapted by several independent groups from USA and former USSR [29-32], where A/Ann Arbor/6/60 (H2N2), B/Ann Arbor/1/66, and A/Leningrad/134/57 (H2N2) were cold-adapted by serial passages at successively lower temperatures to 25℃ in primary chicken kidney cells or embryonated chicken eggs [30,33-35]. Nucleotide sequence analysis has shown that multiple genetic alterations were introduced into all six internal genes, conferring both cold-adapted phenotype (ca) and temperature-sensitive phenotype (ts) [36-38]. It appears that stepwise or gradual lowering of incubation temperature leads to an accumulation of multiple genetic mutations in the viral genome resulting in a genetically stable variant [26, 39]. Several studies engaging reverse genetics and site-directed mutagenesis allowed significant understanding on the role of each mutation towards the attenuation phenotype of a LAIV [40-42]. Well-characterized genetic basis for attenuation and the consistent demonstration of attenuation and efficacy in diverse reassortant vaccines supported the clinical use of the cold-adapted strains. After extensive evaluation, cold-adapted A/Ann Arbor/6/60 (ca A/Ann Arbor/6/60) and cold-adapted B/Ann Arbor/1/66 (ca B/Ann Arbor/1/66) reassortant viruses have been clinically used for humans as annual trivalent seasonal influenza vaccines with promising results of safety and effectiveness [10,43]. In the late 20th century, a novel cold-adapted LAIV donor strains for influenza A and B viruses were developed in South Korea [44,45], where X-31 (A/H3N2) and B/Lee/40 were cold-adapted by serial passages in embryonated chicken eggs, and the resulting attenuated strains demonstrated acceptable level of genetic stability, safety, immunogenicity, and protective efficacy in mice.

In addition to the annual seasonal influenza vaccines, the cold-adapted LAIV is also considered a primary prophylactic measure against influenza pandemics, such as those by HPAI H5N1 viruses or the 2009 pdmH1N1. With the HPAI H5N1 virus emerging as a new serious threat to public health, several LAIVs against the HPAI H5N1 were generated in the genetic background of ca A/Ann Arbor/6/60 as pre-pandemic vaccines and were evaluated for their safety, immunogenicity, and protective efficacy, in various animal models including mice, ferrets, and nonhuman primates [11,12,46]. While inactivated H5N1 vaccines were shown to be poorly immunogenic requiring multiple immunizations with the help of appropriate adjuvants to elicit protective antibody responses [47,48], H5N1 LAIVs not only induced strong systemic and mucosal antibody responses against homologous strains, but also induced cross-clade protective immune responses against heterologous strains, either by a single or boost immunization without the help of an adjuvant [11,12]. The cross-reactivity of H5N1 LAIV is believed to be one of the essential requirements for providing efficient protection in the event of a pandemic, especially because of a the large diversity of antigenicity in HA of HPAI H5N1 viruses found in nature [49]. From its first reported human infection with the HPAI H5N1 virus in 1997, severe morbidity and high mortality in human cases led to a common expectation that the HPAI H5N1 might cause the next pandemic. Therefore, the sudden emergence and rapid transmission of the 2009 pdmH1N1 was rather unexpected and led us to reflect on the 1918 H1N1 Spanish flu pandemic [50-52]. Fortunately, the pdmH1N1 has caused much less fatality than expected and is now in its post-pandemic period. Nevertheless, its characteristic features such as efficient human-to-human transmission and disproportionate infection rate to children and young adults call for constant vigilance [53,54]. In addition, experimental evidence that there was considerable genetic compatibility between HPAI H5N1 and pdmH1N1 and that reassortants between the two viruses demonstrated higher transmission ability in mammalian hosts highlighted the need for vaccination prior to the global circulation of each virus or reassortants between them [55-57]. LAIVs against the pdmH1N1 were also developed using several cold-adapted donor strains including ca A/Ann Arbor/6/60, ca A/Leningrad/134/57, and ca X-31, and were shown to be immunogenic and protective in animal models [13,14,58].

Antigenic drift and antigenic shift refer to the ways in which an influenza virus evades host immunity by alteration its antigenicity. Antigenic drift results from amino acid mutations in the antigenic sites of HA or NA and it abrogates the neutralizing activity of antibodies directed against previous versions of the antigens. Antigenic shift, on the other hand, occurs by genetic exchange between two or more different influenza viruses and gives rise to a novel reassortant strain. These two mechanisms present major challenges to developing an ideal influenza vaccine that could confer long-lasting immunity by a single vaccine type, necessitating annual updates of HA and NA to antigenically match newly emerged variant strain. Occasionally, as exemplified by the pdmH1N1, the antigenic shift by genetic reassortment leads to the creation of an antigenically novel virus to which most of the human population have little immunity, causing a pandemic outbreak [53,59,60]. Nevertheless, the existence of relatively less variable regions in viral proteins, such as those in the polymerase complex (polymerase basic [PB] 1, PB2, and polymerase acidic [PA]), NP, M1, or even those in the HA and NA, continually raised the possibility of developing broadly-protective influenza vaccine. LAIVs appear to better meet the requirements for notion of cross-protection than inactivated vaccines. First, the LAIV itself is in the form of a whole infectious virus and is able to carry the native conformation of all the viral constituents including surface antigens as well as internal viral components, which are subsequently processed into viral epitopes by antigen presenting cells. Many of the conserved regions of influenza viral proteins were found to carry immunogenic T cell epitopes or to be recognized by newly identified antibodies, and the prospect of utilizing such epitopes for constructing universal vaccines has been investigated by extensive studies [61-66]. Second, the LAIV is delivered via nasal route and hence is capable of inducing mucosal secretory IgA antibodies, which is more cross-reactive against heterologous strains through their polymeric nature and higher avidity [67,68]. Remarkably, LAIVs against pdmH1N1 exhibited cross-reactive immunity against antigenically distant influenza strains such as the seasonal H1N1, H3N2, and even avian H5 influenza viruses, even without detectable neutralizing antibody responses in mice and ferrets [58,69,70]. In these reports, cross-protection by LAIVs was ascribed to the induction of both cross-reactive systemic and mucosal antibody responses and to CMI. However, the detailed mechanism and relevant immune correlate of cross-protection by the LAIV remain to be fully elucidated, and the feasibility of cross-immunity in immunologically naïve human has not been addressed yet.

In light of cross-protection, establishing diverse LAIV donor strains is of particular importance since different backbone strains would differentially shape immune responses both qualitatively and quantitatively. The overall shape of immune response may include the strength and durability of protective immunity against homologous strains. Moreover, the breadth of cross-reactivity against heterologous strains is likely to be affected considerably by the distinct involvement of internal viral components in CMI. Additionally, the genetic compatibility between six internal genes from donor strain and updated surface genes from wild type virus should also be taken into account, especially for increasing the production yield of a given vaccine. Although there is no published report yet that describes such differences among currently available LAIV donor strains, it is worthwhile to construct proactively a variety of donor strains to guarantee broader protection coverage.

One breakthrough in the research field of influenza virus is the establishment of reverse genetics technique (RG) [71,72]. The RG allows rapid generation of a desired recombinant influenza virus from cloned cDNAs through simple DNA transfection protocols. More impressively, the RG in concert with site-directed mutagenesis made it possible to design and rescue recombinant mutant influenza viruses carrying desired mutations at any position of interest, which drove the recent advances in our understandings of molecular pathogenesis of the influenza virus. Using the RG, a reassortant LAIV can be rapidly generated by co-transfection of cDNA mixture consisting of the six internal genes of a donor strain and two surface genes from a wild type virus, obviating the need for laborious and time-consuming selection procedure used in classical co-infection followed by antibody selection of the desired 6:2 reassortant vaccine (Fig. 2). Currently, the RG system of cold-adapted LAIV donor strains has been successfully established and is used in almost all vaccine constructs including pandemic vaccines [11,73]. The technique has been recently extended to generate a novel alternative cold-adapted LAIV [74], as an alternative to the classical repeated passage at lower temperatures, allowing the conversion of a wild type virus into a genetically homologous live attenuated vaccine strain.

While executing multiple accessory functions in an influenza virus-infected cell, the nonstructural protein 1 (NS1) dose not assemble into a virion particle. The NS1 protein is not obligatory to the completion of a functional infection cycle and therefore has been considered an attractive viral target for attenuating the virulence of the influenza virus and generating a novel LAIV. One of the main functions of the NS1 during the infection cycle is to play an antagonistic role against host interferon (IFN)-mediated antiviral responses, limiting both the IFN production and subsequent antiviral effects of IFN-induced proteins, such as dsRNA-dependent protein kinase R and 2',5'-oligoadenylate synthetase/RNase L [75-78]. The RG allowed the efficient rescue of a series of mutant influenza A or B viruses containing a truncated NS1 protein, which were attenuated only in IFN competent systems while maintaining growth to a high titer in cell- and egg-based substrates with defects in the type I IFN [79-81]. Thus, the NS1-mutant influenza virus emerged as a leading alternative strategy for the generation of safe and productive LAIVs. Since then, extensive studies have been conducted to characterize the biological and immunological features of the NS1 mutant viruses, and to validate their prospects for providing reliable LAIV donor stains [80,82,83]. Immunization with the NS1 mutant LAIVs were immunogenic inducing robust humoral and cell-mediated immunity in mice, pigs, horses, and macaques, and their protective immunity was not limited to homologous infection but was also cross-reactive against heterologous strains [82,84-88].

In addition to serving as a rational LAIV donor strain, the NS1 mutant virus has the potential to be a viral vaccine vector carrying multiple epitopes, whether from the influenza virus or from others, in the NS1 reading frame. Such potency of the influenza virus NS vector was experimentally demonstrated in previous studies, where the NS1 mutant influenza virus expressing the mycobacterium tuberculosis ESAT-6 protein induced CD4+ Th1 immune responses and protected animals against tuberculosis challenges [89,90]. Based on the wealth of knowledge on cross-reactive T cell epitopes of the influenza virus, it may now be possible again to generate novel universal live vaccine candidates by introduction of multiple appropriate epitopes into the NS segment.

In addition to cold-adaptation and modification of NS1, several new attempts were made to attenuate the virulence of influenza virus and to generate novel live attenuated vaccine, which are summarized in Table 2.