Mammalian Adaptation in the PB2 Gene of Avian H5N1 Influenza Virus

Abstract

The substitution of glutamic acid (E) for lysine (K) at position 627 of the PB2 protein of avian H5N1 viruses has been identified as a virulence and host range determinant for infection of mammals. Here, we report that the E-to-K host-adaptive mutation in the PB2 gene appeared from day 4 and 5 along the respiratory tracts of mice and was complete by day 6 postinoculation. This mutation correlated with efficient replication of the virus in mice.

TEXT

The PB2 protein is a major determinant of influenza virus host range. Lysine (K) at amino acid 627 of PB2 (PB2-627) is generally seen in viruses isolated from mammals, while glutamic acid (E) is found in viruses from avian hosts (1); the mammalian-type amino acid PB2-627K enhanced polymerase activity in mammalian cells (2) and enhanced replication and increased virulence in mice (3, 4), whereas PB2-627E was associated with restricted transcription and replication at 33°C and 37°C in human but not in avian cells (5). This cold sensitivity is believed to contribute to poor replication in the upper respiratory tracts (URT) of humans. A pair of wild-type human H5N1 isolates, A/Vietnam/1203/04 (VN1203) (with PB2-627K) and A/Vietnam/1204/04 (VN1204) (with PB2 627E), isolated from a pharyngeal swab and a tracheal aspirate, respectively, from the same patient (6, 7) differed by eight amino acids, including a K-E disparity at PB2-627. Direct sequencing of viral RNA from clinical material showed a mixed population, with avian-type PB2-627E and 701D and human-type PB2-627K and 701N residues (8), suggesting that the latter were independently selected during replication in the human respiratory tract and are markers of mammalian adaptation. When and where the adaptive mutations in the PB2 protein occur remained unclear.

We inoculated mice intranasally (i.n.) with 10³ 50% tissue culture infective doses (TCID₅₀) in a volume of 50 μl of VN1203 and VN1204 and with a VN1203 K627E virus with an engineered mutation at PB2-627; we analyzed the replication and deep sequences of the viruses from the respiratory tracts of mice for 7 days following intranasal administration (Fig. 1). All experiments were conducted using enhanced biosafety level 3 (BSL-3) containment procedures in laboratories approved for use by the U.S. Department of Agriculture and Centers for Disease Control and Prevention and in compliance with the guidelines of the NIAID/NIH Animal Care and Use Committee.

Fig 1.

Experimental design, including infection of mice, tissue harvest, viral titer determination, RNA extraction, PCR, and sequencing. Groups of four mice were infected with one of the three viruses tested, and each group was necropsied at different time points p.i. This resulted in harvested tissue for a total of 28 mice per virus experiment.

The level of replication of VN1203 was slightly higher than that of VN1204 in the URT of mice (Fig. 2A). None of the animals had detectable replication of VN1204 virus on days 1 and 4 in the nasal turbinates (NT) (Fig. 2A); replication of VN1204 virus was undetectable in the tracheas of all four mice on day 1 and in one of four mice on day 4. In addition, one of four mice on days 1 and 4 had no detectable replication in the lungs (Fig. 2B and C). The VN1203 virus replicated more efficiently than VN1204 from day 1 through 5 in the lower respiratory tract (LRT) (Fig. 2C), but on days 6 and 7, the titers of both viruses were similar. Over 7 days, the areas under the curve for the VN1204 virus in NT, tracheas, and lungs were 70%, 89%, and 77%, respectively, compared to the VN1203 virus, and the reduction in virus titer in lungs was statistically significant (P < 0.05). Therefore, through day 5 postinoculation (p.i.), the replication of VN1204 possessing PB2-E627 was less efficient in the LRT than that of VN1203 possessing PB2-627K.

Fig 2.

Level of replication of the A/VN/1203/04 (VN1203) and A/VN/1204/04 (VN1204) viruses in the respiratory tracts of mice. Groups of four mice were inoculated i.n. with 10³ TCID₅₀ in a volume of 50 μl of VN1203 (●) and VN1204 (△) viruses. Virus titers in the nasal turbinates (A), tracheas (B), and lungs (C) of four mice per group, sacrificed from day 1 to 7 p.i., are expressed as mean log₁₀ TCID₅₀/g of tissue (for nasal turbinates and lungs) or log₁₀ TCID₅₀/0.7 ml (for tracheas). The lower limit of detection is indicated by the dashed horizontal line. The asterisks indicate days on which infectious virus was not detected.

The deep sequence data (Fig. 3) for the inoculum showed that VN1203 had only 627K, while VN1204 had only 627E; 100% of PB2 amplicons from the respiratory tract tissues of mice infected with the VN1203 virus retained K at PB2-627 over the 7 days p.i. However, in mice inoculated with VN1204, a substitution leading to PB2-627K was apparent by day 6 p.i. in the NT and tracheas and on day 5 p.i. in the lungs. The deep-sequence data from day 4 in the NT suggesting the appearance of an E-to-K substitution is difficult to interpret because, as stated above, none of the mice had infectious virus recovered from the NT at this time point. Thus, the mammalian host adaptation mutation (E to K at PB2-627) of an avian H5N1 virus appeared independently in the URT and LRT of mice. Interestingly, after the mutation occurred, by day 6 p.i. the VN1204 virus replicated as efficiently as the VN1203 virus in the LRT of mice (Fig. 2C).

Fig 3.

Number of sequences coding for residues E (>99% GAG, with <0.2% of the reads containing GAA [glutamic acid]) and K (>99% AAG, with < 0.2% of the reads containing AAA [lysine]) at PB2-627 in virus stock (A) or infected mice (B) over a 7-day infection period in each tissue sample tested. One half of each organ was homogenized in phosphate-buffered saline (PBS), and viral RNA was extracted and pooled from four mice for each day (day 1 through 7 p.i.). The region around residue 627 in the PB2 gene was amplified by reverse transcription-PCR (RT-PCR) (nt 1802 to 2169) using barcoded primers; the resulting ∼500-bp PB2 amplicons were pooled and sequenced on the 454/Roche GS-FLX sequencing platform. Technical replicates, comprising the same RNA but tested with new RT-PCRs performed with different barcoded primers, were sequenced on the same picotiter plate but in a different well of the 4-well gasket. Barcodes were mapped to the reads with Nucmer (19), and the subsequent coordinate files were used to demultiplex and trim the barcodes. The trimmed and sorted reads were then aligned with the PB2 sequences from a reference strain using fast statistical alignment (FSA) (20). Sequence read numbers were pooled for the two technical replicates. The number of sequence reads was normalized to the number of sequence reads found in the trachea at day 1 for each virus experiment, as these are the time point and tissue for which the virus titer was consistently zero. Any sequences found to be amplified would be representative of noninfectious virions and would thus represent background amplification.

To determine if PB2-627K specifically confers the advantage of efficient replication in a mammalian host, the recombinant VN1203 K627E mutant virus was evaluated. This virus was made VN1204-like only by changing its original K to PB2-E627 while retaining its original backbone sequence. On day 1 after infection with this mutant virus, none of the mice had detectable replication in the URT and LRT (Fig. 4A). The mutant virus was detected from day 2 p.i. in the tracheas and lungs, with titers of 10^1.9 to 10^3.1 TCID₅₀/0.7 ml and 10^2.2 to 10^4.1 TCID₅₀/g, respectively (Fig. 4A). It is notable that the mutant virus replicated less efficiently than both the VN1203 and VN1204 viruses, particularly in the lungs; areas under the curve for VN1203 K627E virus in NT, tracheas, and lungs were 55, 61, and 46%, respectively, compared to the VN1203 virus and the reduction in virus titer was statistically significant (P < 0.05) in both URT and LRT. As shown in Fig. 4, a PB2-E627K mutation appeared by day 5 in the lungs and was predominant by day 6 p.i. in all parts of the respiratory tracts of mice infected with VN1203 K627E virus. The representation of 627E to 627K across all three tissues was similar to that observed for VN1204 in the same tissues.

Fig 4.

(A) Level of replication of the A/VN/1203/04 PB2 K627E (VN1203 K627E) mutant virus in the respiratory tracts of mice. Groups of four mice were inoculated i.n. with 10³ TCID₅₀/50 μl of VN1203 K627E virus. Virus titers in the nasal turbinates (●), tracheas (△), and lungs (■) of four mice per group sacrificed from day 1 to 7 p.i., are expressed as mean log₁₀ TCID₅₀/g of tissue (for nasal turbinates and lungs) or log₁₀ TCID₅₀/0.7 ml (for tracheas). The lower limits of detection are indicated by the dashed horizontal lines. The asterisks indicate days on which infectious virus was not detected. (B) Number of sequences coding for residues E (>99% GAG, with <1% of the reads containing GAA [glutamic acid]) or K (>99% AAG, with <1% of the reads containing AAA [lysine]) at PB2-627 in infected mice over a 7-day infection period for mutant virus VN1203 (K627E). Samples were processed for sequencing as described in the legend to Fig. 3. The number of sequence reads was normalized to the number of sequence reads found in the tracheas at day 1, as these are the time point and tissue for which the virus titer was consistently zero. Any sequences found to be amplified would be representative of noninfectious virions and would thus represent background amplification.

A PB2-E627K substitution was present in five of eight isolates from fatal human H5N1 cases and in three of four isolates from patients who survived. Interestingly, three of four H5N1 viruses isolated from patients without this substitution contained N at PB2-701, while none of the viruses with the E627K substitution had the substitution at PB2-701, indicating that D701N can compensate for the absence of 627K in mammalian hosts (9). Notably, in our study the sequence at PB2-701 of all three viruses retained aspartic acid (D) in the upper and lower respiratory tracts of mice (data not shown) indicating that the D-to-N substitution at PB2-701 was not critical for mammalian adaptation. We also sequenced the M1 gene of the three viruses and found no changes (data not shown). The PB2-627K mammalian host adaptive mutation has also been seen in HPAI H7 viruses (10–12).

Genetic stability of the PB2-E627K mutation has been evaluated in different mammalian host species. With a single passage in mice, a PB2-627E H5N1 poultry isolate (13) acquired the E627K mutation and the resulting virus replicated faster, with systemic dissemination and enhanced lethality (14). However, the stability of PB2-627E in other mammalian models was variable; when A/Indonesia/5/2005 (H5N1) bearing PB2-627E was serially passaged in ferrets, there was no change at PB2-627E (15). When VN1203 wild-type virus and viruses with the engineered mutation PB2-K627E with and without D701N were evaluated in experimentally infected guinea pigs, there was no change at PB2-627E in contact guinea pigs, but there was evidence for a PB2-E627K change by day 6 in 3 of 4 inoculated animals (16).

We found direct evidence that the PB2-E627K shift is complete by day 6 p.i. in all parts of the respiratory tract during mammalian adaptation of this avian virus. Currently licensed anti-influenza drugs are effective when used within 30 h of onset of symptoms (17). However, a significant proportion of H5N1 virus infected patients do not access medical care within 2 days of onset of illness (18). While available antiviral drugs reduce the level of viral replication, they do not shut it down completely. Our findings indicate the importance of the development of antiviral drugs that will interrupt viral replication rapidly, in a critical window.