Influenza A Strain-Dependent Pathogenesis in Fatal H1N1 and H5N1 Subtype Infections of Mice

Future treatments may involve customizing treatment to the virus pathotype.

To determine if fatal infections caused by different highly virulent infl uenza A viruses share the same pathogenesis, we compared 2 different infl uenza A virus subtypes, H1N1 and H5N1. The subtypes, which had shown no pathogenicity in laboratory mice, were forced to evolve by serial passaging. Although both adapted viruses evoked diffuse alveolar damage and showed a similar 50% mouse lethal dose and the same peak lung concentration, each had a distinct pathologic signature and caused a different course of acute respiratory distress syndrome. In the absence of any virus labeling, a histologist could readily distinguish infections caused by these 2 viruses. The different histologic features described in this study here refute the hypothesis of a single, universal cytokine storm underlying all fatal infl uenza diseases. Research is thus crucially needed to identify sets of virulence markers and to examine whether treatment should be tailored to the infl uenza virus pathotype.
A ccording to the World Health Organization, infl uenza annually infects 5%-15% of the global population, causing 3-5 million cases of severe illness and ≈500,000 reported deaths. The persistence of infl uenza A virus (H5N1) in poultry populations over the past 6 years and the ability of those viruses to cause fatal infections in humans, along with the recent pandemic (H1N1) 2009 outbreaks, have raised fears of a renewed catastrophic infl uenza outbreak comparable to that of 1918, which caused death in 0.2%-8% of those infected in various countries and ≈50 million deaths worldwide (1). Standard infl uenza symptoms include fever, cough, headache, sore throat, and dehydration, with some reports of diarrhea, vomiting, and bleeding from the mouth or throat. In benign cases, not all of these symptoms are exhibited. In severe cases, additional signs typical of either secondary bacterial pneumonia or acute respiratory distress syndrome (ARDS) occur. Notably, these 2 manifestations are those that cause death in patients with infl uenza, whether seasonal or pandemic or caused by the 1918 subtype H1N1 strain or by recent subtype H5N1 strains.
The catastrophic lethality of the 1918 pandemic makes it paramount that we understand the disease pathogenesis of both severe forms of infl uenza. Because most secondary bacterial pneumonias can be controlled with antimicrobial agents, prevention and treatment of infl uenza-associated ARDS are the major medical challenges that must be addressed to reduce the infl uenza-related death rate. This requires more knowledge about the pathogenesis of ARDS. Alterations in human and mouse lungs have been described for fatal virus infections with pandemic virus strains (subtypes H1N1, H2N2, and H3N2 strains of 1918, 1957, and 1968, respectively) or subtype H5N1 strains. They are all characterized by similar lung dysfunctions and lesions (2,3). The lung becomes fl ooded as its alveolocapillary membranes leak, and the alveoli fi ll with body fl uids. Consequently, the exchange of carbon dioxide and oxygen is reduced, and fatal acute lung failure ensues. The histologic fi ndings depend on the stage of the disease. Edema, epithelial necrosis, fi brin, and hyaline membranes are found during the early exudative phase, and fi broblast and type II cell hyperplasia are found during the proliferative phase. This array of morphologic alterations is known as diffuse alveolar damage. Moreover, mice infected with the 1918 infl uenza virus or with a recent subtype H5N1 human isolate also show considerable similarities in overall lung cellularity, composition of lung immune cell subpopulation, and cellular immune temporal dynamics (4). On the basis of these mostly retrospective studies, the pathogenesis of infl uenza-associated ARDS is widely viewed as being the same whatever the infecting strain.
In this study, we closely monitored ARDS in mice, caused by inoculation of identical doses of 2 different infl uenza strains rendered highly pathogenic toward mice by adaptation. The 2 strains elicited dramatically different disease courses and histopathologic signatures, although both strains caused death in 100% of those infected, evoked the expected diffuse alveolar damage, and led to comparable virus titers in the lungs. The pathogenesis underlying infl uenza-associated fatal ARDS thus depended on the infecting strain.

Animals
Eight-week-old female FVB/J mice weighing 20-25 g were obtained from Charles River Laboratories (L'Arbresle, France). Challenge studies were conducted under BioSafety Level 3 laboratory conditions and in facilities accredited by the Belgian Council for Laboratory Animal Science, under the guidance of the Institutional Animal Care and Use Committees of the Veterinary Agrochemical Research Center and University of Liège. The mice were housed in microisolator cages ventilated under negative pressure with HEPA-fi ltered air. The light/dark cycle was 12/12 h, and the animals were allowed free access to food and water. Before each inoculation or euthanasia procedure, the animals were anesthetized by intraperitoneal injection of a mixture of ketamine (50 mg/kg) and xylazine (30 mg/kg).

Viruses
Two infl uenza A virus strain subtypes that had low pathogenicity for laboratory mice were used in this study: a clade 1 avian infl uenza virus (H5N1) (A/crested_eagle/ Belgium/1/2004), and a porcine infl uenza virus (H1N1) (A/swine/Iowa/4/76). Both viruses were fi rst propagated in the allantoic cavity of 10-day-old embryonating hen eggs and then adapted to the mice by lung-to-lung passaging. At each passage, a set of mice were inoculated intranasally with 50 μL of either allantoic fl uid or lung homogenate containing infl uenza A virus. At 5 days postinoculation (dpi), the mice were killed humanely by an overdose of pentobarbital, followed by exsanguination. The lungs were combined and homogenized in phosphate-buffered saline (PBS)-penicillin-streptomycin, the homogenates were centrifuged at 3,000 g for 10 min, and the supernatant was used for the next passage. The process was stopped when the mice showed a substantial loss of bodyweight on 4 dpi. This occurred after 5 (H5N1) or 31 (H1N1) passages. Lung homogenates from the last passage were homogenized and divided into aliquots for direct use in pathotyping studies, and their titers were determined by standard plaque (subtype H1N1) or median tissue culture infective dose assays (H5N1). Serial dilutions of each adapted virus stock were then injected into FVB/J mice, and the 50% mouse lethal dose (MLD 50 ) was calculated according to the method of Reed and Muench (5).

Pathotyping Studies
For assessment of virus-induced pathogenicity, 2 series of mice were inoculated intranasally with 10 MLD 50 of virus by instillation of 50 μL of diluted stock. Mice were monitored daily for changes in bodyweight to assess virusinduced illness. At selected intervals, 5 (virus titration or histopathology) or 10 (virus titration + dry/wet weight ratio) mice were given an overdose of sodium pentobarbital and exsanguinated by cutting the brachial artery. Lungs and pieces of heart, liver, spleen, pancreas, kidney, brain, and adipose tissue from 5 mice were fi xed in 4% neutralbuffered, ice-cold paraformaldehyde, routinely processed, and embedded in paraffi n for histopathologic evaluation. Five-micrometer sections were stained with hematoxylin and eosin (HE) or periodic acid-Schiff (PAS) for lesion detection. For virus detection, sections were stained by a streptavidin-biotin complex immunoperoxidase method. An in-house immunoglobulin (Ig) G-purifi ed polyclonal rabbit antiserum raised against recombinant infl uenza virus nucleoprotein was used as the source of primary antibodies, and horseradish peroxidase (HRP)-conjugated anti-rabbit IgGs (Dako, Glostrup, Denmark) were used as secondary antibodies. Peroxidase was indicated by the bright red precipitate produced in the presence of 3-amino-9-ethyl-carbazole, and sections were counterstained with Mayer hematoxylin. For virus titrations, lungs from 5 mice were weighed, homogenized in 1 mL PBS, and clarifi ed. The supernatants were used for virus titration by plaque or median tissue culture infectious dose assays. Because the appearance of a biphasic expiratory pattern has been shown to announce death within ≈24 h (6), this qualitative sign was chosen, for humane reasons, as the endpoint of the experimental disease. On this endpoint day, lungs from 5 mice were sampled and weighed, and homogenates thereof were desiccated for dry weight determination.

Clinical, Gross Pathologic, and Virologic Observations
The infl uenza A virus strains (subtypes H1N1 and H5N1) used in this study were isolated, respectively, from a diseased pig in the United States in 1976 and from a crested eagle smuggled from Thailand in 2003 (7). Both were nonpathogenic for FVB/J mice (MLD 50 >10 6 PFU/50% tissue culture infective dose [TCID 50 ]). After adaptation, the strains showed a similar pathogenic outcome in FVB/J mice, i.e., close MLD 50 values: 3.2 PFUs for the subtype H1N1 strain and 6.4 TCID 50 for the subtype H5N1 strain. These results allowed a relevant comparison of their respective pathologic signatures. Overall, virus-associated illness, bodyweight loss, and gross lesions caused by inoculation of 10 MLD 50 were similar for both viruses, except that body condition and respiratory function deteriorated far more rapidly after subtype H5N1 inoculation, the endpoint being reached on 4 dpi for subtype H5N1-induced disease and 8 dpi for subtype H1N1-induced disease. The pathologic processes caused no symptoms for the fi rst 2 (H5N1) or 3 (H1N1) days and then gave rise to general signs such as gradually slower, less frequent, and more erratic spontaneous displacements and a ruffl ed coat. By 3 dpi (H5N1) or 5 dpi (H1N1), all mice became lethargic and abruptly showed clinical signs of respiratory disease, including respiratory distress, labored breathing, and forced expiration. Mice inoculated with subtype H5N1 lost 10% of their bodyweight during the last 48 hours before the endpoint day. In mice that were inoculated with subtype H1N1, weight loss was acute and biphasic: a 10% loss occurred between virus inoculation and the appearance of respiratory symptoms, and an additional 20% was lost during ARDS ( Figure 1). Autopsies performed on the endpoint day of subtype H1N1 disease consistently showed dark, purplish, bulky, noncrepitant, liverlike lungs, fi ndings compatible with a diagnosis of massive pulmonary congestion and consolidation. In subtype H5N1-inoculated mice, the lungs at endpoint were bulky, noncrepitant, and diffusely pinkish gray, which suggests a diagnosis of congestion with massive pulmonary edema. Mice inoculated with either virus had a lung wet weight at endpoint approximately double that of controls, but this weight gain was achieved during the last ≈24 hours in mice inoculated with subtype H5N1, whereas mice inoculated with subtype H1N1 showed a progressive lung weight increase over 96 hours, from 4 dpi to the endpoint day (Figure 2). At the endpoint, the dry/wet weight ratio of the lungs was ≈22% lower for subtype H5N1-infected mice (17.6% ± 1.1%) than for subtype H1N1-infected mice (21.4% ± 1.4%). No obvious gross lesions were observed in the heart, liver, spleen, kidney, brain, or perivisceral fat. The lung virus loads measured on 2, 4, 6, and 8 dpi are shown in Figure 3. The time required to reach the peak virus titer was the same for both virus strains. Death occurred at the peak lung virus concentration for subtype H5N1, but subtype H1N1-associated disease did not become fatal until 4 days after this peak, when virus clearance was already substantial ( Figure 3).

Histopathologic Observations
An exhaustive list of the histopathologic lesions caused by the 2 viruses is given in the online Appendix Table (www.cdc.gov/EID/content/16/4/595-appT.htm). Some changes in lung morphology were identical for both viruses. First, a clear topographic extension of the lesions was perceptible between the fi rst and the last day of infection, with centrifugal spreading from the terminal bronchioles or the alveoli adjacent to the airways. Qualitatively, all alterations characterizing the exudative phase of the histopathologic condition termed diffuse alveolar damage were identifi able, with intense congestion of the alveolar capillaries, marginated intracapillary neutrophils, necrosis of the alveolar epithelium, interstitial and alveolar edema, hyaline membranes, and invasion of the alveoli by (mostly) mononucleate cells. On the other hand, we did not observe cuboidalization of the alveoli (hyperplasia of type II pneumocytes) or hyperplasia or squamous metaplasia of the airway epithelia. These results indicate extremely rapid disease progression, nearly complete elimination of type II pneumocytes, or both. Despite these similarities, when sections of lung tissue samples taken on the last day from infected mice were pooled by subtype, an examiner unaware of which infection he was looking could easily distinguish one from the other (Figures 4, 5). The criteria for attributing lung lesions to the subtype H1N1 strain were the following: 1) earlier and much more extensive degeneration, necrosis, and desquamation of the airway epithelium; 2) a much higher cell density of the peribronchial, peribronchiolar, interstitial, and intra-alveolar infi ltrates; 3) the presence of dense cuffs of mononucleate cells around the arterioles; 4) far less extensive alveolar edemas; and 5) the rarity of alveolar hemorrhages. The lesions caused by the subtype H5N1 strain were distinguishable by the late and mild regressive alterations of the airway epithelium, the extent of alveolar edema, the low cell density of the infl ammatory infi ltrates, the high number of alveolar hemorrhage foci, and the unusual appearance of the pulmonary arterioles (which seemed to have been dissected from the surrounding tissues because of the magnitude of the perivascular edema).
On the other hand, no arteriole showed any cuff of infi ltrated mononucleate cells. Some blood-vessel walls also showed hemorrhage inside the muscle layer. No other organ examined was found to carry any histopathologic lesions except, notably, the liver in subtype H5N1-infected mice ( Figure 5). These livers displayed multifocal necrosis, with necrotic foci consisting of aggregates of hypereosinophilic, pyknotic, and caryorhectic hepatocytes, admixed with a few neutrophils and lymphocytes. Such foci were also seen in the spleen in some animals. Strikingly, numerous PAS-positive islets were detected throughout the livers of subtype H5N1-infected animals, each overlapping with a necrotic focus. Patterns of centrolobular, hydropic, granular (2 dpi), centrolobular (3 dpi), and panlobular (4 dpi) microvesicular fatty degeneration were also observed in the livers of all subtype H5N1-infected animals. Interstitial hemorrhages were seen in the renal medulla.

Detection of Viruses in Tissues
The results of immunohistochemical tests were homogeneous for mice infected with the same strain. Overall, they showed that the subtype H1N1 strain swarmed centrifugally from the bronchioles throughout the lungs over 4-5 days, but remained strictly confi ned to the lungs. The subtype H5N1 virus, in contrast, conquered the whole lung over 24-48 hours; infected some bronchioles only later; and spread to the liver, pancreas, kidneys, spleen, brain, and perivisceral fat.

Topologic Distribution of Subtype H1N1 Antigens over Time
The virus was fi rst detectable in the epithelium of the bronchi and bronchioles on 3 dpi. By 5 dpi, the stain was more conspicuous and appeared also in the alveolar epithelium of the areas adjacent to the airways. By 7 dpi, the virus was detectable in the epithelia of almost all bronchi and bronchioles and in the alveolar epithelium in exten-sive areas of the lungs. In the alveolar structures, staining showed the virus in type I and type II pneumocytes and in alveolar macrophages ( Figure 6, panels A, C, and E). Nonrespiratory organs sampled on 3, 5, or 7 dpi remained strictly virus negative.

Topologic Distribution of Subtype H5N1 Antigens over Time
The virus was detectable from 2 dpi in some type II pneumocytes in peribronchiolar alveoli, some interstitial/ alveolar macrophages, and some endothelial cells in the vicinity of the positive alveoli. In contrast, no nonrespiratory organ examined showed any virus-positive cells. By 3 dpi, staining of the airway epithelium was still discrete and limited, whereas the alveolar epithelium showed more pronounced staining, diffusely distributed throughout the lung. In the liver, multiple nests of positive hepatocytes were detectable, corresponding exactly with the abovementioned necrotic PAS-positive foci. A few renal tubular epithelial cells were also positive. On 4 dpi, the alveolar epithelium was still diffusely stained, but more intensely than on 3 dpi. For the fi rst time, staining of the bronchiolar epithelium was also visible, but not all bronchioles-far from all, in fact-showed this staining. Type II pneumocytes and alveolar macrophages were more often positive than type I pneumocytes ( Figure 6, panels B, D, and F). The appearance of the kidneys and liver was the same as on 3 dpi, with more conspicuous staining. Additionally, virus- . Panels B and E, subtype H1N1: Dense granulocytic and lymphocytic cell infi ltrates in the interstitium and around vessels and airways with focally denuded lamina propria due to epithelial necrosis and desquamation (original magnifi cation ×100). Panel C, subtype H5N1: Airway epithelium is intact; note the striking difference in the number of infi ltrated infl ammatory cells between subtypes H1N1-and H5N1-infected lungs. Dramatic congestion of the vessels is visible, with extensive interstitial and alveolar edema (original magnifi cation ×100). Panel F, subtype H5N1: Alveoli are completely fi lled with edema and hemorrhages; cellular infi ltrates are conspicuously absent (original magnifi cation ×200). Panel H, subtype H1N1: An airway with a totally denuded lamina propria is shown (top, left), with its lumen fi lled with granulocytic and lymphocytic exsudate (original magnifi cation ×200). A prominent periarteriolar lymphocytic cuff is visible (bottom right). Panel I, subtype H5N1: Moderate infl ammatory cell infi ltrate, with no cuffi ng of any airway or vessel; an airway with a still intact epithelium is shown, located just beside a vessel with dramatic peripheral edema (original magnifi cation ×200).

Discussion
Two infl uenza A viruses of different subtypes, derived from different species and showing no pathogenicity toward mice, were forced to evolve by serial passaging in mouse lungs. The 2 adapted viruses obtained showed practically identical virulence levels, with similar MLD 50 values. On the basis of this criterion, they appear to be more virulent than most other viruses used to date in murine models (4,(8)(9)(10)(11)(12)(13)(14)(15)(16). Their virulence is of the same order of magnitude as those of the A/Vietnam/1203/2004 (H5N1) and A/Vietnam/1204/2004 (H5N1) viruses, whose respective MLD 50 s are 0.7 and 2.1 PFUs (17). In both cases, inoculation of 10 MLD 50 causes biphasic weight loss, culminating in death with a loss of ≈10% (H5N1) or ≈25% (H1N1) bodyweight. Viral amplifi cation is maximal for both viruses on 4 dpi, roughly corresponding to the typical inoculation-to-peak lag of natural murine respiratory viruses (6,18). On the other hand, the 2 viruses adapted in the lungs showed replication kinetics that differed substantially from what is observed with natural viruses, with a quasiplateau from 2 to 5/6 dpi instead of the classical Gaussian profi le. Notably, this peculiar amplifi cation kinetics profi le has been described previously for mice infected with mouse-adapted forms of the A/Puerto Rico/8/34 (H1N1) virus (19), the A/South Carolina/1/18 (H1N1) virus (4), and several human subtype H5N1 strains showing high or low pathogenicity (4,14). These reports suggest that this profi le is typical of infl uenza virus amplifi cation by the murine respiratory system.
A fi nal common feature of infection with the 2 virus subtypes was diffuse alveolar damage, which dominates both histopathologic profi les; these results corroborate the pathologic data found in the literature. Seasonal human infl uenza epidemics typically consist of a transient tracheobronchitis caused by preferential attachment of the virus to the laryngeal, tracheal, and bronchial epithelia. In contrast, those infl uenza viruses which are highly pathogenic toward humans, from the pandemic viruses of 1918 (H1N1), 1957 (H2N2), and 1968 (H3N2) to the subtype H5N1 strains isolated from humans since 2003, additionally colonize the bronchiolar and alveolar epithelia, preferentially or not, and cause diffuse alveolar damage as an additional primary lesion (20)(21)(22)(23). The same lesion has been found in experimental animals injected with a recent subtype H5N1 strain (14,(24)(25)(26).
Although both viruses share the same pathogenicity, replication kinetics, and concentration peak, and although they both evoke diffuse alveolar damage by the endpoint day, they differ dramatically in terms of the ARDS course and pathologic signature. Flagrant differences make it easy to distinguish infections by the 2 subtypes. In subtype H1N1 infection, the disease becomes fatal at a point when the pulmonary edema is much less intense and leaves a histopathologic picture characterized by much more dense infl ammatory cell infi ltrates, generating cuffs around the bronchioles and blood vessels. Second, subtype H1N1 colonizes the epithelia of both the upper and lower airways, without any obvious preference, whereas subtype H5N1  remains confi ned essentially to the alveoli and terminal bronchioles. Within the alveoli, unlike the subtype H1N1 strain, the subtype H5N1 strain shows a preferential tropism for type II pneumocytes and alveolar macrophages. Lastly, whereas subtype H1N1 remains strictly confi ned to the respiratory system, subtype H5N1 spreads to other organs. These differences demonstrate unambiguously that the 2 highly virulent infl uenza A viruses studied here cause 2 different forms of ARDS. This fi nding suggests that the physiopathologic data obtained when studying 1 virulent strain should not be extrapolated automatically to other strains. The observed differences also suggest that diverse constellations of critical mutations in the viral genome might lead to the same fatal result. This work addresses the question of possible differences between 2 fatal diseases caused by infl uenza A viruses, although some previous evidence that pointed in the same direction has already been reported. For example, the pandemic human strains of 1918, 1957, and 1968, on the one hand, and the recent subtype H5N1 strains, on the other, show different tropisms: panepithelial for the former strains (20,27,28) and limited to the bronchiolar and alveo-lar epithelia for the latter strains, a result compatible with our own observations on mouse-adapted viruses. Likewise, a panepithelial tropism has been observed for the A/South Carolina/1/18 (H1N1) virus in mice (29), whereas a preference for the bronchioles and alveoli has been noted for recent subtype H5N1 strains that have been injected into macaques, mice, ferrets, and cats (14,25,(30)(31)(32)(33)(34)(35)(36)(37). In addition, the observed strict confi nement of our subtype H1N1 strain to the respiratory system confi rms previously reported data that refute the existence of polysystemic dissemination of non-H5 viruses that are lethal to humans or laboratory animals (20,27,29,38). Conversely, our observation that the subtype H5N1 strain spreads beyond the respiratory system confi rms similar observations of both humans (22,22,39) and laboratory animals (14,24,25,(30)(31)(32)(33)(34).
Although other subtype H5N1 and subtype H1N1 viruses infect other susceptible hosts, they may not show trends similar to those observed here. These results, when integrated with the diverse pieces of evidence reported elsewhere, suggest that fatal infections caused by different highly virulent infl uenza A viruses do not necessarily share the same pathogenesis. To be convinced, one has only to Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 16 note the ease of distinguishing, in the absence of any virus labeling, the histopathologic sections typical of the 2 strains used here (Figure 4). These different histopathologic signatures and different pathogeneses probably refl ect the presence of specifi c sets of virulence markers that will have to be decrypted to anticipate the emergence of a pandemic. In this respect, sequence analysis of both strains will lead to insight on specifi c residues that are relevant for the adaptation and virulence of an infl uenza strain in a new host. Furthermore, the differences between these 2 strains suggest that >1 universal cytokine storm underlies fatal infl uenza diseases. Thus, it might be advantageous to tailor the therapeutic approach to the infl uenza virus pathotype.