Unique and unexplained characteristics of the 1918 pandemic suggest that the risk for lethal secondary bacterial pneumonia after influenza infections depended on the nature, timing, and intensity of immune responses to the pandemic strain; and subsequently on the likelihood of exposure during transient periods of increased susceptibility to bacterial strains against which affected persons had no protective antibodies. In 1918, nearly all humans were immunologically susceptible to infection with the A/H1N1 pandemic strain; not surprisingly, the pandemic spread rapidly worldwide. The rapid spread of the pandemic with high illness attack rates in most age groups indicates that an influenza virus antigenically similar to the pandemic strain did not widely circulate among humans within at least several decades before 1918.

We hypothesize that in 1918 many persons had second lifetime exposures to an immunodominant T-cell epitope that was conserved on an internal protein of the 1918 pandemic strain and a heterosubtypic other strain (e.g., 1889 pandemic strain). When persons were reexposed to the identical epitope in 1918, epitope-specific memory CD8+ T-cells produced excessive cytokines, chemokines, immune cell activation, and epithelial cell necrosis. The immunopathologic effects of the dysregulated T-cell response transiently increased susceptibility of infected hosts to respiratory bacterial strains against which they lacked protective antibodies.

In contrast, persons who were first exposed in 1918 to the immunodominant T-cell epitope of hypothesized concern may have had primary T-cell responses that controlled virus replication without increasing susceptibility to bacterial invasion of the lower respiratory tract. Persons who had multiple prior exposures to influenza viruses and other respiratory infectious agents before 1918 had diversely partitioned memory CD8+ T-cell repertoires and extensive portfolios of bacterial strain–specific antibodies. Their immune responses to infection with the 1918 pandemic strain may have controlled virus replication without increasing their susceptibility to bacterial invasion.

Modern genetic analyses have estimated that 3 distinct variants of influenza A (H1N1) viruses co-circulated in the early 1900s (8,32). These variants were the respective prototypes of all pandemic, seasonal, and classical swine influenza A (H1N1) viruses since 1918. The first epidemic wave of the 1918 pandemic may have been the last wave of the 1889–90 Russian flu pandemic. If so, the first wave spread widely and rapidly in the face of background immunity to an influenza strain that had been circulating among humans for nearly 3 decades (4,18).

Alternatively, the first wave may have been caused by an antigenically distinct seasonal strain of influenza A (H1N1) (8,32). If so, antibodies against hemagglutinin of the seasonal strain did not provide complete protection against infection with the pandemic strain. However, because many internal proteins of human influenza viruses are conserved and strongly immunogenic (e.g., matrix 2, nucleoprotein [NP]), antibodies and memory CD8+ T lymphocytes that were produced during the first wave may have altered clinical expressions, decreased susceptibility to secondary bacterial pneumonia, and reduced deaths during the second wave (33).

The peak of mortality rates among young adults (W-shaped mortality curve) remains a unique and unexplained characteristic of the 1918 pandemic (20). Before World War I, there was relatively little global interconnectedness. Even in the most industrialized countries, many persons lived their entire lives in their birth communities and had relatively little exposure to outsiders. The situation sharply and permanently changed with the social disruptions and population dislocations precipitated by worldwide armed conflict.

Persons born before 1901 (the last year of widespread circulation of the 1890 Russian flu pandemic strain) and after 1875 (the first year after widespread epidemics of a poorly characterized influenza-like illness) were 18–43 years of age in 1918. Worldwide, members of these birth cohorts had high influenza attack rates and were likely to die from secondary bacterial pneumonia during the 1918 pandemic period (Figure).

Persons born before 1875 were >43 years of age in 1918. Before the 1918 pandemic period, they likely had been exposed to more heterosubtypes of influenza A and more respiratory bacterial strains than their younger counterparts. In general, during the pandemic period, middle-age and elderly adults had lower influenza attack rates and were less likely to die than their younger counterparts (Figure).

Also in 1918, persons born after 1901 were less likely than those older to have been exposed to the 1890 pandemic strain or displaced by war-related activities. During the pandemic period, persons <17 years of age had relatively high influenza attack rates but relatively low mortality rates (except for infants) (Figure).

We hypothesize that, soon after infection with the 1918 pandemic influenza strain, infected persons experienced a transient increase in susceptibility of the lower respiratory tract to invasion by bacteria to which they were immunologically naive. Thus, for example, those who were relatively new to their living or work environments (e.g., military recruits, soldiers on troop ships, patients on hospital wards) at the time they were infected with the 1918 pandemic strain had a relatively high risk of death from secondary bacterial pneumonia (12). During the course of their influenza illnesses, such persons were likely to be exposed to bacterial strains to which they lacked protective antibodies (10). Thus, for example, residents of rural areas were relatively unlikely to be exposed to novel strains of bacteria while recovering from influenza, and they had low pandemic-related mortality rates. In contrast, military recruits from the same rural areas were likely to be exposed to novel strains of bacteria while recovering from influenza, and they had relatively high pandemic-related mortality rates (24).

This interpretation explicates the somewhat counterintuitive finding that nurses, medical officers, and the crews of troop ships had high influenza attack rates but relatively low mortality rates during the lethal second wave of the 1918 pandemic. Before being infected with the pandemic influenza strain, these persons were often exposed in their occupational settings to high concentrations of diverse strains of respiratory infectious agents. Because of their extensive portfolios of respiratory bacteria strain–specific antibodies, they were naturally immune to and protected from secondary pneumonia caused by these agents (12,23,24).

The hypotheses presented here are consistent with the historical record and scientifically plausible. For example, studies in humans have identified an immunodominant NP-derived CD8+ T-cell epitope that is consistently presented by high frequency HLA class I molecules and recognized by cytotoxic T lymphocytes. The epitope is present on the NP of the 1918, 1976, and 2009 human pandemic strains and on most swine strains, but not on most other human strains of the past century (34).

Studies in pigs suggest that the NP of most swine influenza strains contains a strongly immunogenic CD8+ T-cell epitope. For example, pigs that were primed with a DNA vaccine that expresses NP, and subsequently challenged with an influenza A strain with the same NP, had dysregulated, pathogenic immune responses (35). Also, pigs that were primed with an inactivated swine influenza A vaccine (A/swine/Iowa/15/1930 H1N1) and subsequently challenged with a later generation swine influenza A strain with markedly different surface proteins (A/swine/Minnesota/00194/2003 H1N2) showed development of enhanced (immunologically potentiated) pneumonia that were not observed after challenge with the homologous strain (36). The findings have been reproduced by using pandemic (H1N1) 2009 virus as the challenge strain and adding a recombinant matrix 2 protein to the vaccine construct (37).

Studies in mice have documented that T-cell–mediated immunopathologic responses can contribute to severe pneumonitis when mice are exposed to a highly glycosylated influenza virus and subsequently infected with a poorly glycosylated strain. Infection with a recent seasonal influenza virus (H1N1), followed by infection with pandemic (H1N1) 2009 virus, elicited severe immunopathogenic responses (38). Finally, studies in ferrets have documented that those infected with pneumococci after acquiring influenza, but not before, showed development of lethal secondary pneumonia and other invasive complications (39).

In summary, we hypothesize that mortality risk after infection with the 1918 pandemic influenza A (H1N1) strain depended on the number, nature, and diversity of prior infections with influenza virus and respiratory bacteria. Specifically, mortality rates during the lethal second wave were highest among persons with prior exposures to heterosubtypic influenza strains that enhanced immunopathogenic effects when a person was infected with the 1918 pandemic strain and had limited exposures to other respiratory infectious agents. In such persons, infection with the pandemic strain caused high viral loads, dysregulated and pathogenic cell mediated immune responses, and transient increases in susceptibility to invasive bacterial infections. If such influenza virus–infected hosts were subsequently exposed to bacterial strains to which they had no protective antibodies, they were at high risk of acquiring life-threatening secondary bacterial pneumonia.

The unique circumstances that enabled the unprecedented mortality rates of the 1918 pandemic no longer exist on a global scale. For example, in modern times, even the most isolated communities (e.g., Pacific islanders, indigenous populations of North America, Australia, and New Zealand) are interconnected through myriad commercial and sociopolitical activities. As a result, most populations are exposed to annual seasonal influenza viruses, and most young adults are exposed to numerous viral and bacterial respiratory pathogens. Thus, compared with the situation in 1918, adults in modern communities have more diversified immune repertoires against influenza strains and bacterial respiratory pathogens. The hypotheses presented may explain at least in part the relatively low mortality rate associated with pandemic (H1N1) 2009 virus. During the 2009 pandemic, many persons who died had underlying medical conditions, including obesity, asthma, cardiovascular diseases, diabetes, and pregnancy; histopathologic changes consistent with diffuse alveolar damage; and evidence of bacterial co-infections (40).

Finally, the findings of this report are relevant to the research and development of a universal influenza vaccine. Candidate vaccines that contain antigens that are highly conserved across influenza A strains and strongly immunogenic must be closely monitored to ensure that T-cell–mediated immune responses to future seasonal and pandemic strains are protective but not pathogenic.