Human Coronavirus and Acute Respiratory Illness in Older Adults with Chronic Obstructive Pulmonary Disease

Methods

Patients ⩾50 years of age who met spirometric criteria for COPD were recruited at 20 VA Medical Centers, as described elsewhere [35]. All patients gave written informed consent, and the study received approvals by responsible committees on human experimentation and followed procedures in accordance with the recommendations found in the Helsinki Declaration of 1975 as revised in 1983

Pulmonary-function tests included forced expiratory volume in 1 s (FEV₁), percentage of predicted FEV₁, and the ratio of FEV₁ to forced vital capacity (FEV₁:FVC) [35–37]. Spirometric criteria for COPD were an FEV₁ that was ⩽80% of the predicted value and an FEV₁:FVC that was <0.70

Immunizations occurred from October 1998 through January 1999. All patients received TIV for the 1998–1999 influenza season. Patients were randomly assigned, in a 1:1 ratio, to receive intranasally either LAIV (MedImmune Vaccines [formerly Aviron]) (TC group) or placebo (TP group) [35]. From November 1998 through April 1999, the patients were evaluated when they had either 3 symptoms of acute respiratory illness or fever accompanied by 2 symptoms of acute respiratory illness. A nasal and oropharyngeal (NOP) swab specimen and a serum specimen were obtained for detection of virus and antibody, respectively, and a second serum specimen was obtained 3–4 weeks after the onset of acute respiratory illness; all specimens were stored at −70°C

LDI was defined as the sudden onset of respiratory illness plus (1) an NOP-swab culture positive for wild-type influenza virus A or B and/or (2) a 4-fold increase in the end-point titer of serum hemagglutination inhibition (HAI) antibodies to influenza A or B, measured as described elsewhere [35, 38, 39]. A change in titer of serum HAI antibody was not used to diagnose an LDI-caused illness if the acute respiratory illness occurred <28 days after vaccination for influenza

HCoV-associated illness was defined as the sudden onset of respiratory illness plus (1) an NOP-swab culture positive for either HCoV-229E or HCoV-OC43, by RT-PCR using the SARS/Coronaplex kit (Prodesse), which detects SARS coronavirus and HCoV-OC43 and HCoV-229E, and/or (2) a 4-fold increase in the end-point titer of serum antibody to HCoV, by ELISA. RNA for RT-PCR assay was extracted from NOP specimens according to the manufacturer’s recommended procedure (QIAamp Kit; Qiagen). Available NOP specimens from illnesses that had a 4-fold rise in antibody to HCoV-NL63 were cultured in LLC-MK2 cells (a gift from Lia van der Hoek, University of Amsterdam, the Netherlands). RNA extracted from supernatant fluids by use of the QIAmp Ultrasens Virus Kit (Qiagen) was subjected to RT-PCR by use of the GeneAmp RNA PCR Kit (Applied Biosystems) and Spike (S) gene nested primers for amplification of cDNA. The HCoV-NL63 S gene–specific outer primers comprised positions 22557–22582 (positive sense) and positions 23043–23063 (antisense), as described elsewhere [20]. The inner primers, designed by use of the Netherlands HCoV-NL63 isolate (GenBank accession number AY567487) comprised positions 22519–22538 (positive sense) and positions 22909–22928 (antisense). Specimens with viral bands confirmed by gel electrophoresis of the RT-PCR products were considered to be positive. RT-PCR products were purified by use of the QIA quick PCR Purification Kit Protocol (Qiagen) and were sequenced, with the primers, by Alpha BioLaboratory

The HCoV antigens used for the antibody ELISA were produced as follows. HCoV-229E and HCoV-OC43 were obtained from American Type Culture Collection (ATCC) and were grown in MRC-5 and HCT-8–cell monolayers (ATCC), respectively. HCoV-NL63 was grown in LLC-MK2–cell monolayers. Virus-infected cells were frozen and thawed 3 times, the supernatant fluid was clarified of cell debris by centrifugation, the virus was concentrated by overnight centrifugation, and the virus pellet was resuspended in PBS. The concentrated virus was inactivated by psoralen compound (Sigma), followed by irradiation by long-wavelength UV light, as described elsewhere [40, 41]. Mock antigen was prepared, in the same way, from uninfected cells

His₆-tagged recombinant N protein of HCoV-HKU1 was used as antigen in the ELISA to detect antibody to HCoV-HKU1. Expression vector pET-28b (+) (Novagen), encoding the N protein of HCoV-HKU1 cloned into the EcoRI and NotI sites in-frame and downstream from a series of 6 histidine residues, as described elsewhere [27, 28], was a gift from K. Y. Yuen (University of Hong Kong, Hong Kong). The recombinant N protein was expressed by transformation of BL21 (DE3) Singles Competent Cells (Novagen) by the plasmid and was purified by the Ni²⁺ -loaded HiTrap chelating system (Amersham Biosciences), according to the manufacturer’s instructions. In brief, transformed cells were disrupted, and the protein sample was prepared by isolation of inclusion bodies via sonication and washing. The protein sample was loaded onto the HiTrap chelating HP column prepacked with precharged Ni⁺² Sepharose High Performance. The protein then was purified and refolded by serial buffer washes of the column and by liquid chromatography in an FPLC system (Pharmacia LBK Biotechnology), with elution of protein, which was collected in fractions. The fractions were analyzed for the presence of 53-kDa protein by SDS-PAGE. Protein concentration was determined by the Bio-Rad Protein Assay (Bio-Rad Laboratories) and was adjusted to 3 μg/mL in the ELISA. Mock antigen was produced from the same plasmid DNA vector, but without the N protein gene sequence, by the same procedure

Each viral antigen and its respective control were used to coat flat-bottom 96-well Maxisorp Immuno-plates (Nalge Nunc International). The sequence of reagents consisted of serum in a series of 8 2-fold dilutions to generate a broad dose-response curve, mouse anti–human IgG (Fc-specific) conjugated with horseradish peroxidase (Accurate Chemical and Scientific), and peroxidase substrate (KPL). Optical density (OD) was measured at 405 nm by use of a Tecan SLT400 spectrophotometer (Research Triangle Park, North Carolina). The anti-HCoV antibody titer in the ELISA was calculated by the reference-line least-squares-fit method [42]. The cutoff OD was set at 0.3 for all viral antigens, a level that was at least twice the background OD for the respective mock antigens, and corresponded to the linear range of OD versus reciprocal-dilution curves

The first episodes of acute respiratory illness that occurred >7 days after vaccination were included in the clinical analyses and to categorize patients in terms of “illness group.” Subsequent illnesses are not described, both because the earlier illness could affect the characteristics of the later illness and because the number of later illnesses was small. The illness groups comprised those whose first episode of acute respiratory illness was associated with (1) either HCoV-229E or HCoV-OC43, (2) LDI, (3) LDI and either HCoV-229E or HCoV-OC43, or (4) neither LDI nor HCoV-229E nor HCoV-OC43. The respiratory illnesses associated with HCoV-NL63 and HCoV-HKU1 are described individually

The severity of illness was assessed by use of the self-reported, 6 symptom–based, chronic lung disease severity index (CLDSI), which was developed, as part of the Veteran’s Health Study [43–45], to evaluate functional status and well-being and the effects that chronic lung disease has on quality of life. The CLDSI score ranges from 6 (best) to 27 (most severe) [43]. Hospitalizations during the study were identified on the basis of serious-adverse-event forms. Hospitalization was considered to be temporally associated with an acute respiratory illness if it occurred either during the illness or <1 month thereafter

Descriptive statistics were used for the initial analysis of variables; 2×2 χ² tests or Fisher exact tests were used to compare categorical characteristics of the illness groups, and the Wilcoxon&amp;rank-sum test was used to compare continuous characteristics. P values are reported; as an arbitrary correction for multiple comparisons, P<.01 was considered to be statistically significant. An FEV₁ change between study visits was considered to be clinically significant if it was ⩾15% [37, 46]. Univariate logistic regression and χ² analysis were used to identify the univariate association of illness group with each of 11 clinical symptoms and signs of acute respiratory illness and the influenza-vaccine group. Multivariate analysis using stepwise logistic regression identified factors independently associated with HCoV infections. P<.10 was necessary for the factor to be included in the logistic model, and P<.05 was necessary for it to be retained in the model

Results

A total of 715 acute respiratory illnesses occurred among 2215 vaccinated patients; 678 of them were acute respiratory illnesses that were clinically assessable in 585 patients (503 patients had 1 acute respiratory illness and 82 had more than 1 respiratory illness). We tested for HCoV-229E and HCoV-OC43 infections in 679 illnesses among 586 patients, but only 665 among 575 patients that were tested for HCoV-229E and HCoV-OC43 infections were clinically assessable. Overall, 90 (14%) of the 665 fully assessable illnesses were positive for either HCoV-229E or HCoV-OC43 infection (table 1), and the number of illnesses associated with HCoV-OC43 was ∼3 times more than that associated with HCoV-229E. Of the 648 illnesses for which paired serum specimens were available, 74 (11%) had ⩾4-fold changes in the titer of antibody to either HCoV-229E or HCoV-OC43 (table 1). Of the 504 illnesses assessed by RT-PCR, 50 (10%) were positive for either HCoV-229E or HCoV-OC43 (table 1), and none were positive for SARS-coronavirus. None of the 17 illnesses for which only an NOP specimen was available were positive by RT-PCR. Of the 161 illnesses for which only paired serum specimens were available, 18 (11%) were positive for HCoV infection (1 for HCoV-229E and 17 for HCoV-OC43). Of the 487 illnesses for which both NOP-swab and paired serum specimens were tested, 72 (15%) were positive for either HCoV-229E or HCoV-OC43 (table 2). The proportion of these 487 illnesses that was found to be positive by antibody testing (56 [11%]) was not significantly different from that found to be positive by RT-PCR (50 [10%]), but only 34 (47%) of the 72 illnesses positive for either HCoV-229E or HCoV-OC43 were positive by both antibody testing and RT-PCR

No patient experienced more than 1 HCoV-associated illness (figure 1). The HCoV-229E– and HCoV-OC43–associated illnesses occurred throughout the follow-up period at 19 of the 20 study sites. The majority of illnesses (50) associated with HCoV-229E and -OC43 occurred during December 1998 and January 1999 (figure 1). Until March 1999, the number of HCoV-OC43–associated illnesses was greater than the number of HCoV-229E–associated illnesses (figure 1)

Our analysis of 585 first episodes of acute respiratory illnesses included 81 that were associated with HCoV-229E and/or HCoV-OC43 and 94 that were associated with LDI; 19 (20%) of the 94 LDI-associated illnesses were confirmed by culture, and the remainder were confirmed by serologic assay; 88 LDI-associated illnesses had 1 type of influenza virus (A/H3N2 in 59, A/H1N1 in 5, and B in 24), 3 had both type A and type B of influenza virus, and 3 had both A subtypes. A total of 12 of the first episodes of acute respiratory illness were associated with both LDI and either HCoV-229E or HCoV-OC43

Demographic characteristics and health-care outcomes (hospitalization, prednisone use, and death) of the illness groups are listed in table 3 The baseline demographic characteristics of the group of 1630 patients who had no acute respiratory illness diagnosed during the study were similar to those of patients who did have such an illness diagnosed during the study (table 3). In the HCoV, LDI, and non-HCoV/non-LDI groups, prednisone use during the study, but not necessarily during the acute respiratory illness, increased from the levels at enrollment (table 3); the proportion of patients who used prednisone at some time during the study was smaller in the no-illness group and the LDI group than in the non-HCoV/non-LDI group. The proportion of patients who were hospitalized during the study was significantly smaller in the no-illness group than in the non-HCoV/non-LDI group (P<.001) (table 3). Of the patients who were hospitalized, a majority were hospitalized either during or ⩽1 month after acute respiratory illness that was associated with HCoV infection and/or LDI (table 3). All deaths in the HCoV and LDI groups occurred >1 month after the acute respiratory illness; of the 4 deaths in the non-HCoV/non-LDI group, 2 occurred ⩽1 month after the acute respiratory illness

Of patients with illness only related to HCoV, each of the 11 symptoms and signs of acute respiratory illness was found in >50% of patients with HCoV-related illness, and >80% of these patients had new or increased cough, sputum production, nasal congestion, and fatigue and/or malaise (table 4). In stepwise logistic regression analysis, myalgia was found to be significantly less frequently associated with HCoV illness (table 4). Patients in the LDI group were less likely to receive LAIV+TIV vaccines. For each of the 11 signs and symptoms, the proportion of patients in the HCoV group was not significantly different from that in the non-HCoV/non-LDI group (table 4)

Compared with the mean FEV₁ before the acute respiratory illness, both that at the visit for the first episode of acute respiratory illness and that 3–4 weeks later did not change in the HCoV group; however, in both the LDI group and the non-HCoV/non-LDI group, it was significantly lower (worse) at the visit for the first episode, improving 3–4 weeks later (table 5). At the visit for the first episode, the proportion of the LDI group that had ⩾15% worsening in FEV₁ was significantly greater than both that of the HCoV group and that of the non-HCoV/non-LDI group. In the HCoV, LDI, and non-HCoV/non-LDI groups, the mean CLDSI score increased (worsened) significantly at the visit for the first episode, improving 3–4 weeks later in the LDI and non-HCoV/non-LDI groups (table 5); the proportion of patients with a ⩾15% worsening in CLDSI score at the visit for the first episode was largest in the LDI group

Eight patients had a 4-fold rise in antibody titer to HCoV-NL63. In these patients, 7 of the illnesses were first episodes of acute respiratory illness, and 4 were positive, by RT-PCR, for HCoV-NL63. The sequences of the RT-PCR products were located between positions 22569 and 22882 and were very similar to that of HCoV-NL63 (GenBank accession number AY567487); there was 98%–100% similarity at the nucleotide level. One patient had a 4-fold rise in antibody titer to HCoV-HKU1 in association with the first episode, without concomitant evidence of infection by either HCoV-229E or HCoV-OC43 (table 6). All illnesses were associated with 5–10 symptoms of acute respiratory illness. Of these 9 patients, 1 was hospitalized not in association with HCoV, 3 received prednisone (including 1 who received it in association with HCoV-HKU1), and none died

Discussion

HCoV infections were clinically significant in this population of older patients with COPD and other chronic illnesses. The frequency of HCoV-229E and HCoV-OC43 infections that were associated with acute respiratory illnesses was similar to that for LDI infection. We may have underestimated the number of HCoV-associated illnesses; illnesses not meeting our criteria for evaluation were not assessed, and an end-of-study serum sample was not obtained, so we were unable to estimate the frequency of HCoV infections in the 1630 patients who did not have an acute respiratory illness. HCoV-associated illnesses were identified at diverse geographic sites within the continental United States and Puerto Rico. More than 80% of the HCoV-associated infections were characterized by new or increased cough, sputum production, nasal congestion, and fatigue and/or malaise. In the multivariate logistic regression analysis, HCoV-associated illness was statistically less likely than LDI to be associated with myalgia. Hence, HCoV infections were difficult to distinguish from LDI in this influenza-vaccinated population, whose influenza-associated illnesses may have been less severe than those in an unvaccinated population. Illnesses not associated with either HCoV or LDI were not easily distinguishable, statistically, from HCoV. In previous analyses, influenza was more likely to be associated with fever, myalgia, and a worsening in FEV₁ than were noninfluenza acute respiratory illnesses [47, 48]. Acute respiratory illness, no matter the cause, was associated with significant worsening in the CLDSI score, but, on the basis of changes in FEV₁, HCoV appeared to be less severe

Serologic and RT-PCR assays were used as complementary tests for detection of infection by HCoV-229E and HCoV-OC43. More illnesses were positive for HCoV by serologic than by RT-PCR criteria. Delays in collection of NOP specimens after the onset of symptoms of acute respiratory illness, as well as the long interval (∼7 years) between collection and assay of the samples, may have resulted in numbers of viral copies that were below the threshold for detection by the RT-PCR assay

HCoV-OC43 and HCoV-229E infections were associated with more illnesses in our older, chronically ill population than might have been expected on the basis of some previous reports [4–6, 10, 11, 23, 26, 30], although rates similar to those found by the present study have been reported. For instance, 22.9% of 70 virus-associated respiratory-tract infections in patients with COPD that were reported by Greenberg et al. were due to either HCoV-229E or HCoV-OC43 [8]. In a community-based study of elderly patients conducted during 1998–1999, Graat et al. [12] reported that 17% of 108 respiratory infections were due to HCoV-229E and HCoV-OC43 and that this frequency was second to that for rhinoviruses. As in the present study, HCoV-OC43 also predominated over HCoV-229E in a number of other studies [5, 6, 8, 15, 23, 30, 49], whereas HCoV-NL63 and HCoV–HKU1 predominated in 2 recent studies [26, 50]. We identified a small number of illnesses associated with the recently described HCoV-NL63 and HCoV–HKU1. HCoV-NL63 and HCoV-HKU1 cocirculated during the 1998–1999 winter respiratory-virus season at geographically diverse sites in the United States, and the spectrum of illness was similar to that for the HCoV-229E and HCoV-OC43 strains

Respiratory infection due to HCoV has been associated with hospitalization. Glezen et al. [9] have reported that, in 26 (2.5%) of 1029 patients hospitalized because of acute respiratory conditions, the infections were related to either HCoV-229E or HCoV-OC43 and that 14 of these 26 patients were adults >45 years of age. Falsey et al. [11] have reported that 8 of 100 hospitalized elderly patients had concomitant HCoV-229E or HCoV-OC43 infection. In the present study, hospitalization was associated with respiratory infection in all illness groups, both during the month immediately after the visit for the first episode of acute respiratory illness and throughout the study follow-up. There was a lower rate of hospitalization in the group without an acute respiratory illness that qualified for evaluation by the present study; during the study follow-up, the overall mean number of hospitalizations per patient in the HCoV-infected group was more than double that in the no-illness group. In all illness groups, prednisone use increased during the study

Although commonly thought to be associated with cold months [13, 18, 20, 21, 25], HCoV infection has been reported as also occurring during other times of the year [22, 23, 50], and the seasonality of HCoV in tropical and subtropical areas is not so restricted [26, 30]. The present study’s follow-up period extended from October to April, and HCoV infections were detected throughout that period, although they were most frequent during December and January

The results of the present study underscore the clinical importance of HCoV infection in older patients with COPD. Clinical characteristics of HCoV were difficult to distinguish from those characterizing influenza, at least in our influenza-vaccinated population. Although HCoV-229E and HCoV-OC43 were the most frequent HCoV strains identified, HCoV-NL63 and HCoV-HKU1 cocirculated during the 1998-1999 season. Further study of this older, chronically ill population is necessary to determine the genetic evolution of circulating viral strains, the impact that contemporary strains have on the clinical characteristics and severity of illness, and the seasonality of infections