Nasal Wash Cytokines during Respiratory Viral Infection in Pediatric Allogeneic Hematopoietic Cell-Transplant Recipients

Abstract

Allogeneic hematopoietic cell-transplant (alloHCT) recipients are at increased risk of complications from viral respiratory-tract infections (vRTIs). We measured cytokine concentrations in nasal washes (NWs) from pediatric alloHCT recipients to better understand their local response to vRTI. Forty-one immunologic analytes were measured in 70 NWs, collected during and after vRTI, from 15 alloHCT recipients (median age, 11 yr) with 19 episodes of vRTI. These were compared with NW cytokine concentrations from an independent group of otherwise healthy patients. AlloHCT recipients are able to produce a local response to vRTI and produce IFN-α2 and IL-12p40 in significant quantities above an uninfected baseline early in infection. Compared with otherwise healthy comparator-group patients, alloHCT recipients have higher NW concentrations of IL-4 when challenged with vRTI. Further study of these immunologic analytes as well as of type 1 versus type 2 balance in the respiratory mucosa in the context of vRTI during immune reconstitution may be of future research interest in this vulnerable patient population.

Clinical Relevance

Pediatric allogeneic hematopoietic cell-transplant (alloHCT) recipients experience unique consequences of viral respiratory-tract infections (vRTIs). In this novel study, we measured cytokines in nasal washes from alloHCT recipients to assess the local mucosal response to vRTIs. We found that alloHCT recipients elaborate a local response to vRTI but may have an imbalance in type 1 versus type 2 responses to infection that leads to the unique consequences of vRTI. This has formed a basis for future larger prospective efforts to study the local immune response in this vulnerable patient population.

Hematopoietic cell transplants (HCTs) are curative therapies for a variety of malignant and nonmalignant pediatric conditions. Although curative for these primary diseases, there are a number of potential complications of these life-saving procedures. Profound immunosuppression, toxicities related to conditioning regimens, infections, immune dysregulation, and graft-versus-host disease (GVHD), among others, contribute to serious complications that affect multiple organ systems. In particular, the prolonged immunosuppression and immune dysregulation leave allogeneic HCT (alloHCT) recipients especially vulnerable to infectious complications (1).

Transfusion-dependent marrow aplasia immediately follows conditioning chemotherapy (and/or total body/lymphoid irradiation) for alloHCT. After approximately 2–4 weeks of aplasia, dependent on the cell source and dose (2), the innate immune system compartment begins to recover (3, 4). Shortly after neutrophil engraftment, natural killer cells engraft and rapidly reach normal concentrations (5, 6).

Recovery of adaptive immunity lags behind innate immune recovery, which renders alloHCT recipients particularly vulnerable to viral infection (7). Mature donor T cells proliferate after alloHCT and provide some defense against invasive infection; however, stable T-cell function with recovery of naive and tolerant T cells does not occur until ∼3–6 months after alloHCT (4, 7). This process is dependent on marrow engraftment of hematopoietic precursors, followed by thymic maturation, and is correlated with thymic function (7, 8). Involved in the humoral response to viral and other infections, B-cell numbers are typically normal by 1 year after transplant; however, it may take up to 2 years to fully recover B-cell function (4).

Resulting primarily from deficient T-cell function in the first year after cell infusion, alloHCT recipients are highly susceptible to severe symptomatology from common viral respiratory-tract infections (vRTIs) (9–12). A feared complication of certain vRTIs is progression from upper-respiratory-tract infection (URTI) to lower-respiratory-tract infection (LRTI) (11). The rates of progression from URTI to LRTI in alloHCT recipients varies from <5% in the case of rhinoviruses to ≤55% in the case of respiratory syncytial viruses (RSVs). The associated mortality with progression to LRTI is ≤50% (13). Risk factors for progression of URTI to LRTI include the patient’s age and the presence of lymphopenia, among others (11). In addition, vRTIs may cause chronic pulmonary conditions in alloHCT recipients (14).

Concentrations of cytokines, chemokines, and growth factors in nasal wash (NW) samples obtained from otherwise healthy pediatric patients with influenza have been longitudinally measured previously. In one study, concentrations of MCP-3 (monocyte-chemotactic protein-3) and IFN-α2 in NW samples at diagnosis correlated positively with influenza disease severity (15). In addition, cytokine clustering analyses in serum have demonstrated associations of certain cytokine modules with outcomes in pediatric patients with severe influenza infection (16). In addition to influenza infection, nasal cytokine concentrations have been associated with disease severity in RSV bronchiolitis in otherwise healthy individuals (17).

NW cytokine concentrations have not previously been reported in pediatric alloHCT recipients, a population known to be at higher risk of morbidity from vRTIs. The primary goal of this study was to test the hypothesis that pediatric alloHCT recipients produce a characteristic reduced local response to vRTI. Secondary goals included comparisons of NW cytokine concentrations at uninfected and infected time points with an independent comparator group of otherwise healthy age-matched individuals.

Methods

Selection Criteria

We identified St. Jude Children’s Research Hospital patients who had undergone alloHCT and who had an NW sample in an institutional review board–approved biorepository. Additional retrospective chart review to obtain clinical data from the electronic medical record (EMR) was approved by the St. Jude Children’s Research Hospital Institutional Review Board. This biorepository contains residual fluid from NWs obtained from patients in the course of routine clinical care from January 1, 2009, to August 17, 2013. Samples with both a positive and negative clinical virologic test result were available in the biorepository for the time between December 28, 2011, and August 17, 2013. As shown in Figure 1, to allow each patient to serve as his or her own control, we included alloHCT patient samples from the biorepository only if the patient had at least one NW when symptomatic and tested positive for a vRTI (referred to as an “infected” sample) and at least one NW obtained after symptoms had resolved and viral testing results were negative (referred to as an “uninfected” sample). Analogous data from a previous independent cohort of influenza patients and their household contacts called FLU09 (15), was used for comparison to test secondary hypotheses (referred to as the “comparator group”) (Figure 1). This study enrolled adult and pediatric patients with community-acquired influenza infection as well as uninfected household contacts. Patients selected from this study group for comparison were randomly sampled between the ages of 0 and 18 using the sample_n() function without replacement in R (version 3.6.1; R Foundation for Statistical Computing). To achieve this, FLU09 samples were first divided into two groups: <10 and 10–18 years of age. Then, a defined number of samples were randomly selected within those groups, with a maximum sample size of 20, to be as close to the alloHCT sample size and average age as possible, without having a one-on-one match per patient with alloHCT.

Figure 1.

Study design. To address the primary hypothesis that allogeneic hematopoietic cell-transplant (alloHCT) recipients are unable to respond to viral respiratory-tract-infection (vRTI) challenge, cytokine concentrations in uninfected and infected nasal washes (NWs) from the alloHCT recipients were compared. Secondary objectives included comparisons of NW cytokine concentrations in both uninfected and infected NW samples between alloHCT recipients and patients in a comparator group consisting of randomly sampled and age-matched patients enrolled in a previous study (15). The figure was created using BioRender.com.

Chart Abstraction and Definitions

Clinical data, including demographics, transplant-related variables, results of microbiologic and radiologic tests, as well as clinical symptoms and examination findings, were abstracted from the EMR for each patient. For an individual patient to contribute more than one infectious episode, symptoms must have completely resolved, and the patient must have had an NW collected after the previous episode that had no virus detected.

Virus Identification

For both the alloHCT and comparator cohorts, respiratory viruses were identified by either direct-fluorescent-antigen or PCR testing.

MILLIPLEX for Multiplexed Cytokine Concentration Measurement

The Luminex xMAP system was used with a MILLIPLEX MAP human-cytokine immunoassay (Millipore), HCYTOMAG-60K, for detection of 41 cytokines, chemokines, and growth factors from NW samples, according to the manufacturers’ protocols.

Statistics

Primary objective: to determine whether alloHCT recipients produce a local response to vRTI

Descriptive statistical analyses were performed in R (version 3.6.1) and SAS 9.4 (SAS Institute). Cytokine concentrations in the first infected NW sample were compared with the first uninfected sample for each discrete infectious episode for each individual patient (Figure 1). Three approaches were used to account for cytokine concentrations below the lower limit of detection (LLOD) in the analysis for the primary objective: using zero, using 50% of the LLOD, and using the LLOD. Data are presented using the LLOD concentration for the analyte when concentrations were below the LLOD; however, the statistical significance of reported results was maintained for each of the above approaches. Statistical testing was performed using a Wilcoxon signed-rank test. P values were then adjusted using the Benjamini-Hochberg false-discovery-rate approach (18). After adjusting for multiple comparisons, a P value of less than 0.05 was considered significant.

Secondary objectives: comparisons between alloHCT recipients and comparator group

Descriptive statistics

Descriptive statistical analyses for these objectives were performed in R (version 3.6.1). Two comparisons of cytokine concentrations were made between alloHCT and comparator-group patients in uninfected and infected NW samples (Figure 1). For the data for all analytes measured, see Tables E1, E2, E4, and E5 in the data supplement. Only cytokines that had ≥50% of values above the LLOD were included in the analyses. For statistical comparisons between cohorts, we identified the highest LLOD among all of the plates for the individual cytokine and set all cytokine concentrations below this LLOD to this concentration. To test whether this approach introduced bias, we evaluated other approaches to managing analytes below the LLOD. These included setting the value to zero if the measurement fell below the LLOD for the individual plate, using the LLOD for the plate as the value, removing those cytokines below the LLOD from the analysis, and setting the value to zero if the LLOD was below the highest LLOD for all of the plates. In this manuscript, we highlight analytes that maintained consistent trends using all of these approaches. As groups were independent of each other and not precisely paired, the Wilcoxon rank-sum test was used to compare groups. P values were then adjusted for multiple testing across cytokines using the Benjamini-Hochberg false-discovery-rate approach (18). After adjusting for multiple comparisons, a P value of less than 0.05 was considered significant.

Cytokine clustering analysis

Cycluster identifies similarly regulated cytokine modules in multiplexed cytokine data using hierarchic clustering and patient-level bootstrapping (16). Each data subset (alloHCT, uninfected; alloHCT, infected; comparator, uninfected; and comparator, infected) was analyzed individually. Before hierarchic clustering, inclusion and exclusion criteria were applied to each sample and analyte. First, patient samples were removed if there was more than one “not-applicable (NA)” value, meaning that the plate did not register a concentration for that analyte. Second, samples were removed if more than 70% of the analytes were above or below the limits of detection (LOD) for the assay plate. Third, analytes were only carried forward if no more than one-third of the values were above or below the LOD. Finally, all remaining values above or below the LOD were set at the LOD.

Results

Study-Participant Characteristics

A total of 15 patients and 19 infectious episodes meeting the above criteria were included. The median age of patients at the time of alloHCT was 11 years (range, 1–18 yr), and alloHCTs were performed between 2010 and 2013. Myeloablative (n = 9) and reduced-intensity conditioning (n = 5) were the most common conditioning regimens; the remaining patient underwent a nonmyeloablative regimen. The most common stem-cell source was bone marrow (60%), and the majority of patients received unmanipulated grafts (73.3%) (Table 1).

Table 1.

AlloHCT Patient and Transplant Characteristics (N = 15 Patients)

Characteristic	Number [Median (Range) or n (%)]
Age, yr	11 (1–18)
Sex
Female	6 (40)
Male	9 (60)
Primary diagnosis
Hematologic malignancies*	12 (80)
Others†	3 (20)
Primary disease status at the start of conditioning
Remission	11 (73.3)
Active	2 (13.3)
Relapse	1 (6.6)
Refractory	1 (6.6)
Year of transplant
2010	1 (6.6)
2011	4 (26.6)
2012	7 (46.6)
2013	3 (20)
Transplant number
First	12 (80)
Second	3 (20)
Stem-cell source
Peripheral blood stem cell	4 (26.6)
Bone marrow	9 (60)
Umbilical cord blood	2 (13.3)
Donor type
Haploidentical	4 (26.6)
Matched, unrelated	8 (53.3)
Matched sibling	3 (20)
Conditioning regimen
Myeloablative	9 (60)
Reduced-intensity conditioning	5 (33.3)
Nonmyeloablative	1 (6.6)
Receipt of total body or total lymphoid irradiation
Yes	6 (40)
No	9 (60)
Graft manipulation
Unmanipulated	11 (73.3)
CD3 depletion	4 (26.6)

Definition of abbreviation: alloHCT = allogeneic hematopoietic cell transplant.

^*Including five patients with acute lymphoblastic leukemia, five patients with acute myelogenous leukemia, one patient with non-Hodgkin lymphoma, and one patient with juvenile myelomonocytic leukemia.

^†Including one patient with hemophagocytic lymphohistiocytosis, one patient with myelodysplastic syndrome, and one patient with therapy-related myelodysplastic syndrome.

Patients were randomly sampled from the comparator group as described in the Methods. The median age of patients sampled for comparison at an uninfected time point was 7.49 years (range, 0.15–17.9 yr). The median age of patients for comparison at Day 1 of infection was 11.26 years (range, 0.33–14.42 yr) (Table 2).

Table 2.

Comparator-Group Subject Characteristics (N = 40)

FLU09 Patients Included for Comparison of Infected Samples (N = 20)		FLU09 Patients Included for Comparison of Uninfected Samples (N = 20)
Characteristic	Number [Median (Range) or n (%)]	Characteristic	Number [Median (Range) or n (%)]
Age, yr	11.26 (0.33–14.4)	Age, yr	7.49 (0.15–17.96)
Influenza season		Influenza season
2009–2010	3 (15)	2009–2010	2 (10)
2010–2011	10 (50)	2010–2011	8 (40)
2011–2012	1 (5)	2011–2012	4 (20)
2012–2013	3 (15)	2012–2013	2 (10)
2013–2014	3 (15)	2013–2014	4 (20)

vRTI Characteristics

Of the 19 infectious episodes, 14 (73.6%) episodes occurred within 1 year of alloHCT. The median days after transplant at vRTI diagnosis was 185 days. The median absolute lymphocyte count (ALC) on Day 1 of vRTI for the 19 infectious episodes was 1,302 cells/mm³ (range, 0–7,800 cells/mm³); a patient’s ALC was <100 cells/mm³ on the first day of vRTI in five episodes. Patients received immunosuppressive therapy (IST) for GVHD on Day 1 of infection in 13 (68.2%) of 19 episodes. Of these 13 GVHD IST regimens, 7 were for GVHD treatment and 6 were for GVHD prophylaxis. Of the six infectious episodes during which patients were not receiving GVHD IST, two episodes occurred within 1 year of alloHCT. The most common virus identified was RSV in 4 episodes (21%), and coinfection with an adenovirus was present in 3 (15.8%) of the 19 episodes (Table 3).

Table 3.

AlloHCT Infection Characteristics (N = 19 Episodes)

Characteristic	Number [n (%) and/or Median (Range)]
Time after transplant, d
0–365	14 (73.6); 73.5 (4–264)
>365	5 (26.3); 499 (396–895)
All patients	19 (100); 185 (4–895)
Absolute lymphocyte count on day of diagnosis, cells/mm3	1,302 (0–7,800)
Immunosuppressive therapy on day of diagnosis
Mycophenolate mofetil and cyclosporine	3 (15.9)
Sirolimus	2 (10.6)
Mycophenolate mofetil	2 (10.6)
Cyclosporine	1 (5.3)
Extracorporeal photopheresis	1 (5.3)
Extracorporeal photopheresis and cyclosporine	1 (5.3)
Mycophenolate mofetil and tacrolimus	1 (5.3)
Methylprednisone and extracorporeal photopheresis	1 (5.3)
Methylprednisone and mycophenolate mofetil	1 (5.3)
None	6 (31.8)
Receipt of IVIG on day of diagnosis	11 (57.9)
Receipt of granulocyte or granulocyte–monocyte colony-stimulating factor on day of diagnosis	3 (15.8)
Presence of GVHD on day of diagnosis
Acute GVHD	4 (21.1)
Chronic GVHD	4 (21.1)
None	11 (57.8)
Viruses identified
Respiratory syncytial virus	4 (21.1)
Parainfluenza 3	3 (15.8)
Human metapneumovirus	3 (15.8)
Adenovirus	3 (15.8)
Influenza A	2 (10.5)
Parainfluenza 2	1 (5.3)
Adenovirus and influenza B	1 (5.3)
Adenovirus and parainfluenza 3	1 (5.3)
Adenovirus and human metapneumovirus	1 (5.3)
Symptoms over the course of infection
Upper respiratory tract*	19 (100)
Lower respiratory tract†	6 (31.6)
Duration of symptoms, d‡	14.5 (1–26)
Duration of virologic test-result positivity, d	18 (1–77)
Admission to the pediatric ICU
No	17 (89.5)
Yes	2 (10.5)

Definition of abbreviations: GVHD = graft-versus-host disease; IVIG = intravenous immunoglobulins.

^*Any one of the following: rhinorrhea, nasal congestion, coryza, cough, or fever.

^†Any one of the following: tachypnea, supplemental oxygen requirement, rales or crackles on auscultation, or radiographic evidence of airspace infiltrate.

^‡Only for the episodes occurring within 1 year of transplant (N = 14).

Every alloHCT patient had at least one URTI symptom, which had served as the indication for obtaining an NW sample, and disease progressed to LRTI in six (31.6%) episodes. Clinical symptoms were present for a median duration of 14 days. Virologic test results, obtained by either direct-fluorescent-antigen or PCR testing, were positive for a longer duration of time (median, 18 d; range, 1–77 d) than the clinical symptoms were present. The patient with 1 day of virologic test-result positivity had 9 days of clinical symptomatology, as documented in the EMR. Two (10.5%) patients required admission to the pediatric ICU for the management of LRTI; one of these patients was 895 days removed from alloHCT but was receiving methylprednisolone and extracorporeal photopheresis for chronic GVHD (Table 3).

Influenza A infection was far more common than influenza B infection in the comparator group. Symptom severity in the comparator group varied, although no patients required pediatric ICU admission during their infection (Table 4).

Table 4.

Comparator-Group Infection Characteristics (N = 20)

Characteristic	Number (%)
Influenza strain
Influenza A (H3)	11 (55)
Influenza A (pH1)	6 (30)
Influenza A (not otherwise specified)	1 (5)
Influenza B	1 (5)
Influenza A and B	1 (5)
Disease severity
Mild	10 (50)
Moderate	5 (25)
Severe	3 (15)
Data unavailable	2 (10)

AlloHCT Recipients Respond to Challenge with Respiratory Viruses

In total, we measured concentrations of 41 immunologic analytes in 70 NW samples from 15 pediatric alloHCT patients with 19 episodes of vRTI diagnosed after alloHCT. Individual cytokine concentrations were variable over a range of concentrations, although some cytokines, including IL-12p70 and VEGF (vascular endothelial growth factor), had a smaller dynamic range (Figure 2A).

Figure 2.

AlloHCT recipients have significantly increased NW concentrations of IFN-α2 and IL-12p40 in response to vRTI. (A) Plot of natural log (ln)–transformed concentrations of analytes on the multiplex panel for all 70 NWs included in this study. (B) Box plots of the two cytokines found in significantly higher concentrations in the first infected NW sample compared with the first uninfected NW sample. Statistical significance was determined by using the Wilcoxon signed-rank test and was adjusted for multiple comparisons across cytokines using the Benjamini-Hochberg false-discovery-rate approach (18). For each of these analytes, P values, both raw and adjusted, were not different using each of the approaches to address cytokines below the lower limit of detection. (C) Plot of ln-transformed IFN-α2 concentrations for patients with prolonged viral infection, which was defined as having a positive virologic test result for more than 14 days. EGF = epidermal growth factor; FGF2 = fibroblast growth factor 2; Flt3.Ligand = fms-like tyrosine kinase 3 ligand; GCSF = granulocyte colony-stimulating factor; GMCSF = granulocyte–macrophage colony-stimulating factor; IL1RA = IL-1 receptor antagonist; IP10 = IFN-γ–induced protein 10; MCP = monocyte-chemotactic protein; P_adj = adjusted P value; PDGF = platelet-derived growth factor; RANTES = regulated upon activation, normal T cell expressed and secreted; sCD40L = soluble CD40 ligand; TGF-α = transforming growth factor α; VEGF = vascular endothelial growth factor.

To determine whether vRTIs provoke a local nasal mucosal immune response in alloHCT recipients, we first compared NW cytokine concentrations between infected and uninfected samples in alloHCT recipients. The median time elapsed between the last infected NW and the first uninfected NW was 10 days (range, 1–37 d). Of the 41 cytokines measured, concentrations of 21 were significantly elevated on Day 1 of infection compared with resolution of infection. After adjusting for multiple comparisons among cytokines, IL-12p40 (P = 0.002; adjusted P value [P_adj] = 0.0432) and IFN-α2 (P = 0.002; P_adj = 0.0432) remained significantly elevated on Day 1 of infection compared with resolution (Figure 2B and Table 5). For patients with prolonged viral shedding, which we defined as having a positive virologic test result for more than 14 days, NW concentrations of IFN-α2 remained stably elevated for the duration of virologic test-result positivity (Figure 2C).

Table 5.

Statistical Comparison of Cytokine Concentrations from AlloHCT Recipients on Day 1 of vRTI and at Resolution of Infection

Analyte	P Value*	P Valueadj*
EGF	0.0266	0.061
CCL11	0.0322	0.061
FGF-2	0.0203	0.061
Flt3 ligand	0.0068	0.061
Fractalkine	0.0674	0.1055
GM-CSF	0.0547	0.0895
CXCL1	0.1677	0.1887
G-CSF	0.0137	0.061
IFN-α2	0.0024	0.0432
IFN-α	0.071	0.1065
IL12-p70	0.0322	0.061
IL-10	0.0371	0.0668
IL-12p40	0.002	0.0432
IL-15	0.0098	0.061
IL-17A	0.0244	0.061
IL-1RA	0.1099	0.1465
IL-1α	0.3223	0.3315
IL-1β	0.2031	0.2156
IL-2	0.0313	0.061
IL-4	0.5693	0.5693
IL-6	0.1016	0.1463
IL-7	0.0156	0.061
IL-8	0.2036	0.2156
IP10	0.1189	0.1529
MCP1	0.1099	0.1465
MCP3	0.1465	0.1817
CCL22	0.1514	0.1817
CCL3	0.0273	0.061
CCL4	0.0171	0.061
PDGF-AA	0.1602	0.186
PDGF-AB/BB	0.0103	0.061
RANTES	0.0244	0.061
TGF-α	0.0313	0.061
TNF-α	0.0391	0.067
VEGF	0.0195	0.061
sCD40L	0.0273	0.061

Definition of abbreviations: EGF = epidermal growth factor; FGF-2 = fibroblast growth factor-2; Flt3 = fms-like tyrosine kinase-3; G-CSF = granulocyte colony-stimulating factor; GM-CSF = granulocyte–macrophage colony-stimulating factor; IL-1RA = IL-1 receptor antagonist; IP10 = IFN-γ–induced protein 10; MCP = monocyte-chemotactic protein; P Value_adj = adjusted P value; PDGF = platelet-derived growth factor; RANTES = regulated upon activation, normal T cell expressed and secreted; sCD40L = soluble CD40 ligand; TGF-α = transforming growth factor α; VEGF = vascular endothelial growth factor; vRTI = viral respiratory-tract infection.

^*Statistical testing was performed using a Wilcoxon signed-rank test. P values were then adjusted using the Benjamini-Hochberg false-discovery-rate approach (18). After adjusting for multiple comparisons, a P value of less than 0.05 was considered significant (indicated in bold).

We then evaluated whether certain cytokine concentrations were associated with the presence of LRTI symptoms or prolonged viral shedding. In the unadjusted analysis, VEGF was associated with the presence of LRTI symptoms (P = 0.021); however, after adjustment for multiple comparisons, the association was no longer significant and would require further validation (P_adj = 0.338; data not presented). We also identified trends toward increased concentrations of VEGF in patients with prolonged viral shedding, although these were not significant after adjusting for multiple comparisons (P = 0.023; P_adj = 0.516; Figure E1A).

The Influence of Transplant-related Variables on NW Cytokine Concentrations

We then evaluated the effect of a patient’s conditioning regimen (myeloablative vs. nonmyeloablative), time after alloHCT (greater than or less than 180 d from transplant), ALC (above or below 400/mm³), or presence of ISTs on uninfected NW cytokine concentrations. We did not observe any differences in cytokine concentrations in uninfected samples that varied by conditioning regimen, ALC, or day after alloHCT. However, there were trends toward higher concentrations of four analytes in NWs from patients on IST. These included PDGF-AB/BB (platelet-derived growth factor isoform AB or BB; P = 0.012), fractalkine (P = 0.022), MCP-3 (P = 0.029), and CCL11 (P = 0.051). However, none of these findings was significant after adjusting for multiple comparisons (P_adj = 0.155 for all comparisons) (Figure E1B).

We then performed the same comparisons in infected NW samples from alloHCT recipients. We did observe multiple differences in cytokine concentrations above and below an ALC of 400/mm³. These included FGF-2 (fibroblast growth factor-2; P = 0.0041), CCL11 (P = 0.0097), PDGF-AB/BB (P = 0.0097), CCL22 (P = 0.012), flt-3 (fms-like tyrosine kinase-3) ligand (P = 0.012), fractalkine (P = 0.021), IL-15 (P = 0.027), MCP-3 (P = 0.027), GM-CSF (granulocyte–macrophage colony-stimulating factor; P = 0.035), MCP-1 (P = 0.036), and sCD40L (soluble CD40 ligand; P = 0.049). In all cases, concentrations were significantly higher in NWs obtained with an ALC < 400/mm³. However, after adjusting for multiple comparisons, these differences were no longer statistically significant (P_adj = 0.0744 for FGF2, CCL11, flt-3 ligand, CCL22, and PDGF-AB/BB; P_adj = 0.10 for fractalkine, MCP-3, and IL-15; P_adj = 0.11 for GM-CSF and MCP-1; P_adj = 0.13 for sCD40L). We did not observe differences in analytes in infected NW samples that varied by conditioning regimen, time after alloHCT, or the presence of ISTs (Figure E2).

Comparison of NW Cytokine Concentrations between AlloHCT and Otherwise Healthy Comparator-Group Patients

We first compared cytokine concentrations in uninfected NW samples between alloHCT recipients and the comparator group (Tables E1–E3). There were significant differences detected in the concentrations of six different analytes. These included G-CSF (granulocyte colony-stimulating factor; P = 0.013), IL-10 (P = 0.045), PDGF-AA (P = 0.0015), PDGF-AB/BB (P = 0.045), IL-4 (P = 0.036), and IL-8 (P = 0.0088). Of these analytes, only IL-4 was present in higher concentrations in alloHCT NWs compared with NWs from otherwise healthy subjects. After adjusting for multiple comparisons, none of these analytes were found to be different by statistically significant amount between patient groups in uninfected NW samples; however, PDGF-AA approached significance (P_adj = 0.054) (Figure 3A).

Figure 3.

Comparison of NW cytokine concentrations from AlloHCT recipients and otherwise healthy individuals. (A) Box plots demonstrating cytokines found to be present in different concentrations in uninfected NW samples from alloHCT recipients compared with otherwise healthy comparator-group patients. Statistical significance was determined by Wilcoxon rank-sum testing and then adjusted for multiple comparisons across cytokines using the Benjamini-Hochberg false-discovery-rate approach (18). (B) Box plots demonstrating cytokines found to be present in different concentrations in infected NW samples from alloHCT patients compared with samples from otherwise healthy comparator-group patients. Statistical significance was determined by Wilcoxon rank-sum testing and then adjusted for multiple comparisons across cytokines using the Benjamini-Hochberg false-discovery-rate approach (18).

To then assess whether alloHCT recipients produce a unique local cytokine response to vRTI, we compared NW cytokine concentrations in infected samples between alloHCT and comparator groups (Tables E4–E6). Here, we observed that concentrations of six analytes were significantly different on Day 1 of infection between the two groups. These included FGF-2 (P = 0.014), GM-CSF (P = 0.03), MCP-3 (P = 0.042), IL-12p40 (P = 0.036), IL-1ra (IL-1 receptor antagonist; P = 0.026), and IL-4 (P = 0.00027) (Figure 3B). Notably, all of these analytes were present in higher concentration in the alloHCT patient cohort. After adjustment for multiple comparisons among analytes, only IL-4 remained elevated by a statistically significant amount in alloHCT-cohort NWs (P_adj = 0.009).

Modular Hierarchic Clustering of Uninfected NW Analytes

We then performed hierarchic clustering analyses to describe correlations among measured immunologic analytes in uninfected NWs using Cycluster, which identifies clusters of coregulated cytokines using hierarchic clustering with patient-level bootstrapping. After applying the inclusion and exclusion criteria for removing analytes from clustering analyses described in the Methods section, we included 29 uninfected alloHCT analytes (Figure 4A). In these alloHCT NW samples, IL-4, which was present in higher concentrations in uninfected NW samples from alloHCT recipients, was found in cluster 3 with CCL11, sCD40L, and IL-12p70. In addition, IL-4 (and CCL11) correlated positively with proinflammatory chemokines (cluster 2: CCL4, CCL3, CXCL1, and IL-8) as well as IL-1ra. IL-4 correlated negatively with proinflammatory cytokines in cluster 6 analytes (IP-10, IL-10, and IL-6) as well as in IFN-γ.

Figure 4.

Hierarchic clustering of analytes in uninfected NW samples. (A) Correlations between cytokine concentrations in uninfected alloHCT NWs. Correlations between cytokines conserved between (B) alloHCT subjects and (C) otherwise healthy comparator-group subjects in uninfected NWs.

The process of optimizing the data sets for hierarchic clustering analysis was conducted with respect to cohort and infection group to best assess the coregulation within a group; however, this resulted in different analytes being included in comparator-group analyses, making it difficult to compare coregulated clusters across groups. Thus, we performed subsequent modular hierarchic clustering analysis on the 15 analytes conserved between the uninfected samples from the two cohorts (Figure 4B). Proinflammatory chemokines (CCL4, CXCL1, and IL-8) were similarly positively correlated in both patient groups. The positive correlation among IL-1α, IFN-α2, and IL-1ra (cluster 3 analytes), was notable in the alloHCT analysis. These analytes correlated negatively with the proinflammatory chemokines in cluster 4. An analogous cluster was not observed in NWs from otherwise healthy patients. In these samples there was a negative correlation between IL-1α and both IL-1ra and IFN-α2.

Modular Hierarchic Clustering of Infected NW Analytes

We performed hierarchic clustering analyses to describe correlations among immunologic analytes in infected NW samples from alloHCT recipients (Figure 5A). In addition to the 29 analytes in the uninfected clustering analysis, 7 others remained for clustering, including IL-2, TNF-α, IL-7, PDGF-AA, G-CSF, IL-1β, and TGF-α (transforming growth factor α). We observed a cluster of positively correlated proinflammatory chemokines, cytokines, and IFNs (cluster 4), which implicates this group of analytes in the early innate local response to vRTI in alloHCT recipients. There was also a positively correlated cluster that included type 1–associated cytokines (e.g., IL-12 and IL-2) together with chemokines and growth factors (cluster 5). Analytes in cluster 4 either positively (IFNs) or negatively (proinflammatory cytokines) correlated with analytes in cluster 5. IL-4, which was present in significantly higher concentrations in infected alloHCT NW samples compared with samples from otherwise healthy patients, correlated positively with the IL-8 family of cytokines and IL-1α (cluster 1). In addition, IL-4 negatively correlated with the proinflammatory chemokine and cytokine cluster (cluster 4) and positively correlated with other cytokines that influence T-cell development (e.g., IL-7 and IL-15).

Figure 5.

Hierarchic clustering of analytes in infected NW samples. (A) Correlations between cytokine concentrations in infected alloHCT NWs. Correlations between cytokines conserved between (B) alloHCT subjects and (C) otherwise healthy comparator-group subjects in infected NWs.

As noted above, because different analytes were included in each subgroup analysis, we performed hierarchic clustering analyses using the 15 conserved cytokines remaining after applying the inclusion and exclusion criteria in infected NW samples (Figure 5B). Although there were certain similarities in correlations between closely related analytes in both patient groups in this analysis (e.g., IL-8 positively correlating with CXCL1 and CCL4 correlating positively with CCL3), there were differences in the correlative relationships between IL-1α and IL-1ra as well as between IFN-α2 and IFN-γ. For example, IL-1α correlates positively with IL-8 and CXCL1 in alloHCT NWs but correlates negatively with IL-8 and CXCL1 in comparator NWs, and IL-1ra correlates positively with sCD40L in alloHCT NWs but correlates negatively in comparator NWs. These highlighted differences indicate distinct IL-1 signaling in alloHCT recipients as compared with otherwise healthy patients in the context of vRTI.

Discussion

Compared with immunocompetent patients, alloHCT recipients experience unique consequences of vRTIs, including progression to LRTI, prolonged viral shedding, and development of chronic lung disease. These clinical observations suggest that pediatric alloHCT recipients respond differently to viral insults to the respiratory tract. In an effort to begin to understand the potential immunologic contributions to these clinical observations, this study longitudinally measured concentrations of 41 immunologic analytes in NW samples from 15 pediatric alloHCT recipients with 19 episodes of vRTI. We found that pediatric alloHCT recipients included in this study were able to produce a local response to viral infection, as evidenced by increased and sustained IFN-α2 production and significant elevation of IL-12p40. However, when compared with otherwise immunocompetent pediatric patients, alloHCT recipients may have a different nasal cytokine milieu in the absence of vRTI and higher concentrations of IL-4, IL-12p40, and IL-1ra in the context of vRTI (Figure 6).

Figure 6.

Proposed model of alloHCT response to vRTI. We hypothesize that because of immune reconstitution, mucosal-barrier injury, and immunosuppressive medications, immune- and epithelial-cell function within the nasal mucosa is dysregulated in alloHCT recipients and leads to the unusual clinical consequences of community vRTIs in alloHCT recipients. In particular, type 1 versus type 2 balance in the respiratory mucosa after alloHCT may be of interest for future study in this vulnerable patient population. The figure was created using BioRender.com.

IFN-α2 concentrations were significantly elevated on Day 1 of vRTI compared with resolution of infection in pediatric alloHCT recipients. Interestingly, concentrations of IFN-α2 remained stably elevated over the course of viral shedding, irrespective of clinical symptomatology. In addition, we noted a trend toward higher concentrations of IFN-α2 in infected alloHCT NW samples compared with samples from comparator-group patients. IFN-α2 is expressed by nearly all cell types in the human body in response to a variety of stimuli, including viral-pathogen PRR (pattern recognition receptor) activation (19). IFN-α2 subsequently acts through STAT1 (signal transducer and activator of transcription 1) and STAT2 signal transduction and eventually leads to the transcription of IFN-stimulated genes and induction of a general antiviral state (19). These data suggest that cells of immune and nonimmune origin within the upper respiratory tract in pediatric alloHCT recipients are able to respond to viral PRRs with a type 1 IFN response.

AlloHCT recipients had a trend toward higher concentrations of IL-4 in uninfected NW samples and significantly higher concentrations of IL-4 in infected NWs as compared with comparator-group NW samples. IL-4 is a regulator of T-cell function and induces T-helper cell type 2 (Th2) T-cell differentiation via STAT6 signal transduction (20). Th2-polarized CD4⁺ T cells subsequently produce additional IL-4 (together with IL-5, IL-6, and IL-13) and are believed to be involved in extracellular immunity, asthma, and allergic diseases (21, 22). Serum concentrations of IL-4 have been shown to be significantly elevated in human alloHCT recipients with acute GVHD (23). As a regulator of T-cell responses, IL-4 influences the immune response to intracellular pathogens. IL-4 influences the outcome of infection in animal models of viral infection, such as influenza and RSV, and studies have demonstrated increased pathogenicity and delayed viral clearance with overexpression of IL-4 (24, 25). We hypothesize that increased local concentrations of IL-4 in the upper respiratory tract of alloHCT recipients may contribute to the delayed viral clearance in alloHCT recipients.

IL-12p40 concentrations were significantly elevated in NWs from alloHCT patients on Day 1 of vRTI compared with at resolution of infection, and there was a trend toward higher concentrations of IL-12p40 in infected NW samples from pediatric alloHCT recipients compared with samples from otherwise healthy comparator-group patients. IL-12p40 can bind with IL-12p35 to form IL-12p70 (the active form of IL-12) or with IL-23p19 to form IL-23 (26, 27). IL-12 is produced by antigen-presenting cells, including macrophages, dendritic cells, and airway epithelial cells (28) and is a “proinflammatory” cytokine (29) required for cell-mediated immunity to viral (30) and other intracellular pathogens (31). IL-12p40 expression is driven by NF-κβ and further augmented by IFN-γ (32, 33). IL-12p40–deficient mice display worse outcomes when challenged with RSV or human metapneumovirus, suggesting that it is protective against vRTI (34, 35). However, when IL-12p40 expression is stimulated, it is typically produced in excess (36, 37), and monomers and homodimers (IL-12p80) of IL-12p40 are antagonistic to the IL-12 receptor (38). As such, local ratios of IL-12p40, IL-12p80, and IL-12p70 may impact IL-12 signaling (39). Indeed, T cells from IL-12p40–transgenic mice with high circulating concentrations of IL-12p40 monomers and homodimers have lower production of IFN-γ and increased production of IL-4 and IL-10 in response to antigens (40). In murine models of alloHCT, IL-12p40 is implicated in bone-marrow engraftment (41) and the development of GVHD (42), and it is present in significantly higher concentrations in the context of acute GVHD (23). Our multiplex panel did not include IL-12p35, IL-12p80, IL-23p19, or IL-23. Measurement of these analytes may provide further insight into local respiratory mucosal IL-12/IL-23 signaling (and Th1 vs. Th17 balance) in alloHCT recipients.

Although there were no differences in IL-1α or IL-1β concentrations between the two patient groups in either infected or uninfected NW samples, we observed a trend toward higher concentrations of IL-1ra in infected NW samples from alloHCT recipients. In addition, we observed differences between the correlations of IL-1a and IL-1ra with other analytes in alloHCT NWs compared with NWs from otherwise healthy patients in the context of vRTI. IL-1α, which acts locally, and IL-1β, which acts systemically, are proinflammatory cytokines that are implicated in host defense to intracellular pathogens (43, 44). IL-1ra is an endogenously produced competitive antagonist with high binding affinity to the IL-1 receptor and acts as an antiinflammatory mediator (45). IL-1ra is produced by both immune and nonimmune cell types in response to a number of signals, including IL-1, exposure to LPS, and IL-4 (44, 46, 47). Relative concentrations of IL-1ra compared with IL-1α and IL-1β play a role in the outcome of certain infections (43). Serum concentrations of IL-1ra reach a nadir during marrow aplasia after HCT conditioning and increase with hematopoietic reconstitution. Concentrations of IL-1ra are further influenced by infectious complications of HCT (including cytomegalovirus reactivation), GVHD, and the administration of intravenous immunoglobulins. Cytokines and antigens from viral or bacterial infections may be involved in upregulation of IL-1ra, and the nadir observed during marrow aplasia reflects a dependence of immune reconstitution for IL-1ra induction (48). Taking these previous data into account, relative local concentrations of IL-1ra, which are influenced by infectious complications and transplant-related variables, may influence the outcome of vRTI in alloHCT recipients.

As this was a retrospective study of samples banked in a biorepository, there are limitations to this study. In the process of stringently applying inclusion criteria to address the primary aim of comparing samples obtained on the first day of infection with samples obtained when no virus was detected, we did not capture all patients with vRTIs with samples available in the biorepository. In addition, there was heterogeneity among the alloHCT patient group with regard to the infecting virus type, time after transplant, ALC, and IST. Each of these variables influences the local immune response to a respiratory virus. We attempted to address some of these potential transplant-related variables that might bias an immune response (Figures E1 and E2). We did identify some trends; however, these data should be interpreted with caution, given the small number of analytes available for comparison for some of these analyses. It has been shown previously that NW cytokine concentrations may vary depending on the infecting virus and may account for some of the differences we observed between alloHCT patients and otherwise healthy patients when comparing cytokine concentrations in infected NW samples (49, 50). Each virus could have its own correlate of severity; thus, the combination of many viruses could explain, at least in part, the marginal significance observed in some of our results comparing alloHCT recipients with the comparator group. This highlights the need for a prospective study, with a larger sample size, and thus more power, to detect robust correlates of severity or unique correlates for common respiratory pathogens. Finally, there were differences in the LLOD between plates from different cohorts, and to address this issue, we evaluated multiple approaches and highlighted analytes with consistent findings across analytic approaches.

This study demonstrates the utility of examining the nasal cytokine milieu as a way of understanding immune responses to life-threatening viral infections in immunocompromised hosts. Pediatric alloHCT recipients produce IFN-α2 and IL-12p40 in response to viral challenge and have increased local concentrations of IL-4 early in vRTI. Furthermore, modular hierarchic clustering analyses demonstrate distinct patterns of correlation among and between immunologic analytes in both uninfected and infected NW samples, which suggests differences in immune signaling in alloHCT recipients compared with an otherwise healthy group, at baseline and during influenza infection. Although the numbers for comparisons were small, we did note some trends toward changes in cytokine concentrations that depended on transplant-related variables, including IST and ALC. We hypothesize that the immune reconstitution coupled with a unique combination of known effects of bone-marrow transplant–conditioning regimens, including endothelial damage, mucosal injury, and immune suppression, may contribute to these findings in uninfected and infected NW samples. Further interrogation of the cytokines identified in this small study, as well as of general mucosal type 1 versus type 2 balance in the context of vRTI during immune reconstitution after alloHCT, may be of future research interest.