CXCR3 Chemokine Ligands during Respiratory Viral Infections Predict Lung Allograft Dysfunction

Abstract

Community-acquired respiratory viruses (CARV) can accelerate the development of lung allograft dysfunction, but the immunologic mechanisms are poorly understood. The chemokine receptor CXCR3 and its chemokine ligands, CXCL9, CXCL10, and CXCL11 have roles in the immune response to viruses and in the pathogenesis of bronchiolitis obliterans syndrome, the predominant manifestation of chronic lung allograft rejection. We explored the impact of CARV infection on CXCR3/ligand biology and explored the utility of CXCR3 chemokines as biomarkers for subsequent lung allograft dysfunction. Seventeen lung transplant recipients with CARV infection had bronchoalveolar lavage fluid (BALF) available for analysis. For comparison, we included 34 BALF specimens (2 for each CARV case) that were negative for infection and collected at a duration post-transplant similar to a CARV case. The concentration of each CXCR3 chemokine was increased during CARV infection. Among CARV infected patients, a high BALF concentration of either CXCL10 or CXCL11 was predictive of a greater decline in FEV₁ 6 months later. CXCR3 chemokine concentrations provide prognostic information and this may have important implications for the development of novel treatment strategies to modify outcomes following CARV infection.

Lung transplantation is a treatment option for advanced lung diseases that can improve survival and enhance quality of life. Unfortunately, due to a high incidence of infectious and non-infectious complications, the median survival following lung transplantation is less than 6 years (1). The development of bronchiolitis obliterans syndrome (BOS) is the most important factor limiting the long-term survival with approximately 50% of patients affected within 5 years of transplantation (2). BOS manifests as progressive airflow obstruction of the lung allograft leading to respiratory disability and ultimately death.

Community-acquired respiratory viruses (CARV) have been increasingly recognized as common pathogens after lung transplantation. Apart from the direct effects on morbidity and mortality, CARV infections have been linked to the development of BOS (3–5). BOS is classically thought to be the result of alloimmune-mediated injury. It has therefore been speculated that the immune response to viral replication in the lung allograft leads to enhanced allo-recognition.

CXCR3 is a G protein-coupled receptor that is primarily expressed on activated lymphocytes. Prior studies have shown a role for CXCR3 and its interferon-inducible CXC chemokine ligands CXCL9 (monokine induced by human interferon-gamma/MIG), CXCL10 (IFN-gamma-inducible 10,000 molecular weight (MW) protein/IP10), and CXCL11 (IFN-gamma-inducible T-cell alpha chemoattractant/ITAC) in regulating leukocyte trafficking to the lung during respiratory viral infection (6–8). We and others have also demonstrated CXCR3/ligand biology is important in the continuum of acute to chronic lung allograft rejection (9–11).

We hypothesized that CARV infection after lung transplantation would be associated with augmentation of CXCR3/ligand biology in the lung allograft. More importantly, the concentrations of CXCR3 ligands in bronchoalveolar lavage fluid (BALF) during CARV infection would be useful biomarkers of the future development or progression of lung allograft dysfunction manifest as a decline in forced expiratory volume in 1 second (FEV₁).

MATERIALS AND METHODS

Study design and subject selection

With institutional review board approval and informed written consent, lung transplant recipients transplanted at the University of California Los Angeles (UCLA) Medical Center after January 2000 were prospectively enrolled into an observational cohort to investigate mechanisms of lung allograft dysfunction with the collection of BALF for subsequent research analyses. Within this cohort, we performed a retrospective nested case control study to determine the effect of CARV infections on alterations in CXCR3 chemokine concentrations within the lung. Among the CARV infected cases, we also explored the utility of CXCR3 ligands for predicting subsequent lung allograft dysfunction. We defined CARV infection as a positive test for respiratory syncytial virus (RSV), parainfluenza, influenza A or B, or adenovirus from a respiratory specimen. Patients diagnosed with CARV infection that had a BALF specimen collected for research analyses were included in the study. BALF specimens collected before January 1^st 2009 were eligible. For comparative analyses, each CARV-positive subject was matched with 2 lung transplant recipients (for increased power) that were never diagnosed with CARV infection and had a bronchoscopy performed at a similar duration post-transplantation that was negative for infection. When there were more than 2 eligible negative recipients for a given CARV-positive subject, CARV-negative recipients were selected based upon a date of transplant closest to the matching CARV case. Follow-up data was capped at December 31^st, 2009.

Standard Care of Lung Transplant Recipients

Post-transplant immunosuppression and antimicrobial prophylaxis was administered according to UCLA lung transplant program protocols as previously described (12). Bronchoscopy was performed according to a surveillance protocol and when clinically indicated. Details are provided in the online supplement.

Lung function measurement and diagnostic definitions

Pulmonary function testing (PFT) was performed every 1–2 weeks during the first 3 months after transplantation and every 4–8 weeks thereafter. BOS was diagnosed and staged by PFT data according to the International Society for Heart and Lung Transplantation (ISHLT) guidelines (13, 14).

Acute rejection grading and treatment

Acute rejection (AR) was diagnosed and graded by a pathologist experienced in lung transplantation according to the standard ISHLT criteria (15, 16). Details of AR treatment are provided in the online supplement.

Diagnosis of CARV infection

CARV infections were identified by reviewing virology records from all respiratory specimens of all eligible subjects enrolled in the UCLA lung transplant observational cohort. During this time period, CARV detection was performed using standard microbiologic techniques in the UCLA clinical virology laboratory. Cell cytospin preparations were stained with virus-specific immunofluorescent labeling for RSV, parainfluenza (1, 2, and 3), influenza (A and B), and adenovirus. The cell free supernatant was further tested for viral presence by inoculation into culture tubes of human epidermoid cancer cells (HEp-2) and rhesus monkey kidney cells and examined for cytopathological effect (CPE). Specimens with CPE underwent viral identification by immunofluorescent labeling as described above.

CARV infections were further classified by the presence or absence of lower respiratory tract signs or symptoms. The diagnosis of CARV lower respiratory tract infection (LRTI) required CARV infection and additional documentation of new shortness of breath, hypoxemia, or radiographic infiltrate.

BALF Collection, Processing, and Analysis

Attempts were made to collect BALF specimens for research analysis at the time of each lung transplant bronchoscopy. Bronchoalveolar lavage was performed following a standardized protocol (online supplement).

Unconcentrated BALF was analyzed for protein concentrations of chemokines using Luminex or ELISA technologies with a CXCL9 bead assay (Bio-Rad Life Science Research, Hercules, CA, USA), a CXCL10 bead assay (Millipore, Billerica, MA, USA) and the CXCL11 DuoSet kit (R&D Systems, Minneapolis, MN, USA) according to the manufacturer’s instructions.

Immunohistochemistry

Immunohistochemistry was performed on paraffin embedded slides for the localization of CXCR3 and its chemokine ligands as previously described (17). Details are provided in the online supplement.

Statistical analysis

Descriptive statistics are expressed as means ± standard deviation (SD) unless otherwise noted. Conditional logistic regression was used for paired comparisons between CARV cases and matched negative controls. For these analyses, non-normally distributed variables were log-transformed when appropriate.

To explore associations between chemokine concentrations and lung allograft dysfunction, we determined the percent change in FEV₁ 6 months after CARV diagnosis. BALF chemokine log concentrations were correlated with the percent change in FEV₁, after adjusting for duration post-transplant, using linear regression models. High and low categories were defined as greater than or equal to the median concentration for each chemokine. The percent change in FEV₁ at 6 months was then compared between high and low chemokine groups using the Wilcoxon test.

Analyses were performed with SAS statistical software, version 9.1 (SAS Institute Inc., Cary, NC) and GraphPad Prism version 5.0 (GraphPad Software Inc., La Jolla, CA).

RESULTS

Study Patient Characteristics

A total of 293 lung transplant recipients were screened for this study (5 heart-lung, 177 bilateral, and 111 single). Thirty patients were diagnosed with CARV infection prior to January 1^st 2009, 17 of which had a BALF specimen available for research analyses and were included in the study. Each CARV-positive case was matched with 2 lung transplant recipients that were never diagnosed with CARV for a total of 34 CARV-negative recipients. Baseline demographic and clinical characteristics at the time of transplantation were similar in the CARV-negative and CARV-positive groups (Table 1). However, at the time of CARV diagnosis or control bronchoscopy, CARV cases had a significantly higher cumulative AR score, representing the sum of all prior AR grades, and were more likely to have a clinical indication for bronchoscopy as compared to control subjects (Table 2).

TABLE 1.

SUBJECT CHARACTERISTICS AT TRANSPLANTATION

	CARV-negative (n=34)	CARV-positive (n=17)	P Value
			0.83
Age at transplant, years ± SD	55.8 ± 12.1	56.6 ± 11.8	0.72
Female, n (%)	18 (53)	8 (47)
Pre-transplant disease, n (%)			0.25†
Restrictive	18 (53)	9 (53)
COPD/α1-AT	10 (29)	8 (47)
CF/bronchiectasis	3 (9)	0 (0)
Vascular	3 (9)	0 (0)
Type of Transplant, n (%)			0.48
Bilateral	25 (74)	11 (65)
Single	9 (26)	6 (35)
Induction Therapy, n (%)			0.52
Thymoglobulin	21 (62)	12 (71)
Basiliximab	13 (38)	5 (29)
Cardiopulmonary bypass, n (%)	24 (71)	11 (65)	0.67
Highest PGD Grade in first 72 hours			0.16
0 or 1	19 (56)	6 (35)
2	5 (15)	3 (18)
3	10 (29)	8 (47)

^†Comparing COPD/a1-AT (Yes vs. No)

TABLE 2.

SUBJECT CHARACTERISTICS AT STUDY BASELINE^†

	CARV-negative (n=34)	CARV-positive (n=17)	P Value
Days to BAL, median (IQR)	201 (140, 516)	202 (126, 653)	0.26
Bronchoscopies, median (IQR)	5 (4, 6)	5 (4, 7)	0.25
Indication for bronchoscopy, n (%)			0.03††
Surveillance	20 (59)	5 (29)
Follow up of acute rejection	5 (15)	2 (12)
Clinical indication	9 (26)	10 (59)
Cough	4	7
Dyspnea/hypoxia	4	10
Fever	1	3
Isolated radiographic abnormality	2	0
Maintenance immunosuppression, n (%)			0.16†††
Corticosteroids + tacrolimus	6 (18)	0 (0)
Corticosteroids + tacrolimus + MMF	28 (82)	14 (82)
Corticosteroids + tacrolimus + sirolimus	0 (0)	3 (18)
Concurrent AR, n (%)	2 (6)	2 (12)	0.49
Cumulative AR score, median (IQR)	0 (0, 2)	2 (1, 3)	0.02*
Existing BOS, n (%)	4 (12%)	4 (24%)	0.99

^†Study Baseline indicates the time of CARV diagnosis or matched bronchoscopy.

^††Comparison of clinical indication vs. surveillance or follow up of acute rejection.

^†††Comparison of 3 drugs (MMF or sirolimus) vs. 2 drugs

^*dichotomized as 0 vs. >0

More than half (10/17) of the CARV-positive recipients had lower respiratory tract symptoms at the time of CARV diagnosis. Parainfluenza was the most common viral pathogen, responsible for 10 cases of CARV infection (Table 3). Six CARV cases were coinfected with another potential pathogen at the time of CARV diagnosis; 2 with Aspergillus fumigatus, 1 with Aspergillus terreus, 2 with Pseudomonas aeruginosa, and 1 with Streptococcus pneumoniae. By definition, controls were negative for infection.

TABLE 3.

RESPIRATORY VIRUSES IN THE ABSENCE OR PRESENCE OF LOWER RESPIRATORY TRACT SYMPTOMS

	Any Respiratory Virus	Parainfluenza	RSV	Influenza A or B
LRTI symptoms absent, n	7	4	1	2
LRTI symptoms present, n	10	6	2	2
Total, n	17	10	3	4

BALF CXCR3 Ligand Concentrations are Increased During CARV Infection

We measured the concentrations of CXCL9, CXCL10, and CXCL11 in BALF at the time of CARV diagnosis (n=17) or matched control bronchoscopy (n=34). The concentration of each chemokine was significantly greater in BALF from CARV-positive recipients (P=0.006, P=0.003, and P=0.01; respectively) (Figure 1A). Among CARV infected patients, the presence or absence of LRTI did not significantly affect BALF chemokine concentrations (Figure 1B).

Figure 1.

CXCR3 chemokine ligand alterations in BALF during CARV. (A) Concentrations of CXCL9, CXCL10, and CXCL11 in BALF were all elevated at the time of CARV diagnosis as compared to BALF collected from a time matched bronchoscopy in control subjects (P=0.006, P=0.003, and P=0.01; respectively). (B) At the time of CARV diagnosis, concentrations of CXCL9, CXCL10, and CXCL11 were similar in CARV cases with and without lower respiratory tract infection (LRTI) (P values all non-significant). (C) At the time of the last bronchoscopy prior to a CARV diagnosis or matched control bronchoscopy, there were no differences in CXCL9, CXCL10, or CXCL11 BALF concentrations (P values all non-significant). (D) Among CARV cases, the levels of CXCL9, CXCL10, and CXCL11 were increased at the time of CARV diagnosis as compared to the last bronchoscopy prior to CARV diagnosis (P=0.007, P=0.002, and P=0.04, respectively).

We also measured chemokine concentrations in the BALF collected from the last bronchoscopy performed prior to a CARV diagnosis (n=14) or the matched bronchoscopy in CARV-negative recipients (n=30). There was no significant difference in the concentration CXCL9, CXCL10, or CXCL11 between CARV-negative recipients and those patients who went on to develop CARV infection (Figure 1C). Furthermore, among CARV-positive recipients, the concentrations of CXCL9, CXCL10, and CXCL11 were each significantly greater at the time of CARV diagnosis than at the last bronchoscopy performed prior to CARV diagnosis (P=0.007, P=0.002, P=0.04; respectively) (Figure 1D). Concentrations of each chemokine were similar in CARV positive recipients with and without co-infection (data not shown).

Immunolocalizaton of CXCR3 and its Ligands during CARV Infection

We performed immunohistochemistry to identify the cells involved in the production of CXCR3 chemokine ligands, as well as those cells expressing CXCR3 within CARV infected lung allografts. We obtained paraffin embedded lung tissue from lung transplant recipients infected with CARV (1 influenza LRTI; 6 parainfluenza – 3 with LRTI; and 2 RSV, 1 with LRTI). Typical histopathologic findings ranged from peri-airway mononuclear cell infiltration and epithelial reserve cell hyperplasia to organizing pneumonia and/or DAD. For each CARV infection, staining for CXCL9, CXCL10, and CXCL11 consistently demonstrated strong expression by hyperplastic bronchial epithelial cells and sub-epithelial mononuclear cells. As expected, CXCR3 expression was prominent on peribronchial infiltrating mononuclear cells. Interestingly, there was also strong expression of CXCR3 by hyperplastic bronchial epithelial cells. Representative examples of immunohistochemical staining for each CARV infection are shown in figure 2.

Figure 2.

Representative immunohistochemistry from CARV infected human lung allografts. A) Case of Influenza LRTI demonstrating CXCL9, CXCL10, and CXCL11 expression by allograft hyperplastic bronchial epithelial cells, as well as by peribronchial mononuclear cells. CXCR3 was expressed by peribronchial infiltrating lymphocytes, as well as by bronchial hyperplastic bronchial epithelial cells. Examples of isotype control antibody demonstrate no significant nonspecific staining. B) Representative example of parainfluenza infection in a lung transplant recipient without lower respiratory tract symptoms, but with histopathologic evidence of bronchial epithelial reserve cell hyperplasia. CXCL9, CXCL10, and CXCL11 were expressed by these hyperplastic epithelial cells and by subepithelial mononuclear cells, as was CXCR3 (not shown). C) Representative example of RSV LRTI with strong expression of CXCR3 by hyperplastic bronchial epithelial cells and infiltrating mononuclear cells. Similar expression was seen for CXCL9, CXCL10, and CXCL11 (not shown). A representative low power view of RSV LRTI incubated with isotype control antibody (mouse IgG) demonstrates no nonspecific staining.

BALF CXCR3 Ligand Concentrations during CARV Infection are Associated with a Subsequent FEV₁ Decline

We determined whether CXCR3 chemokine ligand concentrations in BALF during CARV infection would be predictive of subsequent changes in FEV₁. Change in FEV₁ was calculated as the percent change from the last FEV₁ measure prior to CARV infection to a follow-up measure at approximately 6 months. The FEV₁ measurement closest to 180 days after CARV diagnosis was selected (median 174 days, range 150–213). Two patients had no follow-up FEV₁ measures available; one with BOS stage 3 died 140 days after CARV diagnosis due to BOS and was assigned a 6 month FEV₁ of 0 liters (i.e. −100% change); the second, also with BOS stage 3, was retransplanted 153 days after CARV diagnosis, and the pre-CARV FEV₁ measurement was carried forward (i.e. 0% change). We first correlated the log concentrations of CXCL9, CXCL10, and CXCL11 with change in FEV₁ over 6 months. For these analyses, we adjusted for the duration post-transplant for sample collection given a strong correlation between duration post-transplant and change in FEV₁. In these analyses, log concentrations of both CXL10 and CXCL11, but not CXCL9 correlated with the change in FEV₁ (Table 4).

TABLE 4.

CORRELATION OF BALF CXCR3 LIGAND LOG CONCENTRATIONS DURING CARV INFECTION WITH CHANGE IN FEV₁ AT 6 MONTHS

	Estimate	Rsq	Std. Error	P value
Log CXCL9	−3.66	0.23	4.99	0.48
Log CXCL10	−10.79	0.57	3.10	0.004
Log CXCL11	−15.66	0.59	4.28	0.003

^†Each chemokine submitted to a separate linear regression model, adjusted for the duration post transplant that sample was drawn.

Next we categorized CXCR3 chemokine ligand concentrations into high and low groups based upon the median values among CARV-positive recipients. Using these methods, high CXCL9, CXCL10, and CXCL11 concentrations were defined as greater than or equal to 5.05 ng/ml; 5.13 ng/ml; and 0.56 ng/ml; respectively. The median change in FEV1 at 6 months was significantly different between high and low concentration groups for both CXCL10 and CXCL11 (−14.9% vs. 3.0%, p=0.04; and −14.9% vs. 3.0%, p=0.01; respectively), but not CXCL9 (−13.4% vs. −5.9%, p=0.19) (Figure 3). Neither the presence of LRTI (−11.6% vs. −5.9%, p=0.49) nor co-infection with another pathogen was associated with a significant change in FEV₁ (−12.4% vs. −8.4%, p=0.37).

Figure 3.

FEV1 changes in 6 months by high and low concentration groups for CXCL9, CXCL10, and CXCL11. (A) For CXCL9, the median change in FEV1 was not significantly different for high (−13.4, IQR −15.8 to 4.0) and low (−5.9%, IQR −22.5 to 15.4) concentration groups (p=0.19). (B) For CXCL10, the median FEV1 declined in the high concentration group (−14.9%, IQR −43.2 to −10.7), and increased slightly in the low concentration group (3.0%, IQR −10.8 to 5.5), and this difference was statistically significant (p=0.04). Similarly, for CXCL11, the median FEV1 declined in the high concentration group (−14.9%, IQR −63.6 to −10.7) and increased slightly in the low concentration group (3.0%, IQR −10.8 to 5.5), and this difference was significant (p=0.01).

DISCUSSION

Several studies have suggested that CARV infection, especially CARV LRTI, is a risk factor for the development of BOS (3–5, 18), the most common cause of chronic lung allograft dysfunction. However, we have noted that not every instance of CARV infection, or CARV LRTI for that matter, leads to sustained allograft dysfunction. In other words, some patients with CARV infection make a full recovery while others develop progressive dysfunction manifest as a sustained decline in FEV₁. By itself, the presence or absence of LRTI does not sufficiently discriminate who develops progressive airflow obstruction and who recovers, underscoring the need for biomarkers to improve our ability to predict outcomes following CARV infection.

CXCR3/ligand biology is known to have a role in the recruitment of activated lymphocytes during respiratory viral infections (6–8), and during acute and chronic lung allograft rejection (9–11). Thus, we hypothesized that these chemokines would be a link between CARV infection and subsequent lung allograft dysfunction. We therefore explored alterations in BALF concentrations of CXCR3 chemokines during CARV infection. CXCL9, CXCL10, and CXCL11 were each markedly elevated during CARV infection, relative to matched control bronchoscopies and to the last bronchoscopy performed before the diagnosis of CARV.

Importantly, we found that hyperplastic bronchial epithelial cells and peribronchial infiltrating mononuclear cells were major cellular sources of these chemokines, although other cell populations in the BAL and interstitium may also play a significant role. These data suggest that respiratory viral infection stimulates epithelial cells to express interferon-inducible ELR-negative CXC chemokines that recruit CXCR3 expressing lymphocytes that may initiate and propagate lung allograft dysfunction.

Somewhat unexpectedly, CXCR3 was also expressed by bronchial epithelial cells. Interestingly, other investigators have shown that human bronchial epithelial cells express CXCR3 (19). Furthermore, these authors showed that when CXCR3 ligand concentrations are “low”, the predominant effect on the bronchial epithelium was proliferation, consistent with a reparative mechanism. On the other hand, a “high” CXCR3 ligand concentration inhibited epithelial cell proliferation, which may lead to airway denudation and damage (19). Collectively, these data suggest that high concentrations of CXCR3 ligands, as seen commonly during CARV infection in our cohort, may have direct deleterious effects on the allograft airway.

In order to provide useful prognostic information at the time of CARV diagnosis, we explored the utility of these chemokines in BALF as biomarkers of subsequent lung allograft dysfunction, manifest as a decline in FEV₁ 6 months after CARV. After adjustment for duration post-transplant, we report a linear correlation between log concentrations of both CXCL10 and CXCL11 with the percent change in FEV₁ 6 months after CARV diagnosis. Specifically, a high concentration of CXCL10 or CXCL11 predicted a decline in FEV₁ at 6 months. Although the concentration of CXCL9 was markedly elevated in BALF during CARV infection, there was no association with change in FEV₁. If our findings are validated in future studies, concentrations of CXCR3 chemokine ligands in BALF may offer useful prognostic information to clinicians and lung transplant recipients.

Our study has inherent limitations common to any retrospective study. First, the number of included CARV cases was only slightly greater than half of the total patients diagnosed with CARV during this study. This discrepancy was due to the requirement of an available BALF sample. We attempt to collect samples for research at the time of each bronchoscopy, but in most cases where BALF was not available, the bronchoscopy was done on a weekend, holiday, or late in the day, when staffing was not available to collect and/or process the specimen. We selected CARV-negative recipients based on a bronchoscopy performed at a duration post-transplantation similar to the time of CARV diagnosis in a matching CARV-positive recipient. Because our program does not perform surveillance bronchoscopy after the first year, all control bronchoscopies after this point were performed for clinical indications. These clinical indications may very likely be associated with an increased risk of subsequent allograft dysfunction. Also, our methods of viral identification during the study have a lower sensitivity for viral detection as compared to viral polymerase chain reaction (PCR) assays. Indeed, recent studies utilizing the viral PCR assay report an incidence of CARV infections in lung transplant recipients between 8–34% during a single season (18, 20), and 52% over a 3 year period (5). We detected CARV infections in approximately 10% of our cohort, and it is possible that cases of missed CARV infection were included as control subjects. Moreover, current PCR multiplex techniques can now identify numerous other viruses (eg. rhinovirus and metapnuemovirus) that were not detected by the techniques utilized during this study. However, the fact that we found significant differences despite these limitations only strengthens the confidence in our findings.

In summary, in this study we have shown that BALF concentrations of interferon-inducible CXCR3 chemokine ligands are increased during CARV infection. Importantly, these chemokine concentrations may provide useful prognostic information about the risk of poor outcomes following CARV infection, although this assertion requires validation. Furthermore, our study provides some insight into the immunologic mechanisms responsible for the link between CARV infection and lung allograft dysfunction. While CXCR3/ligand biology certainly plays an important role in controlling viral infection, antagonizing this pathway after the eradication of the infecting virus may be a promising target for limiting lung allograft dysfunction following CARV infection. Our findings represent an important first step, but ultimately additional investigation into the basic mechanisms responsible for the poor outcomes following CARV infection in lung transplant recipients is required so that the goal of novel prevention and treatment strategies can be achieved.

ABBREVIATIONS

BOS

bronchiolitis obliterans syndrome

CARV

community-acquired respiratory virus

BALF

bronchoalveolar lavage fluid

UCLA

University of California Los Angeles

RSV

respiratory syncytial virus

PFT

pulmonary function test

ISHLT

International Society of Heart and Lung Transplantation

FEV₁

forced expiratory volume in 1 second

AR

acute rejection

CPE

cytopathological effect

LRTI

lower respiratory tract infection

SD

standard deviation

AUC

area under the curve

PCR

polymerase chain reaction