Detection and Gram Discrimination of Bacterial Pathogens from Aqueous and Vitreous Humor Using Real-Time PCR Assays

Abstract

Purpose.: To develop and apply real-time PCR protocols to the detection and classification of the Gram status of bacterial pathogens in aqueous and vitreous humor collected from clinically suspected intraocular infections

Methods.: The analytical specificity of two PCR assays, SYBR Green 16S rDNA–Based Universal PCR (SGRU-PCR), and a Multiplex Gram-Specific TaqMan–Based PCR (MGST-PCR), was determined with 31 clinically important pathogens, including 20 Gram-positive and 11 Gram-negative. Analytical sensitivity was determined with a 10-fold dilution of Staphylococcus epidermidis and Escherichia coli DNA. Assays were further tested on aqueous (n = 10) and vitreous humor (n = 11) samples collected from patients with clinically diagnosed intraocular infections.

Results.: DNA was amplified from all control bacterial isolates when using SGRU-PCR. MGST-PCR correctly classified the Gram status of all these isolates. The SGRU-PCR limit of detection of S. epidermidis and E. coli DNA was 100 fg/μL (E = 0.82 and 0.86; r ² = 0.99) and for MGST-PCR, 1 pg/μL (E = 0.66 and 0.70; r ² = 0.99. For clinical intraocular samples, positivity of culture was 47.6% and for real-time PCR assays, 95.2%. Gram classification was achieved in 100% of MGST-PCR-positive samples. Among microbiologically negative samples, real-time PCR assays were positive in 90% of cases. The false-positive rate in control aqueous was 3.2%, and control samples of vitreous were negative.

Conclusions.: The real-time PCR assays demonstrated good correlation, with culture-proven results. With the use of these methods, bacterial detection was improved from 47.6% to 95.3%, demonstrating them to be sensitive, rapid tests for diagnosis of bacterial endophthalmitis.

Bacterial endophthalmitis is a rare but severe intraocular inflammation resulting from the introduction of microorganisms into the posterior segment of the eye. Its associated morbidity can have a dramatic impact on the patient's quality of life, causing a substantial decrease in the ability to perform normal daily activities and compromising psychological health. ¹  

Occasionally, clinical signs of intraocular inflammation result from a noninfectious mechanism, such as an immune response, chemical or physical aggression, vasculitis, or neoplasia. Therefore, the correct and early determination of the causative agent of inflammation is important for the prompt initiation of appropriate therapy. ²  

Conventional microbiology methods, such as Gram staining and culture, are routinely used for laboratory characterization of endophthalmitis cases. However, microbiologic methods are time consuming and show poor sensitivity for bacterial detection in aqueous and vitreous humor. ^{3 –6}  

Since the introduction of the PCR (polymerase chain reaction) methodology in 1983 by Kerry Mullis, the use of molecular methods in laboratory medicine has increased rapidly. Now, its use is part of the routine processing of many clinical samples in microbiology laboratories, establishing a new era for the diagnosis of infectious diseases. ⁷ The utilization of PCR for endophthalmitis diagnosis significantly increases the sensitivity of bacterial detection in aqueous and vitreous humor in less time. PCR and nested PCR protocols followed by post-PCR techniques such as RFLP (restriction fragment length polymorphism), DNA-probe hybridization, and DNA sequencing have been applied to the diagnosis of bacterial endophthalmitis. ^{4,6,8 –11}  

Real-time PCR technology is an enhancement of the original PCR design. It is a homogeneous method in which DNA amplification and detection of the target sequence occur together, decreasing handling of PCR products and risks of carryover contamination. ¹² Real-time PCR protocols for the detection of bacterial DNA in clinical samples of aqueous and vitreous humor have recently been described. ^{13 –15} However, broad-range 16S rDNA-based PCR and probe-based Gram classification with the use of real-time PCR technology for detection of bacterial pathogens from endophthalmitis cases has not yet been extensively described. 

The purpose of this study was to develop and apply a SYBR Green 16S rDNA–Based Universal PCR (SGRU-PCR) and a Multiplex Gram-Specific TaqMan–Based PCR (MGST-PCR) to the detection and Gram status classification of bacterial pathogens in aqueous and vitreous humor collected from eyes with clinically diagnosed endophthalmitis. 

Methods

Bacterial Isolates

A total of 31 bacterial isolates were used for standardization purposes, including American Type Culture Collection (ATCC; Manassas, VA) strains and clinical isolates preferentially isolated from ocular samples (Table 1). Species isolated from ocular samples and other sterile body sites stored in our bacterial collection were identified by standard microbiologic methods, and confirmation was performed with an automated system (Phoenix; BD Diagnostic System, Sparks, MD). 

Table 1.

 

Results of SGUR-PCR and MGST-PCR for Control Isolates Used for Determining Analytical Specificity

Organisms	Isolate Type	T m (°C) (SYBR Green)	TaqMan Multiplex
			GP-ProbeFAM	GN-ProbeVIC
Gram-positive
B. cereus	Clinical isolate	84.6	+	−
C. jeikeium	Clinical isolate	83.4	+	−
E. casseliflavus	ATCC 700327	82.8	+	−
E. faecalis	ATCC 29212	84.0	+	−
E. faecium	ATCC 93011	83.4	+	−
P. acnes	Clinical isolate	84.7	+	−
S. anginosus	Clinical isolate	83.8	+	−
S. aureus	ATCC 29213	83.4	+	−
S. bovis	Clinical isolate	84.1	+	−
S. cohini	ATCC 29974	84.3	+	−
S. epidermidis	ATCC 12228	83.4	+	−
S. haemolitycus	ATCC 29968	83.8	+	−
S. hominis	ATCC 29885	84.3	+	−
S. mitis	Clinical isolate	84.1	+	−
S. oralis	Clinical isolate	84.1	+	−
S. pneumoniae	ATCC 49619	84.3	+	−
S. pyogenes	ATCC 18615	84.4	+	−
S. sanguinis	Clinical isolate	84.1	+	−
S. viridans	Clinical isolate	83.8	+	−
S. xylosus	ATCC 29971	84.6	+	−
Gram-negative
A. baumannii	ATCC 19606	83.8	−	+
A. lwoffii	Clinical isolate	84.4	−	+
E. aerogenes	ATCC 13048	84.7	−	+
E. coli	ATCC 25922	85.0	−	+
H. infuenzae	ATCC 49247	83.8	−	+
K. pneumoniae	ATCC 700603	85.4	−	+
M. morganii	Clinical isolate	84.4	−	+
P. aeruginosa	ATCC 27853	83.1	−	+
P. mirabilis	Clinical isolate	83.9	−	+
P. rettgeri	Clinical isolate	84.9	−	+
S. marcescens	Clinical isolate	85.7	−	+
Fungi
F. solani	ATCC 36031	−	−	−
C. albicans	ATCC 90028	−	−	−

Clinical Samples

A total of 10 aqueous and 11 vitreous humor samples were collected from 14 patients with clinically suspected endogenous (n = 2) and postoperative (n = 12) endophthalmitis seen in our clinic between November 2007 and February 2009. All postoperative patients were using topical moxifloxacin or gatifloxacin at the time of sample collection. Only 1 of 12 patients showed delayed-onset endophthalmitis; all others had acute disease. Two patients with endogenous endophthalmitis were not using topical antibiotics. The study complied with the guidelines of the Declaration of Helsinki, and informed consent was obtained from all subjects. Aqueous and vitreous humor samples were obtained by means of anterior and posterior chamber paracentesis, respectively. Vitreous samples were also collected by vitrectomy through the pars plana. An aliquot of the samples was immediately used for Gram and Giemsa staining and cultured in blood agar, chocolate agar, and thioglycolate broth in aerobic and anaerobic conditions for 14 days. Approximately 100 to 500 μL of the samples was aseptically transferred to a sterile microtube and stored at −20°C for PCR procedures. 

DNA Extraction of Bacterial Isolates

Bacterial DNA was extracted using cultured isolates in 1.5 mL Luria-Bertani broth. The pelleted cells were treated with 450 μL TE buffer (Tris-EDTA, 10 mM Tris-Cl, [pH 7.4] and 1 mM EDTA [pH 8.0]) and boiled for 10 minutes. Afterward, 50 μL lysozyme (5 μg/μL) and 5 μL lysostaphin (20 μg/μL) were added, followed by incubation for 30 minutes at 37°C, followed by the addition of 40 μL proteinase K (20 μg/μL) and 100 μL STE buffer (10 mM Tris-HCl [pH 8.0], 1 mM EDTA[pH 8.0], 150 mM NaCl), and the samples were incubated for another 30 minutes at 37°C. After incubation, 120 μL 5 M potassium acetate was added, and the samples were incubated for 15 minutes at room temperature. The supernatant fraction obtained after 3 minutes of centrifugation (13,000 rpm) was transferred to a clean microcentrifuge tube, followed by the addition of 120 μL isopropanol. After centrifugation for 3 minutes, the pellet was washed with 1 mL 80% ethanol. The supernatant was discarded after centrifugation for 1 minute. DNA was resuspended in 50 μL sterile water (MilliQ; Millipore, Billerica, MA) and quantified by spectrophotometry (model ND-1000; Thermo Scientific, Wilmington, DE). Every bacterial DNA extract was diluted to a working concentration of 10 ng/μL. 

DNA Extraction of Clinical Samples

Total DNA was extracted from aqueous and vitreous humor with a DNA purification kit (QIAamp DNA mini kit; Qiagen, Hilden, Germany) according to the tissue protocol, with a few modifications. In brief, 50 μL of sample was mixed with 180 μL ATL buffer and immediately ground in liquid nitrogen for 1 minute. Extraction was achieved after treatment with proteinase K (10 minutes incubation at 56°C) and AL buffer (10 minutes incubation at 70°C). After spin purification, DNA was eluted with 100 μL EB buffer and stored at −20°C. 

Primer and Probe Design

16S rDNA sequences of the following organisms were obtained from GenBank (http://www.ncbi.nlm.nih.gov/blast/ provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD) and aligned on computer (SeqMan Lasergene Software Package; DNASTAR, Madison, WI): Acinetobacter baumannii (EU221389), Acinetobacter lwoffii (AY176770), Bacillus cereus (EF100618), Citrobacter diversus (AF025372), Citrobacter freundii (AF025365), Corynebacterium spp. (AY259129), Enterobacter aerogenes (AM933751), Enterobacter cloacae (AM778415), Escherichia coli (EU014689), Enterococcus faecium (AB326300), Haemophilus. influenzae (AF076035), Klebsiella oxytoca (AB353048), Klebsiella pneumoniae (AY043391), Morganella morganii (AY043390), Propionibacterium acnes (AF076032), Pseudomonas aeruginosa (EU090892), Proteus mirabilis (EU287466), Proteus vulgaris (AJ301683), Streptococcus anginosus (AY691541), Staphylococcus aureus (BA000017), Staphylococcus epidermidis (AM157417), Staphylococcus haemolyticus (EF522132), Staphylococcus lugdunensis (AB009941), Serratia marcescens (EU233275), Streptococcus mitis (EU200182), Streptococcus mutans (DQ677784), Streptococcus pneumoniae (AF003930), Streptococcus salivarius (AY188352) and Streptococcus viridans (AF076036). The universal primers NT-341Fw and 16S-522Rv were designed by aligning in the conservative region of 16S rDNA, yielding a product of approximately 192 bp. Gram-specific probes were designed by aligning sequences according to their Gram classification and were synthesized using different fluorescent agents to allow a multiplex approach. A primer set and a probe for the β-globin gene was designed for use as an internal control (Table 2). 

Table 2.

 

Primers and Probes

Name	Sequence (5′-3′)	Application	Product Size (bp)
Primers
NT-341Fw	GACTCCTACGGGAGGC	16S rDNA universal amplification	192
16S-522Rv	GCGGCTGCTGGCAC	16S rDNA universal amplification	192
Beta-Rv	CCAAGAGTCTTCTCTGTCTC	Endogenous control	100
Beta-Fw	GAAGTTGGTGGTGAGGCC	Endogenous control	100
Probes
GP-Probe	FAM - CTGAYSSAGCAACGCCGCG - MGB	Gram-positive classification	—
GN-Probe	VIC - CCTGAYSCAGCMATGCCGCG - MGB	Gram-negative classification	—
Beta-Probe	NED - AGGTTGGTATCAAGGTTACAAG - MGB	Endogenous control	—

 All primers and probes were developed for the present study.

PCR Optimization, Sensitivity, and Specificity

The two Gram-specific probes were initially tested in uniplex reactions to guarantee correct amplification using control strain DNA. The best primer annealing temperature was chosen by testing the primer set in a temperature gradient amplification. Afterward, the multiplex reaction was set up by applying different probe concentrations. Sensitivity was determined for S. epidermidis and E. coli DNA serially diluted from 10 ng/μL to 1 fg/μL. The reaction efficiency (E) was calculated after the construction of a standard curve with the formula E = 10^{(−1/−slope)} −1. 

The specificity of the Gram-specific reaction was determined after correct amplification of each validating strain DNA. To exclude cross-reaction with fungal pathogens, amplification was performed with Candida albicans (ATCC 90028) and Fusarium solani (ATCC 36031) DNA. Aqueous humor (n = 50) of phacoemulsification surgery and vitreous humor (n = 12) obtained from patients with noninfectious intraocular inflammation were used to calculate the proportion of false-negative results. 

Pretreatment of PCR Mixtures with DNaseI

The amplification of contaminant DNA from molecular biology reagents is a recognized drawback for universal 16S rDNA amplification for clinical purposes. ¹⁶ For universal bacterial detection in small DNA amounts from clinical samples by using real-time PCR platforms, the simultaneous amplification of sample and contaminant DNA can make it difficult to interpret the results of positive clinical samples. ¹⁷ Commercial Taq DNA polymerase preparations can carry a low quantity of E. coli or Thermus aquaticus DNA ^10,16 and also Pseudomonas spp., Sphingomonas spp., Moraxella spp., and Acinetobacter spp., in the real-time PCR mix. ¹⁸ Therefore, pretreatment of commercial Taq polymerase preparations to remove contaminating bacterial DNA is an essential step for the application of PCR protocols in clinical diagnosis. 

The real-time PCR mix can be pretreated using restriction endonuclease or DNaseI digestion, UV radiation or 8-methoxypsoralen. ^17,19,20 The Platinum SYBR Green qPCR Super Mix (Invitrogen) and TaqMan Universal PCR Master Mix (Applied Biosystems, Inc. [ABI], Foster City, CA) were incubated with 5.0, 2.0, 1.0, 0.5, 0.4, 0.3, 0.25, 0.1, and 0.01 U RQ1 RNase-free DNase for 30 minutes at 37°C followed by 50 minutes at 95°C. For each DNaseI concentration, three positive reactions using 10 ng/μL of S. epidermidis and E. coli DNA and three negative reactions were tested. 

Positive and negative control reactions using untreated reagents were also prepared, to allow the evaluation of positive (decontamination) and negative (decrease in efficiency) effects. Comparison of C_t (cycle threshold) values between treated and untreated control samples was made to choose the better DNaseI concentration for each real-time PCR mix treatment. 

SYBR Green 16S rDNA–Based Universal PCR

A 20-μL reaction was set up containing 10 μL Platinum SYBR Green qPCR Super Mix (Platinum Taq DNA polymerase; SYBR Green I dye; Tris-HCl; KCl, 6 mM MgCl₂, 400 μM dGTP, 400 μM dATP, 400 μM dCTP, 800 μM dUTP, uracil DNA glycosylase [UDG], and stabilizers), 0.2 μM of each primer NT-341Fw and 16S-522Rv (Table 2), and 2 μL template. Amplification was performed on a real-time PCR system (model 7500; ABI) with the following conditions: 50°C for 2 minutes and 95°C for 10 minutes (UDG) followed by 40 cycles of 95°C for 15 seconds and 60°C for 60 seconds. Melting curve analysis was performed after amplification, by heating samples from 68°C to 95°C, with a gradual increase of 0.5°C/s. The amplification of every five clinical samples was performed in the presence of a negative control, including all PCR reagents, except DNA template and positive control samples, with 10 ng/μL S. epidermidis and E. coli DNA. 

Multiplex Gram-Specific TaqMan–Based PCR

Multiplex real-time PCR for the differentiation of Gram-positive and Gram-negative pathogens was set up using the TaqMan-MGB method (ABI). The reaction consisted of 10 μL TaqMan Universal PCR Master Mix, 0.25 μM of each primer, NT-341Fw and 16S-522Rv, 0.15 μM GN-probe and 0.05 μM GP-probe (ABI), 2 μL of sample and MilliQ (Millipore) sterile water, to a final volume of 20 μL. Amplification was performed in the real-time PCR system (model 7500 ABI) with the following conditions: 50°C for 2 minutes and 95°C for 10 minutes (UDG) followed by 40 cycles of 95°C for 15 seconds and 60°C for 60 seconds. Amplification of the β-globin gene was performed for every clinical sample at the same time of bacterial detection. Negative and positive controls were run in the amplification of every five PCR clinical samples. 

Sequencing of Amplified Products

After the reaction, a 15-μL PCR aliquot was purified with a PCR purification kit (QIAquick; Qiagen). DNA molecules were sequenced in both directions with dye-termination chemistry (Big Dye Terminator; 3000 Genetic analyzer; ABI). The sequences obtained were edited (SeqMan; DNASTAR, Madison, WI) and searched for similarity in GenBank using the BLAST (blastn algorithm) program with automatically adjusted parameters. The minimal percentage to identify genus and species were used as recommended by CLSI document MM18-A. ²¹  

Results

Analytical Specificity and Sensitivity

Universal amplification was obtained by using the NT-341Fw and 16S-522Rv primer set for control bacterial testing. The T _m (melting temperature) variation range was determined to define an experimental range of possible results. In addition, the GP-probe and GN-probe were shown to be quite specific for Gram differentiation when tested with bacterial DNA control samples. No cross-reactivity with F. solani and C. albicans DNA was shown (Table 1). After prior DNaseI treatment, the limit of detection for S. epidermidis and E. coli DNA was 100 fg/μL for SGRU-PCR and 1 pg/μL for MGST-PCR. 

The efficiency of reactions determined using 10-fold dilutions of S. epidermidis and E. coli DNA was 0.82 and 0.86, respectively, with SGRU-PCR, and 0.66 and 0.70 for the MGST-PCR assay (Fig. 1). 

Figure 1.

 

Standard curve derived from SGUR-PCR and MGST-PCR for S. epidermidis (A) and E. coli (B) DNA. Calculations of efficiency for each reaction are shown. Efficiency was calculated from the slope given by regression analysis using the formula: E = 10^{(−1/−slope)} −1.

Pretreatment of Commercial PCR Mixtures

The treatment of PCR mixtures used in this study showed significant inhibitory effects with high concentrations of DNaseI. On the other hand, low concentrations that do not exhibit inhibitory effects are not sufficient to completely decontaminate the mix. The DNaseI concentrations required to eliminate false-positive amplification are shown in Table 3. In a 20-μL reaction volume, 0.4 U DNaseI for TaqMan Universal PCR Master Mix (ABI) and 0.25 U DNaseI for Platinum SYBR Green qPCR Master Mix (Invitrogen) showed efficient decontamination and a less inhibitory effect when comparing C_t values between treated and untreated mix. 

Table 3.

 

Comparison of C_t Values Using Pretreated and Untreated PCR Mix

DNase I/Mix (U)	Ct
	S. epidermidis			E. coli			NTC
TaqMan Universal PCR Master Mix
Untreated	17	17	17	15	15	15	32	32	32
1.0	26	26	26	25	24	25	Neg	Neg	Neg
0.5	22	23	23	19	20	19	Neg	Neg	Neg
0.4	22	22	22	17	18	18	Neg	Neg	Neg
Platinum SYBR Green qPCR Super Mix
Untreated	11	11	11	19	19	19	33	34	33
2.0	Neg	Neg	Neg	Neg	Neg	Neg	Neg	Neg	Neg
1.0	Neg	38	35	Neg	Neg	Neg	Neg	Neg	Neg
0.5	23	22	22	35	37	36	Neg	Neg	Neg
0.3	36	36	36	26	25	26	Neg	Neg	Neg
0.25	14	14	14	22	22	22	Neg	Neg	38

 TaqMan PCR master mix, ABI, Foster City, CA, and Platinum SYBR Green qPCR Super Mix, Invitrogen, Carlsbad, CA. NTC, nontemplate control; neg, negative.

Microbiologically Positive Samples

Among the 11 microbiologically positive samples collected from eight eyes, 10 showed positive culture isolation, including 3 S. viridans (two patients), 2 coagulase negative Staphylococcus (CNS) (one patient), 2 Weeksella virosa (one patient), 1 β-hemolytic Streptococcus, 1 P. acnes, and 1 S. pneumoniae. One aqueous humor sample (patient 6) showed only Gram-positive cocci. Every bacteriologically positive sample showed positive 16S rDNA detection by SGUR-PCR, with a C_t variation of 18.7 to 33.6 and T _m from 83.3°C to 86.5°C. Gram classification was achieved for all these samples by MGST-PCR. No mixed Gram-positive signal was obtained in five samples (C_t 20.9–33.8) with Streptococcus spp. isolation (patients 1, 3, 4, and 8). Gram-negative signal was obtained in four samples (C_t 23.5–33) including a case of endophthalmitis with positive culture for W. virosa (aqueous and vitreous humor of patient 7) and a case with CNS isolation (aqueous and vitreous humor of patient 2). Direct amplification from this cultured CNS strain showed only Gram-positive signal amplification, demonstrating that no cross-reaction with GN-probe occurred in the clinical sample. Two samples showed mixed Gram-specific amplification, and both were culture positive for Gram-positive organisms. 

Analysis of the sequenced amplicon allowed organism identification for almost all microbiologically positive samples. Only the amplification product from patient 2 did not permit organism identification due to poor-quality sequencing. The sequenced product from a case of delayed postoperative endophthalmitis with P. acnes showed similarity with 16S rDNA sequences of Actinobacteria spp. and Mycobacterium spp., but a RefSeq database search showed 100% identity with P. acnes sequence NC_006085.1 (E = 1e-81) and less than 91% identity with Actinobacteria spp. For the samples in which molecular Gram classification matched the microbiology results, identification using the sequenced products agreed with phenotypic species determination (Table 4). 

Table 4.

 

Results of the Real-time PCR Assays and Organism Identification Based on the 16S rDNA Sequencing of Clinical Aqueous and Vitreous Humor Samples

Patient	Sample	Category	Microscopy	Culture	SGRU-PCR		MGST-PCR (Ct)		Sequencing
					Ct	T m (°C)	GP-FAM Probe	GN-VIC Probe	Organism	Maximum Identity (%)
Microbiologically positive
1	Vitreous	PO	Gram-positivie cocci	Viridans Group Streptococci	20.5	84.0	25.0	Negative	Streptococcus mitis group*	100
1	Aqueous	PO	Gram-positivie cocci	Viridans Group Streptococci	26.7	84.0	33.8	Negative	Streptococcus mitis group	100
2	Vitreous	PO	ND	Coagulase-Negative Staphylococci	30.4	83.3	Negative	33.0	Bad file (mixed sequence)	ND
2	Aqueous	PO	ND	Coagulase-Negative Staphylococci	29.3	83.3	Negative	30.2	Bad file (mixed sequence)	ND
3	Vitreous	PO	Gram-positive cocci	Viridans Group Streptococci	22.6	84.3	24.1	Negative	Streptococcus spp.	99
4	Vitreous	PO	Gram-positive cocci	β-hemolytic Streptococci	18.7	83.6	20.9	Negative	Streptococcus dysgalactiae	100
5	Vitreous	Delayed PO	Negative	Propionibacterium acnes	33.6	86.5	30.0	34.9	P. acnes, Actinobacterium spp., Mycobacterium spp. †	100
6	Aqueous	PO	Gram-positivie cocci	Negative	26.5	83.6	30.2	29.2	Coagulase-negative staphylococci	100
7	Vitreous	PO	Gram-negative bacilli	Weeksella virosa	24.4	84.0	Negative	23.8	Flavobacteriaceae	99
7	Aqueous	PO	Gram-negative bacilli	Weeksella virosa	24.6	84.0	Negative	23.5	Flavobacteriaceae	99
8	Aqueous	PO	Gram-positivie cocci	Streptococcus pneumoniae	19.5	84.0	23.2	Negative	S. pneumoniae and S. mitis group	100
Microbiologically negative
9	Vitreous	Endogenous	Negative	Negative	28.1	83.3	Negative	27.9	Moraxella spp.	100
10	Vitreous	Endogenous	ND	Negative	Negative	Negative	Negative	Negative	ND	ND
3	Aqueous	PO	Negative	Negative	31.6	84.0	Negative	33.6	Bad file (mixed sequence)	ND
6	Vitreous	PO	Negative	Negative	20.9	83.3	23.9	Negative	Coagulase-negative Staphylococci	99
11	Vitreous	PO	Negative	Negative	25.2	84.0	31.6	Negative	Streptococcus mitis group	100
11	Aqueous	PO	Negative	Negative	28.7	84.0	31.5	31.3	Streptococcus mitis group	100
12	Vitreous	PO	Negative	Negative	34.3	79.0	Negative	31.9	Proteus spp.	99
12	Aqueous	PO	Negative	Negative	32.1	84.0	Negative	28.9	Bad file (mixed sequence)	ND
13	Aqueous	PO	Negative	Negative	30.9	84.0	36.8	31.0	Bad file (mixed sequence)	ND
14	Aqueous	PO	Negative	Negative	32.5	84.3	Negative	34.8	Bad file (mixed sequence)	ND

 PO, postoperative; ND, not determined.

*  Includes the following streptococci: S. mitis, S. oralis, S. cristatus, S. infantis, S. peroris, S. orisratti, and S. pseudopneumoniae. ^21,22

†  However, a RefSeq data base search showed 100% identity with P. acnes sequence NC_006085.1 (E = 1e-81) and less than 91% identity with Actinobacteria species and any one Mycobacterium hit.

Microbiologically Negative Samples

Ten clinical samples, five aqueous and five vitreous, collected from eyes with suspected intraocular infections were negative by culture and microscopy. Universal amplification by SGRU-PCR was positive in 9 (90%) of these 10 samples (C_t range, 20.9–34.3, and T _m, 79.0–84.3). Gram classification was achieved by MGST-PCR for all universal-positive PCR samples, where seven showed Gram-negative (C_t 27.9–34.8) and four Gram-positive (C_t 23.9–36.8) amplification signals. Two samples demonstrated mixed fluorescence signal (aqueous humor from patients 11 and 13). Nevertheless, sequenced amplification products of samples from patient 11 showed 100% identity with the Streptococcus mitis group. Only one sample (patient 10) was negative in the real-time PCR assays. Organism identification using sequenced PCR products was achieved in five samples and included one Moraxella spp., one CNS, one Proteus spp., and two Streptococcus mitis group (Table 4). 

Overall Clinical Samples

In total, 21 clinical samples collected from 14 eyes with suspected endophthalmitis were included. Of those samples, 10 were aqueous humor and 11 vitreous humor. Most cases of intraocular infection were postoperative (12/14; 85.7%), and one of them had delayed onset (P. acnes endophthalmitis case). Only two cases were categorized as endogenous endophthalmitis, both showing negative culture results. 

The sensitivity of microscopy and culture for bacterial detection in clinical samples was 44.4% and 47.6%, respectively. There was no significant difference in sensitivity for microbiologic detection between aqueous and vitreous humor. With the use of real-time PCR assays, the sensitivity of bacterial detection was increased to 95.3% (Table 5). 

Table 5.

 

Sensitivity of Assays Used to Detect Bacterial Pathogens from Aqueous and Vitreous Humor

Sample	Gram Staining	Culture	Real-time PCR*
Aqueous humor (n = 10)	4 (50.0)	4 (40.0)	10 (100)
Vitreous humor (n = 11)	4 (40.0)	6 (54.5)	10 (90.9)
Total (n = 21)	8 (44.4)†	10 (47.6)	20 (95.3)

 Data are expressed as the number (%) of samples positive.

*  Results of SGUR-PCR and MGST-PCR.

†  In total, 18 clinical samples were Gram stained.

Gram classification was achieved for every universal PCR-positive clinical sample (n = 20, 100%). Seven samples showed only a Gram-positive signal, nine were only Gram-negative, and four demonstrated mixed amplification (Table 6). When the Gram classification results based on phenotypic methods and MGST-PCR were compared, there was total agreement (no mixed signal samples included) in 63.6% of the cases. 

Table 6.

 

Positivity of SGRU-PCR and Gram Differentiation Achieved by MGST-PCR According to the Classic Microbiology Results

Microbiology Results*	SGRU-PCR	MGST-PCR
		Gram-Positive	Gram-Negative
Positive (n = 11)
Gram-positive (n = 9)	9	7	4†
Gram-negative (n = 2)	2	0	2
Negative (n = 10)	9	4‡	7‡
Total PCR positive	20	11	13

*  Results of culture and microscopy (Gram staining).

†  Two CNS endophthalmitis samples showed Gram-negative signal.

‡  Two samples with mixed Gram-specific signal.

Control Samples

Among the 50 aqueous humor samples collected at the end of phacoemulsification procedures, 2 (4%) samples showed universal 16S rDNA amplification by SGUR-PCR (C_t 39). These two samples did not show a Gram-specific amplification fluorescence signal by MGST-PCR. Amplification of 16S rDNA in all 12 control vitreous humor samples was negative with the real-time PCR assays. 

Discussion

Routinely, the aqueous and vitreous humors are collected from patients with suspected endophthalmitis and processed by culture and microscopy techniques. However, the sensitivity of bacterial detection using these methods can be lower than molecular methods. Gram stain and culture of aqueous and vitreous humor have shown a sensitivity varying between 10% and 56%. ^{3 –6} Moreover, routine microbiologic methods are time consuming, leading to the inappropriate use of broad-spectrum antibiotics for several days. 

The use of a faster and more sensitive method, such as PCR, increases the positivity of bacterial detection in aqueous and vitreous samples and demands less time for the diagnosis if performed in a routine manner. Several investigators have reported the use of the PCR methodology with a significant increase in the percentage of bacterial detection in clinical samples of aqueous and vitreous humor, ranging from 40% to 100%. ^{3 –6,9,11,23,24} In most of these studies, a nested PCR strategy was used, to increase the sensitivity of bacterial detection. However, because of the high sensitivity of this method and the transfer of first-reaction products to a second-reaction tube, the risks of contaminant DNA amplification and cross contamination are substantially increased. ²⁵  

In addition, PCR protocols that include 16S rDNA-based Gram-specific primer and probe sets may help in choosing a more direct antibiotic therapy and have been applied in different settings. ^{15,26 –32} Although nested PCR multiplex and PCR-DNA probe hybridization approaches for Gram differentiation have been developed and applied to the diagnosis of endophthalmitis, ^8,33 both methods include at least two post-PCR steps, increasing the turn-around time of the test, the expense of reagents, and risk of cross-contamination. 

The use of real-time PCR for Gram classification directly on clinical samples allows simultaneous amplification and detection by the use of Gram-specific probes labeled with different fluorescent agents in a closed-system reaction. Gram differentiation using the real-time PCR technique has been described as a complementary laboratory technique for the diagnosis of infections in intensive care unit patients, ¹⁷ sepsis, ^32,34 infectious arthritis, ¹⁵ and urinary tract infections in hospitalized patients. ³⁰ Recently, real-time PCR using a universal probe, Gram-positive probe, and eight genus-specific probes was developed for the diagnosis of bacterial endophthalmitis. ¹³  

In this study, we describe two coupled real-time PCR reactions for universal detection and Gram differentiation of bacteria that causes endophthalmitis. The two sets of reactions showed excellent analytical specificity when tested for the most common ocular bacterial pathogens (Table 1). No cross-reaction was detected when using human and fungal DNA. The limit of detection for S. epidermidis and E. coli DNA was 100 fg/μL when using SGRU-PCR and 1 pg/μL when using MGST-PCR. Even with the lowest concentration of DNaseI enzyme, the master mix pretreatment caused considerable inhibition of the reaction. Efficiency of reaction was approximately 0.8 for SGRU-PCR and 0.70 for MGST-PCR assays. The decrease in analytical sensitivity was attributable to the reduction of Taq polymerase activity, caused by DNaseI inactivation at 95°C for 50 minutes and also to residual DNaseI activity. ^19,20,35 The MGST-PCR assay showed lower efficiency of reaction than did SGUR-PCR, most likely because of the simultaneous use of two probes. ^24,36,37  

The real-time PCR assays developed in this study were shown to be more sensitive for bacterial detection in aqueous and vitreous humor, compared with routine microbiology techniques. Among the 21 clinical samples included, 44.4% were positive by Gram staining, 47.6% were culture positive and 95.3% were positive with real-time PCR assays. A slight difference in the positivity of culture for vitreous samples (54.5%) compared with that for aqueous humor (40.0%) was observed. Reports demonstrating a better bacterial recovery when culturing vitreous for endophthalmitis diagnosis purposes have been published. ^5,6,38,39 A significant difference in the positivity of PCR for bacterial detection in aqueous and vitreous humor was not seen in our work. All aqueous humor samples were PCR-positive, and only one vitreous sample was negative. In general, no significant differences in bacterial detection in aqueous versus vitreous samples have been observed when employing PCR protocols. ^{4 –6,24} A multicenter study that included 100 patients with acute postoperative endophthalmitis showed a slightly lower sensitivity of PCR for aqueous humor (34.6%) compared with vitreous fluid (57%). However, in samples collected after initiation of antibiotic therapy, vitreous humor (70.1%) showed more PCR-positive results than did aqueous (23.4%), when compared with the first collected sample from patients with endophthalmitis not treated with intravitreous antibiotic. ⁹  

Although the sensitivity of the real-time PCR assays for bacterial detection was similar for aqueous and vitreous samples, poor-quality sequencing (mixed sequences) was more frequently seen in aqueous samples. Analyses of 16S rDNA sequenced product obtained from aqueous humor in a previous report demonstrated unsatisfactory quality when compared with PCR products from vitreous samples. ⁶  

The isolation of bacterial pathogens from anterior chamber aspirates at the end of intraocular surgery ranges from 2.0% to 46.25%. ^{40 –44} Therefore, the poorer sequencing data yielded for aqueous humor could be the result of mixed amplification from the real pathogen causing infection and a contaminant organism. Even though no considerable sensitivity difference was noted for PCR bacterial detection in the aqueous and vitreous in this and other studies, a vitreous fluid aspirate is preferable for molecular detection and organism identification in cases of postoperative endophthalmitis. 

Molecular Gram classification using MGST-PCR was achieved in 100% of the aqueous and vitreous humor specimens. Among the Gram-positive, culture-proven samples, 77.7% were correctly classified by the GP-probe when compared with phenotypic results. Two samples (aqueous and vitreous from the same patient) showing a Flavobacteriaceae culture growth were correctly classified by the GN-probe. Four samples showed a mixed amplification signal, but three of them yielded quality sequencing data, where Gram-positive organisms were identified. Therefore, although the GN-probe showed 100% analytical specificity, cross-reactions with Gram-positive organisms could have occurred in a small number of clinical samples. 

SGRU-PCR and MGST-PCR were shown to be particularly helpful in bacterial detection in culture-negative samples. Among 10 microbiologically negative samples, nine (90%) were PCR positive, and five gave an interpretable and searchable sequence data. The organisms identified included one Moraxella spp., one CNS, and two Streptococcus mitis group, recognized causative agents of endophthalmitis. Sequencing of PCR-positive/culture-negative samples also included the identification of a Proteus spp. causing postoperative endophthalmitis, which is an uncommon but possible agent in postoperative endophthalmitis. ^38,45,46  

Rates of false-positive amplification were low (3.2%) in this study. Only two aqueous humor samples collected at the end of phacoemulsification showed a positive amplification signal by SGUR-PCR, with a high C_t value (39), almost five more than the highest C_t value (35) found for the clinical samples used in this assay. The use of a larger number of clinical samples could help to establish a cutoff value. For now, our results encourage us to set a cutoff of 35 cycles for the SGUR-PCR assay. 

The results presented show that the real-time PCR assays designed in this study provide faster and more sensitive testing for the laboratory diagnosis of bacterial endophthalmitis, helping ophthalmologists to differentiate bacteria from other infectious etiologies of intraocular inflammation and giving additional information about the Gram status of the causative microorganism in a homogenous and closed amplification system. This method could be useful for the prompt initial antibiotic therapy in endophthalmitis cases avoiding intraocular toxicity due to the use of unnecessary antibiotics. Further clinical studies with well-controlled case series to evaluate clinical performance including diagnostic sensitivity and specificity, predictive values, and diagnostic accuracy are warranted. 

Footnotes

 Supported by a grant from Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), Brazil.

Footnotes

 Disclosure: P.J.M. Bispo, None; G.B. de Melo, None; A.L. Hofling-Lima, None; A.C. Campos Pignatari, None