Research ArticleDiagnostics

T2 Magnetic Resonance Enables Nanoparticle-Mediated Rapid Detection of Candidemia in Whole Blood

See allHide authors and affiliations

Science Translational Medicine  24 Apr 2013:
Vol. 5, Issue 182, pp. 182ra54
DOI: 10.1126/scitranslmed.3005377


Candida spp. cause both local and disseminated infections in immunocompromised patients. Bloodstream infections of Candida spp., known as “candidemia,” are associated with a high mortality rate (40%), which is mainly attributed to the long diagnostic time required by blood culture. We introduce a diagnostic platform based on T2 magnetic resonance (T2MR), which is capable of sensitive and rapid detection of fungal targets in whole blood. In our approach, blood-compatible polymerase chain reaction is followed by hybridization of the amplified pathogen DNA to capture probe–decorated nanoparticles. Hybridization yields nanoparticle microclusters that cause large changes in the sample’s T2MR signal. With this T2MR-based method, Candida spp. can be detected directly in whole blood, thus eliminating the need for analyte purification. Using a small, portable T2MR detection device, we were able to rapidly, accurately, and reproducibly detect five Candida species within human whole blood with a limit of detection of 1 colony-forming unit/ml and a time to result of <3 hours. Spiked blood samples showed 98% positive agreement and 100% negative agreement between T2MR and blood culture. Additionally, performance of the assay was evaluated on 21 blinded clinical specimens collected serially. This study shows that the nanoparticle- and T2MR-based detection method is rapid and amenable to automation and offers clinicians the opportunity to detect and identify multiple human pathogens within hours of sample collection.


Candidemia is a fungal bloodstream infection where rapid diagnosis is needed. Candida spp. cause a variety of infections ranging from local to invasive and disseminated infections in immunocompromised patients. Bloodstream infections of Candida spp. raise concerns among clinicians owing to their high mortality rate (40%), which is mainly attributed to the long diagnostic time required by blood culture, with half of the cases diagnosed postmortem (13). Consequently, antifungal therapies are often used when not necessary (4). Candidemia is either an indication of disseminated infection or a result of colonization of an indwelling intravenous catheter that requires immediate antifungal therapy. Studies have shown that initiation of antifungal treatment in less than 12 hours can reduce the mortality rate from 40 to 11% (2, 3). Thus, early diagnosis is crucial in the management of candidemia (4).

The current clinical gold standard for the diagnosis of candidemia is blood culture, which requires 2 to 5 days of culture growth followed by morphological examination and biochemical testing of the cultivated microbes. Blood culture has a low sensitivity of about 50% (5) and requires viable circulating Candida cells. Previous efforts to develop a Candida diagnostic assay that circumvents the inherent limitations of blood culture have used conventional molecular diagnostic techniques, such as polymerase chain reaction (PCR) or reverse transcription PCR (RT-PCR), on DNA isolated from whole blood or serum. These methods require time-consuming fungal DNA extraction and purification steps (2 to 24 hours) or are implemented after blood culture (6, 7). Their reported clinical sensitivities range from 13% (8) to 100% (9), but few of these molecular diagnostic tests have been translated to actual products. Further, nucleic acid isolation for standard PCR methods is relatively complex and can lead to false-positive and false-negative results caused by laboratory or sample contamination and DNA extraction inconsistencies (10). The complexity of extensive sample preparation sets barriers against the clinical use of PCR for whole-blood pathogen detection (10).

Other methodologies have been explored for rapid diagnosis of candidemia, including tests based on both antigen and host immune responses. A serological antigen assay from Fungitell has been approved by the U.S. Food and Drug Administration for detection of (1,3)β-d-glucan (BDG), which is a cell wall component of many fungal pathogens, including Aspergillus, Candida, and Fusarium. However, even at BDG levels of >80 pg/ml, the clinical sensitivity of the assay ranges from 50 to 66% (Fungitell, Cape Cod Associates Inc.). Additionally, several home-brew Candida antibody tests are currently in use; however, interpretation of the results is confounded by the 20 to 30% positive detection rates in healthy individuals and suppressed antibody responses found in immunocompromised patients. Therefore, although these assays have the potential to assist in diagnosis, there remains a need for a rapid test with better sensitivity and specificity.

T2 magnetic resonance (T2MR)–based biosensing has been used for the detection of a wide range of target types, including proteins, drugs, pathogens, enzymes, and cells, in a variety of complex matrices, such as culture media, cell lysate, urine, serum, plasma, sputum, fine-needle aspirates, and blood (11, 12). T2MR reduces both time to result and overall test complexity, as recently shown for multiplexed analysis of human tumor fine-needle aspirates (11). Despite these promising reports, T2MR-based detection has been limited in sensitivity to analyte concentrations of ≥1000 targets/ml (1114) and have not produced clinically relevant assay performance in whole blood (12, 15).

To overcome these limitations and produce a robust diagnostic test for blood-borne pathogens, we developed whole blood–compatible nanoparticles that produced %ΔT2 switches higher than 1000% as a result of analyte binding. Other such particles that produce T2 switches have reported no higher than 54% T2 switch (16). Additionally, we coupled the nanoparticle-based hybridization with PCR amplification and a detector that could function in whole blood. This improved the detection sensitivity and turnaround time over that of conventional PCR by eliminating common sample preparation steps. This application of T2MR to the field of Candida diagnostics resulted in a >10-fold decrease in time to result while achieving detection sensitivities of ~1 colony-forming unit (CFU)/ml compared to Candida blood culture sensitivities of between 1 and 100 CFU/ml (1719). This diagnostic method is rapid and amenable to automation and offers clinicians the opportunity to detect multiple human pathogens within complex biological specimens.


Assay design

We designed an assay that lyses the Candida cells, amplifies the DNA, and finally detects the amplified product directly in the whole-blood matrix by amplicon-induced agglomeration of superparamagnetic nanoparticles (Fig. 1A). Nanoparticle clustering yields changes in the T2 relaxation time, making it detectable by MR. We designed a small, portable T2MR instrument for rapid yet highly precise T2 relaxation time measurements in standard PCR tubes. The T2MR instrument performance was measured using a stable standard sample with a mean T2 value of ~200 ms as described in Materials and Methods. The intra-assay coefficient of variation (CV) on a single T2MR instrument was 0.23%. For the same sample, the interassay CV was 0.45%, and the CV for T2 detection across multiple T2MR instruments was 2.57%. Unlike previously reported relaxometers, the T2MR instrument is compatible with standard PCR laboratory disposables (that is, 0.2 ml of polypropylene PCR tubes), making sample manipulation and loading straightforward.

Fig. 1 Detection of analytes within biological matrices with nanoparticle biosensors and a custom MR detector.

(A) Assay workflow for detection of Candida with T2MR. (B) Schematic depicting the T2MR detection particle reagent. Oligonucleotide probes are covalently conjugated to superparamagnetic nanoparticles. For each target, two populations of nanoparticles were generated, with each bearing a distinct target-complementary probe. Upon hybridization to the target strand (amplified in excess by asymmetric PCR), the nanoparticles cluster. The extent of clustering increases with the target DNA concentration. (C) Scatter plot showing T2 detection of a range of concentrations of synthetic oligonucleotide identical in sequence to the PCR amplicon for C. albicans spiked into human blood and buffer (phosphate-buffered saline with 0.1% bovine serum albumin and 0.1% Tween 20). Four independently spiked samples were prepared per concentration of DNA for blood or buffer. Data are means of n = 4 ± 2 SDs.

Lysis of Candida cells was performed by standard mechanical bead beating. For whole-blood PCR amplification, we used a proprietary formulation of a mutant enzyme (T2 Biosystems Inc.), which works similar to other thermostable polymerases (2024). We designed pan-Candida PCR primers to amplify the intervening transcribed spacer 2 (ITS2) region within the Candida ribosomal DNA operon, which is present at 50 to 100 copies per genome. Subsequent speciation was achieved through hybridization of particle-bound, species-specific capture probes nested within the pan-Candida amplicons (Fig. 1B and table S1). The PCR was designed to generate an asymmetric amplicon, enriched for one strand to facilitate probe hybridization. This overall primer and probe design strategy has been used by others (7, 25, 26) to enhance sensitivity over conventional PCR, but not in whole-blood assays. Superparamagnetic particle agglomeration was applied to detect the amplified product, wherein DNA hybridization of the amplified product to the superparamagnetic nanoparticles switched the particles between dispersed and clustered states.

Whole-blood detection of Candida

T2MR can be used for detection of nucleic acids by synthesizing a mixture of two populations of superparamagnetic particles, each functionalized with different oligonucleotide capture probes (Fig. 1B): one conjugated to a capture probe hybridizing to the 5′ end of the single-stranded DNA target and one hybridizing to the 3′ end of the DNA target. Upon hybridization to the oligonucleotides, the target DNA promotes nanoparticle cluster assembly, which leads to a change in the T2 relaxation rate (27). As shown in Fig. 1C, the dose-response curves for titrations of oligonucleotide target in buffer and blood matched, demonstrating lack of interference from heme or other background materials in blood with the particle clustering reaction. Because the T2 relaxivity of blood is much lower than that of superparamagnetic particles, the measured T2 relaxation rates of a blood sample containing a suspension of superparamagnetic particles are dominated by the clustering state superparamagnetic particles, as previously reported (12) and as demonstrated by comparison of a T2MR titration curve in buffer versus blood (Fig. 1C).

Nanoparticle sensors were designed and qualified to minimize inter- and intra-assay CV while maximizing the sensitivity of analyte detection. To date, T2MR nucleic acid detection sensors have used nanoparticles with diameters of <100 nm. These sensors clustered in the presence of the target nucleic acid and were shown to produce %ΔT2 values up to 18% (27). A confounding attribute of particles <100 nm is that clustering can produce both T2MR signal decreases and increases depending on the size regime of particle clusters, leading to unreliable particle sensors (28). On the basis of these observations, we screened a 1050-nm particle (MyOne, Invitrogen) and a proprietary 800-nm particle (T2 Biosystems). Particles were conjugated to capture probes, and their functional performance was measured by hybridization to a complementary oligonucleotide target. Particles with a diameter of 1050 nm yielded variable T2MR signal (CV = 2.3% for zero analyte) with moderate overall %ΔT2 of <20% (fig. S1A). Particles with a diameter of 800 nm yielded more stable T2MR values (CV = 0.7% for zero analyte) with a large overall %ΔT2 of >300% (fig. S1B). Furthermore, baseline %ΔT2 signals were indistinguishable between four 800-nm particle lots, and overall, %ΔT2 was similar across the linear dynamic range (fig. S1B).

The conjugation procedure was optimized to achieve maximum oligonucleotide probe density. Capture probe density was indirectly measured via hybridization, release, and quantification of a Cy5-labeled complement (fig. S2). The concentration of captured oligonucleotide was divided by the particle concentration to obtain the number of functional capture probes per particle, which was ~4 × 105 to 5 × 105 capture probes per particle.

The 800-nm carboxylated particles were conjugated to capture probes complementary to the ITS2 region of the C. albicans genome to investigate the feasibility of whole blood–compatible detection. T2 dose-response curves were generated with titrations of synthetic single-stranded oligonucleotide target identical in sequence to C. albicans ITS2 in both a final concentration of 40% blood and saline (Fig. 1C). The assay performs similarly in blood and buffer, suggesting that the whole-blood matrix does not interfere with the particle clustering reaction or T2MR measurement. Non-optical methods such as T2MR are capable of analyte measurements in opaque matrices such as whole blood.

The performance of the detection nanoparticles was assessed by evaluating the percent change in signal for addition of a given amount of analyte relative to signal baseline. Because T2 values are relaxation times that combine multiplicatively, one should convert T2 values to relaxation rates (R2 = 1/T2) before addition or subtraction. Percent relative change in R2 (%ΔR2) was calculated by dividing the difference of two signal values by the smaller of the two values. This was found to be equivalent to a previously used term defined as %ΔT2 according to Eq. 1:%ΔR2i=R20R2iR2i=1T2o1T2i1T2i=T2iT2oT2o=%ΔT2i(1)where %ΔR2i is the %ΔR2 of the ith data point, %ΔT2i is the %ΔT2 of the ith data point, T2i is the measured T2 relaxation value for the ith data point, and T20 is the T2 value measured for zero analyte concentration. If more than one T2 measurement was obtained for zero analyte concentration, the average of those measurements was used for T20. The percent change is a normalized measure of the magnitude of the T2 switch and can be useful in comparing results from different particle detection systems where baseline T2 values may vary, and differs from normalizing to the maximum T2 value as previous reports have done (16, 29). Dividing %ΔT2 by the measured %CV value of samples with zero target yields a measure of the T2MR detection reaction signal-to-noise ratio.

Percent change values similar to those shown in Fig. 1C were observed for detection of Candida, where %ΔT2 values were typically 1000% (Table 1). The measured signal-to-noise ratio for T2MR detection of C. albicans (1 CFU/ml) was ~200, as calculated by dividing the average %ΔT2 (1226%) from 16 replicate measurements of C. albicans (1 CFU/ml) by the %CV value (6.1%) for 16 replicates of negative blood (Table 1).

Table 1 Limits of detection for five Candida species.

Average and %CV of the measured T2 values were calculated for all measurements of 0 CFU/ml. Average and %CV of the measured T2 values were calculated for positive spikes only if the T2 values were above the cutoff. Samples are Candida spp.–spiked in human whole blood (n = 16 per concentration).

View this table:

Candida assay performance on spiked samples

Assay performance was characterized using whole-blood samples from healthy donors that were spiked with known concentrations of cultured Candida cells. A single operator repeatability study was conducted in which the same samples were run on the same instrument over the course of 10 days. A range of spiked C. albicans concentrations were used: 3 × 104, 3 × 103, 300, 60, 3, and 0 CFU/ml (Table 2). The CVs in measured T2 values were less than 12.8% across all spike levels [n = 30 measurements per concentration (in triplicate per day over the course of 10 days)].

Table 2 Repeatable T2MR detection of C. albicans in human whole blood.

Data are averages (n = 30 per concentration).

View this table:

Subsequently, the limit of detection (LoD) of the assay was determined for five different Candida spp. using 16 replicates for each species prepared at concentrations of 0, 1, 2, and 3 CFU/ml. The LoD was defined as the lowest amount of Candida in a specimen that could be detected with 95% probability. A LoD of 3 CFU/ml was observed for C. albicans and C. tropicalis; a LoD of 2 CFU/ml was observed for C. krusei and C. glabrata; and a LoD of 1 CFU/ml was observed for C. parapsilosis (Table 1).

The impact of sample matrix effects was measured across 105 healthy patients whose blood was spiked with C. albicans and C. glabrata (1 CFU/ml) and C. parapsilosis (6 CFU/ml) (table S2). Detection rates were 39 of 43 of the C. albicans–spiked specimens, 42 of 44 of the C. glabrata–spiked specimens, and 18 of 18 of the C. parapsilosis–spiked specimens. No impact on the assay’s detection sensitivity and reproducibility was observed, indicating no matrix interference across healthy donors. C. albicans, C. glabrata, C. krusei, and C. parapsilosis (1 CFU/ml) and C. tropicalis (2 CFU/ml) were spiked into blood samples from 94 infected patients suspected for sepsis (table S3). Detection rates were 25 of 28 for C. albicans, 11 of 11 for C. tropicalis, 17 of 18 for C. glabrata, 10 of 14 for C. krusei, and 23 of 23 for C. parapsilosis, with no observable impact on the assay’s reproducibility, indicating no matrix interference across infected donors. The assay cutoff was determined by measuring the T2 values in 50 Candida-negative specimens obtained from 50 patients suspected of sepsis (table S4).

Additionally, interference studies for 12 endogenous substances and 13 exogenous substances (table S5), including 11 antibacterial or antifungal compounds, were completed (fig. S3). No significant change in assay performance was observed for any of the conditions tested (P = 0.100, two-sample t test).

T2MR comparison with blood culture

The current standard for clinical candidemia diagnosis is blood culture. In vitro–spiked healthy donor whole-blood specimens were prepared using laboratory reference strains for C. albicans and C. krusei and clinical isolates of C. albicans at concentrations of 0, 33, and 100 CFU/ml. There was 98% positive agreement and 100% negative agreement between T2MR and blood culture (Table 3).

Table 3 Agreement of T2MR with blood culture using spiked whole-blood samples.

Data are number of samples.

View this table:

Performance on blinded clinical samples

A proof-of-principle evaluation of the assay’s clinical performance was conducted with a set of 24 fully blinded whole-blood patient specimens from patients who presented with symptoms of septicemia. Eight samples were from candidemic patients (n = 3), eight were from bacteremic patients (n = 8), and eight were from blood culture–negative patients (n = 8). The measured %ΔT2 values for all 24 patient samples and confirmed sample identifications are shown in Table 4. A decision threshold of 5 SDs above the mean T2 signal of all measured Candida-negative detection reactions was used (table S4). Detection reactions were grouped on the basis of Infectious Diseases Society of America guidelines (30) such that three results were reported as follows: C. albicans– or C tropicalis–positive, C. krusei– or C. glabrata–positive, and C. parapsilosis–positive. In one sample (sample 3) with a high C. albicans signal, some low cross-reactivity was observed for detection with the C. parapsilosis particles. Subsequent investigation identified that the 5′ capture probe for C. parapsilosis had two mismatches when bound to C. albicans amplicon. Redesign of the C. parapsilosis capture probe eliminated this cross-reactivity (fig. S4).

Table 4 Clinical identification and T2MR results for blinded clinical specimens from patients suspected of septicemia.

Data are from 24 patients. “A/T” indicates the detection particle for C. albicans and C. tropicalis, “P” indicates the detection particles for C. parapsilosis, and “K/G” indicates the detection particle for C. krusei and C. glabrata. T2MR values higher than the cutoffs of their respective particles are in bold.

View this table:

Additional serially drawn specimens from three patients were evaluated with both T2MR and blood culture (n = 21 specimens total). C. albicans infection was detected by T2MR before administration of antifungal therapy in all cases (Table 5). The time to result for T2MR measurements was ~2 hours, and the time to result for blood culture was on average 48 hours. After antifungal administration, all surveillance blood culture results became negative within 1 day of starting the antifungal agent, whereas T2MR results remained positive. In two cases, T2MR results showed a persistent presence of Candida cells even in the presence of an administered antifungal agent. These two patients did eventually recover. For one case, the T2MR signal of one sample dropped below the cutoff on day 5, suggesting an absence of Candida cells in that sample. In all cases, the measured T2MR values were above a cutoff established as 5 SDs above the mean of the negative specimens from 50 unhealthy patient samples (table S4). The same cutoff that was used for C. albicans in Table 1 was used for these clinical samples, which was the average baseline T2 value plus 5 SDs of the baseline.

Table 5 T2MR results for clinical specimens from three patients infected with C. albicans compared to blood culture and timing of administration of antifungal medication.

Each day, multiple samples from each patient were measured with blood culture (BC) and T2MR. Patient 1 had five BC samples and five T2MR samples measured. Patient 2 had five BC samples and four T2MR samples measured. Patient 3 had 11 BC samples and 12 T2MR samples measured. A total of 21 BC samples and 21 T2MR samples were measured for all patients. Samples were counted as positive (pos) for T2MR if the measured %ΔT2 value was above the cutoff of 25%. Days the patient was on antifungal therapy are indicated. Positive samples and T2MR values higher than the cutoffs of their respective particles are in bold. n/a, not available.

View this table:


The low abundance of circulating Candida cells (<10 CFU/ml) (10, 31), and the challenges associated with pathogen enrichment, DNA isolation, and purification from whole blood, has yielded PCR-based assays with compromised sensitivity, specificity, and time to result (10, 32). For example, only one Candida PCR test known to the authors has had its LoD measured with adequate number of replicates (n ≥ 16), and the assay had a LoD of ≥30 CFU/ml for all five Candida species (8). In addition, speciation of Candida is necessary to determine the appropriate therapeutic intervention based on the known toxicity and efficacy profiles for antifungal therapeutic compounds. Although the percentage of C. albicans that is resistant to fluconazole is as low as 1.5% (33), C. krusei is intrinsically resistant to fluconazole owing to a modified cytochrome P450 isoenzyme (34) and C. glabrata exhibits increased resistance to azoles caused by drug efflux (35).

We developed a whole-blood T2MR Candida assay capable of detecting five clinically important species of Candida that leveraged the advantages of non-optical detection to eliminate analyte purification, thus enabling more rapid turnaround times than previously reported Candida PCR assays (≤12 hours) and than the current clinical standard of blood culture (≤5 days). Here, time to result on spiked samples for T2MR was 3 hours, whereas for blood culture, it was ~2 days for C. albicans and ~1 day for C. krusei. Asymmetric PCR was used to specifically amplify the ITS2 region of the Candida genome directly in whole blood to achieve clinically relevant detection sensitivities. A T2MR detection method was developed in which two pools of oligonucleotide-derivatized nanoparticles hybridize to each end of the single-stranded amplicon. The amplicons serve as interparticle tethers and induce nanoparticle agglomeration yielding very large (%ΔT2 = ~1000%) changes in the T2 relaxation rate of the protons in water molecules that can be measured precisely (%CV = 0.12%) by the portable T2MR instrument. The largest previously reported %ΔT2 value for a bivalent molecular analyte known to the authors was 54% (16), which is lower than the %ΔT2 values of >1000% observed here. Large changes in T2 can be achieved by selecting particles with diameters of >100 nm that have high relaxivity per particle, as outlined previously (36). Here, we increased the change in T2 further by increasing the particle concentration to facilitate fast binding kinetics, lower baseline T2 values, low variation in the baseline T2, and large cluster formation that leads to large changes in T2 signals. This yielded a T2MR detection system with high signal-to-noise ratios (~200) at low levels (1 CFU/ml) of target pathogen.

The assay was evaluated using reference strains and clinical isolates spiked into healthy donor whole blood. Multiday assay repeatability measured using C. albicans–spiked blood showed excellent precision. The LoD ranged from 1 to 3 CFU/ml for all Candida species tested, which was more than 10-fold greater than previously published studies (8). These performance metrics were unchanged for spikes into unhealthy patient samples, thus demonstrating the robust nature of this T2MR assay.

Performance on blinded clinical specimens examined the capability of T2MR measurements to detect Candida infections directly in infected blood specimens not only on the day a sample was initially drawn but also within hours of the sample draw. This highlights the potential of this assay for monitoring Candida infection before and after administration of antifungal therapy. Additionally, these initial data demonstrate good positive and negative agreement of the assay with blood culture for samples that have been deemed Candida-positive or Candida-negative by blood culture.

The observation of negative blood cultures after antifungal intervention (Table 5) may arise from an absence of viable (live) Candida cells present within the patients’ blood or from antifungal-induced inhibition of Candida cell growth on the blood culture instrument, as reported previously (37). Blood culture requires viable pathogen cells and the absence of any substances that could inhibit cell growth. This T2MR method requires intact Candida cells to preserve the genomic DNA through the initial assay steps of concentration and lysis. The T2MR method did not require cell growth and was not susceptible to interference from antifungals. The fact that T2MR was able to detect Candida within the patients’ blood after antifungal treatment indicates the presence of intact but nonviable Candida cells within the blood that do not grow when cultivated in blood culture.

The design choice of detecting intact Candida cells and not circulating DNA was motivated by reported poor clinical performance of a PCR-based Candida assay that detects all available Candida DNA in blood specimens (38), and the fact that the clinical relevance of circulating pathogen DNA as a marker of infection remains a topic of active investigation (39). The T2MR data suggest that after initiation of antifungal therapy, persistent cells can remain in clinical specimens. Although preliminary, these data suggest that the test may be useful in monitoring treatment and Candida clearance. Follow-up studies will be needed to confirm the presence of intact Candida cells within clinical samples after the initiation of antifungal therapy and fully correlate that with blood culture performance before and after initiation of antifungal therapy.

Before translation is possible, assay performance will need to be compared with blood culture at lower titer levels of organism in spiked samples and with measurements across much larger sets of clinical samples. To realize and fully understand the assay’s impact in clinical settings, these data should be extended by means of a complete sample-to-answer automation to facilitate testing at clinical sites.

In conclusion, we have developed a sensitive and specific MR-based test for the diagnosis of candidemia caused by the five most commonly implicated Candida species. Early clinical results are encouraging and suggest that rapid diagnosis and species identification is achievable. Because the sequence of the oligonucleotide capture probes dictates the target specificity, this detection method can be readily adapted for detection of other nucleic acid–based targets. We anticipate that this nanoparticle-based T2MR method can be fully automated and broadly applied to infectious disease diagnoses in a variety of specimen types.

Materials and Methods

MR relaxometer

A small, portable relaxometer, or T2MR instrument (T2 Biosystems), was used for T2 relaxation measurements within clinical samples with the Carr-Purcell-Meiboom-Gill sequence. T2 relaxation data were fit with a nonlinear least-squares fitting algorithm to obtain the monoexponential decay constant, which corresponded to the measured T2 value. No further data processing or normalization was applied to measured T2 values. The T2MR instrument contains a 0.5-T samarium-cobalt permanent magnet assembly and a radio-frequency probe for excitation and detection with a 5-mm-diameter solenoid coil and a proton frequency of operation of 22 to 24 MHz. The system temperature was held at 37 ± 0.1°C.

Nanoparticle sensor conjugation and characterization

Carboxylated iron oxide superparamagnetic particles, consisting of numerous iron oxide nanocrystals embedded in a polymer matrix comprising a total particle diameter of ~800 nm (T2 Biosystems), were conjugated to aminated DNA oligonucleotides with standard carbodiimide chemistry. DNA-derivatized nanoparticles were stored at 4°C in 1× tris-EDTA (TE) (pH 8) and 0.1% Tween 20. Iron concentration of nanoparticle conjugates was measured by dissolving the particle with 6 M HCl followed by addition of hydroxylamine hydrochloride and 1,10-o-phenanthroline and subsequent spectrophotometric detection. Oligonucleotide-derivatized particles were then subjected to a functional performance test by conducting hybridization-induced agglomeration reactions with diluted synthetic oligonucleotide targets identical in sequence to the fungal ITS2 sequences from the five different Candida species within the sodium phosphate hybridization buffer 4×SSPE (600 mM NaCl, 40 mM sodium phosphate, 4 mM EDTA). Reversibility of the agglomeration reaction was confirmed by subjecting agglomerated reactions to a 95°C heat denaturation step, conducting a T2 measurement, and repeat hybridization at 60°C followed by a second T2 measurement.

PCR primer and nanoparticle capture probe design

Universal pan-Candida PCR primers were designed complementary to 5.8S and 28S ribosomal RNA sequences that amplify the ITS2 region of the Candida genome. For each of the five Candida species, a pair of species-specific oligonucleotide capture probes was designed complementary to nested sequences at the 5′ and 3′ end, respectively, of the asymmetrically amplified PCR product. The capture probe hybridizing to the 5′ end of the amplicon was 3′-aminated, whereas the capture probe hybridizing to the 3′ end of the amplicon was 5′-aminated. A poly(T) linker (n = 9 to 24) was added between the amino group and the first nucleotide base of the capture probe sequence. PCR primers and capture probes were procured from IDT Technologies (table S1). PCR inhibition control design, Candida cultivation, spiked sample preparation, and blood culture vial inoculation are described in the Supplementary Methods.

Whole-blood PCR

The entire assay workflow is shown in Fig. 1A. Detergent lysis of erythrocytes was conducted with a proprietary mixture of Triton X-100 and Igepal CA-630 (Sigma-Aldrich) on ~2 ml of each whole-blood sample, followed by low-speed centrifugation (6000g). The supernatant was removed and discarded. TE buffer (pH 8.0) (150 μl) was then used to wash the blood debris and cells before another brief centrifugation (6000g) and aspiration of the supernatant. After addition of 100 μl of TE buffer (pH 8.0) containing 700 copies of the inhibition control, the suspension was subjected to bead beating mechanical lysis. Lysate (50 μl) was then added to 50 μl of an asymmetric PCR master mix containing 200 μM deoxynucleotides, a 4:1 ratio of forward and reverse PCR primers, and a proprietary formulation of a whole blood–compatible thermophilic DNA polymerase (T2 Biosystems). Thermocycling was conducted with the following cycle parameters: 95°C for 10 min; 40 cycles consisting of 20 s at 95°C, 30 s at 62°C, and 30 s at 68°C; and a final extension at 68°C for 10 min. The PCR products for the five Candida species were ~250 nucleotides in length and were visible via gel electrophoresis and SYBR Green staining only in reactions spiked with relatively high concentrations of Candida cells (>33 CFU).

Hybridization-induced agglomeration assays

A 15-μl sample of the resulting amplification reaction was aliquoted into 0.2-ml thin-walled PCR tubes (Axygen) and incubated in a sodium phosphate hybridization buffer (4×SSPE) with pairs of oligonucleotide-derivatized nanoparticles at a final iron concentration of about 0.2 mM iron per reaction, yielding a baseline T2 value of about 30 to 40 ms. Hybridization reactions were incubated for 3 min at 95°C followed by 30 min at 60°C within a shaking incubator set at an agitation speed of 1000 rpm (VorTemp, Labnet International). Hybridized samples were then subjected to an 8-s vortexing step (2000 rpm) and subsequently placed in a 37°C heating block to equilibrate the temperature to that of the T2MR instrument for 1 min. The samples were then inserted into the T2MR instrument for T2 measurement.

Candida patient sample collection protocol

Blood specimen discards that had been drawn in K2EDTA Vacutainers (BD) on the same day as specimens drawn for blood culture were obtained from the clinical hematology laboratory at the Massachusetts General Hospital (MGH). This study was reviewed by the Institutional Review Board of MGH. For the blinded study, blood sample retains were selected for T2MR if the patient’s blood culture outcome was blood culture–positive for Candida (n = 8 samples from n = 3 patients), blood culture–positive for bacteremia (n = 8 patients), or blood culture–negative (n = 8 patients). A single PCR was conducted with 1 ml of each specimen. The average of two detection reactions conducted for each PCR is shown. Seven hundred fifty copies of a competitive internal inhibition control were added to each PCR. For serially drawn specimens, the specimens were collected and cataloged from patients (n = 3) that had blood culture–positive results for C. albicans. Samples for both studies were stored within the original Vacutainer at −80°C and shipped overnight on dry ice to T2 Biosystems. Clinical sample collection protocols were reviewed and approved by the appropriate Human Research Committees. The patients all received blood draws sent for blood culture followed by microbiological identification of the positive vials.

After initial microbial identification, surveillance blood cultures were conducted to monitor treatment efficacy. Discarded specimens from these patients were stored at 4°C and selected for testing by the T2MR assay if the outcome of the first blood draw was blood culture–positive for Candida. Discarded specimens were selected that were drawn within 24 hours of the blood culture draws. For each specimen, a single PCR was conducted with ~2 ml of input blood. A competitive internal inhibition control was added to each PCR to monitor for interfering substances that could yield a false-negative result.

Statistical analysis

To sufficiently power the LoD study at 80% with an α of 0.05, the minimal sample size required to detect 35% decreases in T2 is 15. Therefore, we conducted a LoD study with 16 replicate runs per Candida species. The assay cutoff was set at 5 SDs above the limit of blank (LoB) as measured from Candida-negative specimens (table S4). The LoD was defined as the CFU level at which ≥95% positive detection was observed. The statistical analysis of assay LoD, LoB, and interference testing was conducted with methods outlined in the Clinical Laboratory and Standards Institute guidances EP12-A2, EP7-A2, and EP17-A. The statistical significance level was 5%, and all analyses were conducted with MiniTab XVI.

Supplementary Materials


Fig. S1. Titration of single-stranded target DNA with probe-functionalized particles.

Fig. S2. Standard curve used to calculate nanoparticle oligonucleotide graft density.

Fig. S3. T2 values for spiked blood samples with and without interfering substances.

Fig. S4. T2MR detection of C. albicans and C. parapsilosis DNA targets with two different C. parapsilosis particles.

Table S1. Pan-Candida primer and species-specific capture probe sequences.

Table S2. Matrix evaluation across different healthy blood donors.

Table S3. Matrix evaluation across different unhealthy blood donors.

Table S4. Unhealthy Candida-negative patient samples used for assay cutoff calculation.

Table S5. Interfering substances that were tested.

References and Notes

  1. Acknowledgments: We thank the T2 Biosystems team for their efforts and A. Desalermos and M. Muhammed for blinded clinical sample collection. Funding: This work was funded by T2 Biosystems Inc. Author contributions: L.A.N., M.A., and T.J.L. designed the assay. L.A.N., M.A., N.A.P., L.R.S., and T.J.L. developed the T2MR particle detection reactions. L.A.N., M.A., N.A.P., M.M., A.S., D.P., M.B., P.W., and T.J.L. developed the lysis and concentration protocols. M.B. and V.D. developed the T2MR instrument. L.A.N., N.A.P., M.A., M.M., and D.P. collected analytical performance data. T.A., J.J.C., and E.M. reviewed clinical records. J.J.C. collected clinical samples from the laboratory. L.A.N., M.A., L.R.S., J.J.C., E.M., and T.J.L. wrote and revised the manuscript. Competing interests: L.A.N., M.A., N.A.P., M.M., A.S., D.P., M.B., V.D., L.R.S., P.W., and T.J.L. are employees of T2 Biosystems Inc., which manufactures the T2MR instrument used in this work. The T2MR instrument and assay technology described herein are covered under the following U.S. patents and applications: US7,564,245; US8,102,176; US12/844,672; US2011/0020787; US12/844,677; US2011/0020788; US12/844,680; US2011/0021374; US12/910,594; US2012/0100546; US13/384,051; US13/363,916; US13/402,566; US13/646,402; US13/649,839; PCT/US11/056933; PCT/US11/056936; and US13/650,734. The other authors declare that they have no competing interests. Data and materials availability: Materials, reagents, and instrumentation manufactured by T2 Biosystems are available to readers under materials transfer agreement or purchase.
View Abstract

Navigate This Article