Benchtop NMR Spectroscopy: A Step Toward Widely Accessible Lipoprotein Analysis

Researchers have successfully adapted high-field nuclear magnetic resonance (NMR) technology for comprehensive lipoprotein analysis to more accessible benchtop systems. This breakthrough paves the way for wider use of this powerful diagnostic tool, potentially revolutionizing cardiovascular disease risk assessment and personalized healthcare.

Cardiovascular disease (CVD) remains a major global health challenge, and analyzing lipoproteins in blood is crucial for assessing an individual's risk. Traditionally, detailed lipoprotein analysis using Nuclear Magnetic Resonance (NMR) spectroscopy has required expensive, large, high-field instruments. These systems demand significant space, high capital costs, and require cryogenic liquids, limiting their use primarily to central laboratories. 

This research explores translating this powerful analytical method to smaller, more affordable benchtop NMR systems. By successfully adapting high-field NMR-based lipoprotein analysis for routine benchtop use, this work aims to make comprehensive lipoprotein quantification more widely available, potentially transforming how cardiovascular disease risk is assessed at both the population and individual levels.

Key Takeaways

• High-field NMR-based lipoprotein analysis was successfully translated to lower-field benchtop systems. Benchtop systems are smaller, cheaper, and don't require cryogenics, making the technology accessible to small laboratories and clinics.
  
  
• 25 out of 28 major lipoprotein parameters were accurately measured within 15 minutes using benchtop NMR, including key cardiometabolic risk markers like LDL-C, HDL-C, total cholesterol, Apo-A1, and Apo-B100.
  
  
• Despite challenges like reduced spectral dispersion and sensitivity at lower field strengths, a quantitative calibration approach yielded stable and reproducible results across multiple sites at the measured temperature.
  
  
• Potentially increases accessibility allowing for molecular phenotyping in the clinic, enhancing the potential for longitudinal personalized medicine.

Overview

Lipoproteins are vital supramolecular structures composed of lipids and proteins that are responsible for managing and transporting lipids throughout the body. Abnormal levels of lipoproteins, such as high low-density lipoprotein cholesterol (LDL-C) and low high-density lipoprotein cholesterol (HDL-C), have long been linked to increased risk of cardiovascular disease. 

Other markers like apolipoproteins Apo-B100 and Apo-A1, and their ratio, are also recognized as valuable CVD biomarkers. Recent research also indicates a link between inflammation and modifications in lipoproteins, suggesting that chronic inflammatory conditions might contribute to cardiometabolic diseases like diabetes and chronic kidney disease. 

For example, increased LDL concentrations and a shift towards smaller, denser LDL particles are seen in obesity, which is associated with chronic inflammation. Lipoproteins have also been identified as accurate predictive markers for the severity and recovery from SARS-CoV-2 infection, highlighting their potential for detecting chronic inflammatory conditions.

Traditionally, the gold standard methods for lipoprotein quantification and characterization are ultracentrifugation and gradient gel electrophoresis. However, NMR spectroscopy has emerged as a powerful tool capable of quantifying over one hundred lipoprotein parameters in a single, relatively quick measurement (around 4 minutes for high-field systems, with the benchtop method achieving results within 15 minutes). 

This analysis includes details about the lipid composition of main and subfractions, particle size and number, and protein content. Historically, this detailed NMR analysis has relied on high-field systems (400 to 600 MHz), which are expensive, bulky, and require cryogenic liquids for operation.

Benchtop NMR systems, utilizing permanent magnets at lower field strengths (60 to 100 MHz), offer a more accessible alternative. They have a smaller footprint, operate at room temperature, and are simpler and more cost-effective to maintain.

While promising clinical applications using biofluids like urine, serum, and plasma have been reported with benchtop NMR for detecting various biomarkers, applying it to the complex analysis of lipoproteins has been challenging due to reduced spectral dispersion and sensitivity at lower field strengths.

This collaborative study aimed to address this challenge by establishing a lipoprotein serum model for benchtop NMR systems. Researchers from three institutes analyzed a total of 389 samples from three cohorts using both high-field (600 MHz) and benchtop (80 MHz) NMR. The cohorts included healthy/free-living populations and a diabetic/obese cohort, chosen to create a wide range of lipoprotein profiles and maximize the model's covariance range. 

High-field data, obtained using the Bruker IVDr Lipoprotein Subclass Analysis (B.I.-LISA) method, were used as the reference to build the benchtop regression model against the 80 MHz spectra. The standardization of acquisition protocols and external quantitative calibration ensured consistency across the different sites, leading to a robust joint model capable of recovering a significant majority of key lipoprotein parameters.

Why it’s Important

The successful translation of comprehensive lipoprotein analysis to benchtop NMR systems has profound implications for both clinical diagnostics and medical research. Currently, advanced lipoprotein profiling is often limited to specialized laboratories due to the cost and complexity of high-field NMR instruments. The availability of accurate and reliable benchtop systems can democratize access to this powerful diagnostic tool, allowing smaller clinics, regional hospitals, and research facilities to perform these analyses in-house. This could lead to:

• Improved Cardiovascular Risk Assessment: More detailed lipoprotein profiles, including subclass analysis and apolipoprotein concentrations, offer a more nuanced view of an individual's cardiovascular risk compared to standard lipid panels. Making this analysis more accessible could lead to earlier and more accurate risk identification.
  
  
• Enhanced Personalized Medicine: The ability to perform frequent and cost-effective lipoprotein analysis opens doors for longitudinal monitoring of patients. This allows clinicians to track an individual's response to lifestyle interventions or pharmacological treatments in greater detail, leading to more personalized and effective therapeutic strategies.
  
  
• Broader Research Applications: Researchers studying metabolic diseases, inflammation, and other conditions where lipoproteins play a role will benefit from the increased accessibility of this technology. It could facilitate larger cohort studies and the exploration of novel biomarkers.
  
  
• Point-of-Care Potential: While this study focused on laboratory-based benchtop systems, further advancements could potentially lead to even more compact and user-friendly devices suitable for near-patient or point-of-care testing, especially when combined with advancements in self-administered capillary blood sampling.
  
  
• Early Detection of Inflammatory Conditions: The study's success in quantifying inflammatory markers like SPC and Glyc alongside lipoproteins is particularly significant. Chronic inflammation is increasingly recognized as a key driver of various diseases, including CVD. The ability to simultaneously assess lipoprotein metabolism and inflammatory status could provide earlier and more comprehensive insights into disease development and progression.

The authors themselves highlight that this development is a "major milestone toward deployment of NMR diagnostic and prognostic tools within the clinical and healthcare landscapes." The ability to perform harmonized measurements across multiple sites and platforms is crucial for clinical applications, and this study successfully demonstrated this feasibility.

Summary of Results

The core of the research involved a collaborative study across three research institutes and a company laboratory. They analyzed samples from three distinct cohorts totaling 389 samples. 

After excluding samples based on quality control and for use as long-term references, 358 samples were used for building the models. The cohorts consisted of a diabetic/obese group (N=110) from Australia/Mauritius, a free-living population (N=121) from Spain, and a healthy/free-living population (N=127) from Germany. This diverse sample set was intentionally used to create a larger variance in lipoprotein profiles, which helps in building robust models.

Each sample was measured using both a high-field 600 MHz NMR spectrometer and an 80 MHz benchtop system. The high-field data, generated using the Bruker IVDr Lipoprotein Subclass Analysis (B.I.-LISA) method, served as the reference. A benchtop NMR model was then developed by using regression analysis to correlate the 80 MHz spectra to the lipoprotein data obtained from the 600 MHz spectra.

Despite the anticipated challenges associated with lower field strength, specifically decreased spectral dispersion and sensitivity, the study achieved significant success. By standardizing the acquisition protocol and employing an external quantitative calibration approach, stable and reproducible results were obtained from the different participating sites.

The resulting joint model, built from the combined cohort data, was able to recover a substantial number of lipoprotein parameters. Specifically, the model recovered 26 out of 28 (93%) main lipoprotein parameters and 62 out of 112 (55%) total parameters. When focusing on major parameters, the study successfully predicted 25 out of 28 major parameters. These included crucial cardiometabolic risk markers like LDL-C, HDL-C, total cholesterol, Apo-A1, Apo-B100, and the Apo-B100/Apo-A1 ratio. The analysis using benchtop systems was completed within 15 minutes.

Spectral Analysis and Reproducibility:

Despite the 7.5-fold decrease in spectral dispersion and approximately 50-fold decrease in signal-to-noise ratio at 80 MHz compared to 600 MHz (Figure 1a), all major lipid peaks necessary for lipoprotein analysis were visible in the 80 MHz spectra within a 15-minute acquisition time.

Figure 1: This figure compares typical NMR spectra of serum at 80 MHz and 600 MHz, highlighting the lipid peaks. It also shows the consistency of 80 MHz spectra from the three participating institutes after quantitative referencing (QuantRef) and demonstrates the reproducibility of measurements over a 20-day period. Specifically, panel (d) shows that 96.4% of measurements over 20 days fell within the 95% confidence interval.

The researchers implemented a quantitative referencing method (PULCON) to ensure comparability of spectral intensities across different samples and platforms. The superimposed 80 MHz spectra from the three sites showed good agreement in spectral intensity for key lipoprotein regions (Figure 1b, 1c). The standard deviation for spectral region integration across long-term reference samples at 80 MHz was 2.3%, which, while higher than at 600 MHz (~1%), was considered acceptable given the lower signal-to-noise ratio.

Regression Model Performance:

A sparse regularized generalized canonical correlation analysis (SGCCA) was used to build the regression model, correlating three spectral regions of interest from the 80 MHz data (aliphatic, GLYC, and SPC regions) with the 112 lipoprotein parameters obtained from the 600 MHz B.I.-LISA method.

Figure 2: This figure summarizes the prediction performance of the SGCCA model. Panel (a) shows scatter plots for the prediction of 28 main lipoprotein parameters and fractions, with correlation coefficients (r2) color-coded based on performance. Panel (b) displays the normalized root mean square error (RMSE) and r2 values for all 112 parameters, highlighting that clinically important parameters were well-recovered.

The model successfully predicted 25 out of 28 major lipoprotein parameters with strong correlation coefficients (r2 values ranging from 0.64 to 0.99). These included:

Total triglycerides (TPTG): r2=0.99
Total cholesterol (TPCH): r2=0.86
ApoA1 (TPA1): r2=0.84
ApoB100 (TPAB): r2=0.81
LDL-C (LDCH): r2=0.78
HDL-C (HDCH): r2=0.78

Overall, 62 out of 112 (55%) lipoprotein parameters achieved a satisfactory r2>0.7. An additional 30 parameters showed moderate prediction performance (0.5<r2<0.7). Parameters that were more difficult to predict were primarily LDL subfractions, some VLDL subfractions, and specific HDL subfractions, which is attributed to spectral congestion and lower visibility at lower field strengths. The inflammatory markers SPC and Glyc were also successfully recovered with r2 values of 0.95 and 0.78, respectively.

Impact of Temperature:

The study also investigated the effect of temperature on the lipoprotein signals at 80 MHz, as NMR spectra are typically acquired at 310 K (37 °C) to maximize lipoprotein peak areas, while most benchtop experiments in this study were conducted at 298 K (25 °C).

Figure 3: This figure illustrates the impact of temperature on the intensity of lipoprotein signals at 80 MHz for a serum sample and its isolated main fractions. Measurements at 310 K (black trace) consistently showed higher signal intensity compared to 298 K (dotted trace), particularly for the LDL fraction.

Increasing the temperature from 298 K to 310 K resulted in a substantial gain in signal intensity for all lipoprotein fractions, with the LDL fraction showing an almost doubling of intensity. This suggests that operating benchtop systems at 310 K, where feasible, could further improve the prediction of parameters, especially for LDL subfractions.

Model Interpretation:

The SGCCA model allows for the interpretation of how different spectral regions contribute to the prediction of specific lipoprotein parameters.

Figure 4: This figure shows the "outer weights" and "outer components" for the data blocks (SPC, GLYC, and CH regions from 80 MHz spectra) and the lipoprotein parameters for the first two components of the model. The color-coding indicates positive (red) or negative (blue) contributions of spectral variables to the latent variables. This visualization helps to confirm the spectroscopic basis of the model's predictions. For example, the model correctly associated specific parts of the SPC peak with HDL lipoprotein parameters.

This interpretability is an advantage over "black box" machine learning algorithms, allowing researchers to verify that the model's predictions are based on meaningful spectroscopic features.

Conclusion

This research represents a significant advancement in making comprehensive lipoprotein analysis more accessible. By successfully translating high-field NMR capabilities to benchtop systems, the study overcomes the major limitations of cost, size, and operational complexity that previously restricted this powerful diagnostic tool to specialized centers.

The ability to accurately measure 25 out of 28 major lipoprotein parameters, including key risk markers, on a benchtop instrument within a rapid 15-minute timeframe is a crucial milestone. This demonstrates that despite the inherent compromises in spectral quality at lower magnetic field strengths, a quantitative calibration approach can yield stable and reproducible results across multiple locations.

The implications of this work are far-reaching. Widespread access to detailed lipoprotein profiles facilitates the adoption of molecular phenotyping in clinical settings. This shift from generalized risk scores to individual biochemical profiles allows for more precise risk stratification, earlier detection of potential issues, and the implementation of truly personalized medicine strategies, including longitudinal monitoring of patient health. Coupled with potential advancements in sample collection methods, this technology has the capacity to significantly enhance preventative healthcare and the management of cardiometabolic diseases. While high-field NMR will continue to be essential for detailed biological exploration of lipoproteins, benchtop NMR is now poised to become an indispensable tool in routine clinical diagnostics.