Boron NMR as a Method to Screen Natural Product Libraries for B-Containing Compounds

Abstract

Natural products are biologically relevant metabolites exploited for biomedicine and biotechnology. The frequent reisolation of known natural products questions whether existing discovery models are still capable of identifying novel compounds. As innovative NMR-based screening techniques can help overcome these challenges, we applied a phase cycling composite pulse sequence to ¹¹B NMR experiments to enhance their sensitivity to screen libraries for novel boron-containing molecules. Aplasmomycin and autoinducer-2 were detected in crude and enhanced microbial fractions, via their boron signals, as proof of concept. Subsequently, a screen of 21 crude plant and 50 crude marine microbial extracts were chosen at random and analyzed with the optimized ¹¹B experiment for feasibility as a high throughput discovery method. Eight of the plant samples and 13 of the microbial samples were identified as boron-containing, suggesting that there is a higher presence of boron metabolites available from natural sources than previously known due to a lack of appropriate discovery methods. As a result, we believe that this optimized ¹¹B NMR experiment can serve as a robust method for quick and facile discovery of novel boron-containing metabolites from a variety of natural sources.

Natural products are secondary metabolites produced by bacteria, fungi, plants, and other organisms for evolutionary advantage.¹ Today, they are important compounds in biotechnology and biomedicine, being prevalent active ingredients in pharmaceuticals, agriculture, and pesticides.^2,3 Given the importance of natural products for these disparate uses, the field is driven by the development of screening approaches that take advantage of biological activity, identification of biosynthetic genes, and analytical chemistry to identify compounds of interest.

While the value of biology-driven screening and discovery efforts is immense, the rediscovery and dereplication of known compounds is a tremendous challenge.⁴ Chemical strategies that rely upon large data sets, such as GNPS, offer a comprehensive, data-centric strategy to try and identify unique chemistry.⁵ Other strategies are more focused on the reactivity of specific functional groups, such as the design of unique probes to carry out chemoselective derivatizations with readily identifiable UV or MS tags for isolation efforts.⁶ The use of probes is an extremely valuable tool; however, not all functional groups are sufficiently reactive for probes. We have been interested in the use of NMR as a direct way to identify functional group specific approaches for underrepresented chemical moieties. In that context, we have found that the rapid identification of boronic natural products in microbial and other collections may be of value.

There are a variety of boron-containing natural products, exhibiting antibiotic, anti-HIV, and quorum sensing properties (Figure 1).^7,8 Examples include: borolithochrome G (1), a boron-containing pigment from a Jurassic-putative red alga; riboflavin BC (2), a more water-soluble formulation of riboflavin; Autoinducer-2 (3), a critical quorum sensing molecule in Gram-negative bacteria; and hundreds of boronic lipids, such as 4, that have been isolated from bacteria, algae, and fungi. While the physiological utility of these molecules is unknown, it is possible that, like riboflavin, the boron improves physiochemical properties. On the basis of these metabolites, as well as those discussed later, we desired a library screening approach to identify boron-containing compounds.

Figure 1.

Examples of boron-containing metabolites.

NMR has served as an important tool for screening, whether in fragment-based approaches for drug discovery or natural products discovery efforts. Exciting new approaches, such as SMART, allow for powerful dereplication of complex natural product mixtures.⁹ Our lab has taken advantage of NMR-active nuclei, such as ¹⁵N¹⁰ and ¹⁹F (unpublished work), to study the mechanism of formation and discovery of biologically interesting compounds. On the basis of this success, we looked to ¹¹B NMR as an approach to find novel boronic compounds in our microbial natural products library. Herein we describe the design and implementation and discuss the challenges and solution to screening with ¹¹B NMR.

Initial studies to detect boron with 3.93 mM pentaphenylfluoroboronic acid (5, Figure 2) yielded large protrusions of asymmetrical noise with the standard Bruker ¹¹B NMR experiment. This noise was due to broadening of resonance peaks from ¹¹B having a nuclear spin of 3/2¹¹ coupled with inhomogeneous frequencies rising from internal probe components within the instrument itself. These extraneous signals were stronger than our compound’s and resulted in the asymmetrical boron NMR noise referred to as the “boron hump” (Figure 2), which typically covers a range of −30 to 30 ppm.¹² Attempts to rectify this by screening in quartz instead of borosilicate NMR tubes did not silence the extraneous noise. We determined that this experiment would not be sensitive enough to detect boron in natural product samples, as we would be expecting low milligram amounts of material incapable of yielding a stronger signal than those from the probe components themselves.

Figure 2.

3.93 mM 5 in CD₃OD, ns = 128, in borosilicate and quartz NMR tubes. The ¹¹B proton decoupled experiment with the standard zgig pulse sequence in both borosilicate (a) and in quartz (b) tubes pick up extraneous probe noise from outside the NMR coil, causing large, asymmetrical peaks, which can overshadow compound peaks. Use of the zgbs pulse sequence with the proton decoupled ¹¹B NMR in quartz tubes (c) is best for suppressing extraneous probe noise.

We hypothesized that optimizing the sensitivity of ¹¹B NMR by silencing the extraneous probe signals and reducing the boron hump would provide the means to detect lower levels of boron material. On the basis of work by Cory and Ritchey, DEPTH pulse sequences can be applied to NMR to select homogeneous regions of radio frequency. They were able to successfully apply a three-pulse DEPTH experiment to Si NMR to suppress glass resonance from siloxysilanes.¹³ We believed this method could likewise suppress extraneous boron resonances from the NMR probe itself. This composite pulse series is the standard 90° pulse with two subsequent 180° pulses with applied phase cycling,^13,14 essentially applying a 90° pulse for the spins inside the coil and 0° for the outside, canceling all extraneous noise from outside the coil.^14,15 With this, excess signals are suppressed, and the resultant NMR signals arise only from the boron sample within the coil. As this pulse sequence was successfully applied to Si NMR, yielding more ample signals,¹³ we sought to achieve the same with boron.

This pulse sequence is known as “zgbs” and was readily available on Bruker’s pulse program library. We applied it to the standard ¹¹B proton decoupled experiment, the nucleus was changed for boron detection, and neither the phase cycling nor delay parameters were modified. We named it “¹¹Bzgbsig”: “zgbs” to indicate the composite pulse sequence and “ig” for “inverse gated,” as this was applied to the proton decoupled experiment. To further reduce noise, we continued conducting all experiments in quartz NMR to ensure signal would not be compromised by borosilicate in regular tubes. Proof of concept was retested with a solution of 3.93 mM 5 in CD₃OD (Figure 2), and total reduction of the boron hump was seen with the applied zgbs pulse sequence.

The ability of this pulse sequence to suppress asymmetrical peak noise in ¹¹B NMR seemed very promising for use as a screening method for boron-containing natural products. As proof of concept, we sought to detect metabolites from crude extracts and enriched fractions using only their boron signals. 3 and Aplasmomycin (6) (Figure 3) were additionally used to illustrate the effectiveness and feasibility of this optimized NMR experiment.

Figure 3.

Structures of additional boronic compounds used to illustrate the effectiveness of the ¹¹B NMR experiment with the applied pulse sequence.

6 is a microbially derived macrolide with a Böeseken complex¹⁶ core that exhibits antibiotic activity. Its structure was originally characterized and published without the boron until it was elucidated with X-ray crystallography, not ¹¹B NMR.^17,18 A bacterial strain in our library was known to produce 6, but over the course of isolation, we found there to be extensive challenges in the analytical chemistry, including validating the presence of boron. With the standard Bruker ¹¹B decoupled pulse sequence on crude, semicrude fraction, or pure 6, only the asymmetrical noise was seen (Figure 4). This dominating, broad signal made it difficult to see and confirm the desired boron signal at 10.5 ppm. This confirmed our suspicion that the standard ¹¹B NMR experiment, without the zgbs pulse sequence, would be useless as a screening method for natural product discovery. Running the same sample with the zgbs pulse sequence (the ¹¹Bzgbsig experiment) in quartz NMR tubes gave the boron peak against a flattened baseline, confirming the experiment’s ability to detect low quantities of boron in crude and pure fractions. Enrichment of 6’s boron NMR is seen with subsequent steps of purification (Figure 5).

Figure 4.

3.22 mM of 6 in CDCl₃ via the standard Bruker ¹¹B decoupling NMR experiment in quartz NMR tubes (blue), ns = 128, yields the boron hump, overshadowing distinctive chemical signals. Application of the zgbs pulse sequence (red), ns = 128, eliminates background resonance noise allowing for clear identification of boron peaks.

Figure 5.

¹¹B NMR of 3.22 mM 6 in CDCl₃ via the standard ¹¹B proton decoupled experiment, ns = 128, throughout various stages of purification. Significant reduction of extraneous signals is observed with the pulse sequence resulting in clearly distinguishable boron peaks.

The sensitivity of this experiment was then probed with its ability to detect Autoinducer-2 (AI-2) in situ. AI-2 (3) is a naturally occurring, boron-containing furan. It is produced in both Gram-positive and -negative bacteria serving as an extracellular signaling molecule for inter- and intraspecies communication.¹⁹ It is highly studied in the reporter bacterial strain Vibrio harveyi, which uses 3 for quorum sensing-mediated bioluminescence.^7,20 In V. harveyi, the concentration of 3 is proportional to the luminescent expression of the bacteria. It is typically detected via the engineered strain,²¹ but the bioassay used for detection is not quantitative and can be thwarted at high concentrations of culture medium. AI-2 is also relatively unstable at lower concentrations and is sensitive to growth conditions. For example, low pH or the addition of glucose to media inhibits luminescence, and the presence of borate can shift chemical equilibrium between stereoisomers,²² thus making it very challenging to standardize quantification of 3 in culture.

Various methodologies have been developed throughout the years to facilitate detection and quantification of 3 including Fe³⁺ supplementation to growth medium;²² LuxP-based fusion proteins;²³ 1,2-phenylenediamine reactions forming HPLC measurable quinoxaline derivatives;²⁴ environmentally sensitive protein receptors with fluorescent dyes;²⁵ GC-MS;²⁶ and chemical probes such as d-desthiobiotin-AI-2.²⁷ But these methods have their shortcomings including sensitivity to environmental inference, invasiveness, and dependence on concentration. NMR being quick, sensitive, and noninvasive would be an efficient and facile way to detect 3. More so, exploiting the boron with ¹¹B NMR would eliminate the need of purification from other naturally occurring metabolites in culture, making it an efficient detection method. ¹¹B NMR has been reported once for 3, but with low resolution after 80 000 scans.¹⁹ If ¹¹Bzgbsig could both hasten acquisition and improve data resolution, it would be a great experiment to employ for detecting 3.

After a 24 h incubation of a 50 mL culture of V. harveyi, 3 was detectable using the ¹¹Bzgbsig NMR experiment with aliquots from aqueous media. This is the first detection of 3 in culture by ¹¹B NMR of which we are aware. A sharp singlet at −1.32 ppm, in DMSO-d, is indicative of the one boron atom in 3 (Supporting Information S4.7). The same was seen when the V. harveyi aliquot was resuspended in CD₃OD (Supplementary S4.9), but a clear pentet with splitting J_B–D = 3.50 Hz is observed when taken in D₂O (Figure 6). The splitting is indicative of deuterium exchange, further confirming that we are observing 3 via ¹¹B NMR. Observing 3 with ¹¹B NMR exemplifies the experiment’s ability to detect quantitative amounts of extracellular metabolites. Additionally, as NMR can quantify material, this method could be an appealing solution for quantification of 3.

Figure 6.

¹¹Bzgbsig proton decoupled NMR of V. harveyi in D₂O, ns = 512. Closer inspection of the peak reveals pentet splitting (insert) from the exchange between D₂O and boron on 3, further proving detection of the molecule.

Boron is a fascinating element with remarkable chemical and biological properties due to its unique electronic structure. It plays an ever-increasing role in drug development, making up the active agents in antibacterial, anticancer, and antiviral therapies, among others.^28,29 The ability to find structurally new boronic acids, boronic esters, and boronates from natural sources has the potential to discover equally novel biologies elicited through mechanisms not yet known.

While boronic natural products could lead to incredible biological activities, high throughput screens (HTS) to identify such metabolites have not yet been executed due to boron’s relatively unknown presence outside of plant and algal natural products. Only a miniscule fraction of the known boronic natural products are from microbial sources such as 6, boromycin,³⁰ or the tartrolons.^18,31

Additionally, there is a lack of efficient screening methods for boron. Potentially boronic compounds are present in natural product libraries but have been lost during HTS or purification. Identification of boronic metabolites cannot rely solely on bioassay-guided fractionation, as boron is easily lost from boranes and boric acids with traditional HTS and isolation methods. For example, boron is easily removed from chelated, five-membered rings with exposure to common acidic buffers in bioassays or chromatography,³² which could be a reason for its minimal presence in natural product discovery. If boron is crucial for a compound’s activity, the “hit” or “target” can be lost due to incompatible isolation schemes. The noninvasive nature of NMR is an appealing alternative, as it will retain the structural integrity of molecules of interest and better inform isolation strategies. In theory, crude extracts from natural extracts can be screened with this optimized ¹¹B NMR experiment, and the presence of boron peaks would be indicative of boronic natural products.

Since plants are known to be producers of boronic metabolites, various plant extracts were screened with this optimized ¹¹Bzgbsig experiment for proof of concept. Twenty-one terrestrial plant samples were collected at random from the UCSC greenhouse. Samples included stem, leaf, and fruit from selected plants, which were frozen, crushed, and extracted with MeOH. The organic residues were then dissolved in CD₃OD, and ¹¹B NMR was performed with ns = 256 scans. Spectra that revealed a boron peak at the end of acquisition were labeled as “boron containing,” and those without boron peaks were labeled as “non-boron containing.” The screen yielded 8 boron containing plant extracts.

After success with the plant extracts, we tested a subset of relatively unannotated crude extracts from our marine microbial extract library. 50 crude microbial samples were chosen at random. 150 mg aliquots of methanol-soluble material were centrifuged, transferred to quartz NMR tubes, and analyzed with the ¹¹Bzgbsig experiment, ns = 512 scans. Of the 50 tested, 13 fractions were seen to contain boron natural products, which were not seen with the standard Bruker ¹¹B experiment (utilizing the zgig pulse sequence). The 13 samples retained their boron peaks after EtOAc solvent partition to prove that it was a complex boronic compound and not boric acid carried over from fermentation media. Thus, 26% of our library subset revealed the potential for novel boron natural products, which was previously unknown. Figure 7 summarizes the percent distribution of boron from the microbial and plant extracts analyzed with this optimized experiment. The results from this screen suggest there are many boron-containing natural products that have been left undiscovered due to insufficient screening methods.

Figure 7.

Percent distribution of boron found in the plant and microbial samples screened using the optimized ¹¹B NMR experiment.

Additionally, this sensitive experiment will ease elucidation and characterization of boron-containing structures. As a result of boron’s diamagnetic properties, the ¹¹B signals can span a range of about 250 ppm depending on the molecule’s configuration.¹¹ Like ¹H and ¹³C NMR, the shift of an ¹¹B NMR peak gives insight into the surrounding chemical environment and the overall molecular structure. Additionally, this pulse program was successfully applied to the standard ¹¹B no proton decoupling experiment (zg Bruker standard experiment). In the case that naturally occurring boranes are identified, for example, the no proton decoupling experiment can show splitting from a B–H bond with the same sensitivity as the proton decoupling experiment (Supporting Information S3.3 and S3.4).

As this optimized method can be used to detect quantitative amounts of boronic species from crude and semicrude natural product fractions, it is promising as a nondestructive HTS method for discovery of new boronic metabolites. Current work with this experiment consists of systematically screening our microbial natural product library to identify, isolate, and characterize novel compounds. We are optimistic at the prospects of discovering novel boron-containing structures, and expanding the scope of natural product chemical space, using this optimized experiment.

In conclusion, the zgbs composite phase-cycling pulse sequence effectively silences extraneous background noise in ¹¹B experiments by suppressing regions of inhomogeneous frequencies within the NMR instrument itself. With this newfound sensitivity, quantitative amounts of boron are now detectable from crude plant and microbial extracts by NMR. Identification of boronic compounds from these sources was an excellent proof of concept for the utility of this experiment as a screening method and suggests it can be expanded beyond the subset of species we examined.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.orglett.2c00885.

• Experimental procedures and ¹¹B NMR spectroscopic data (PDF)

• FAIR data, including the primary NMR FID files, for compounds 3 (ZIP)

• FAIR data, including the primary NMR FID files, for compounds 5 (ZIP)

• FAIR data, including the primary NMR FID files, for compounds 6 (ZIP)

The authors declare no competing financial interest.