Tapping into the native Pseudomonas bacterial biofilm structure by high-resolution multidimensional solid-state NMR

Abstract

We present a multidimensional magic-angle spinning (MAS) solid-state NMR (ssNMR) study to characterize native Pseudomonas fluorescens colony biofilms at natural abundance without isotope-labelling. By using a high-resolution INEPT-based 2D ¹H–¹³C ssNMR spectrum and thorough peak deconvolution at the 1D ssNMR spectra, approximately 80/134 (in 1D/2D) distinct biofilm chemical sites were identified. We compared CP and INEPT ¹³C ssNMR spectra to differentiate signals originating from the mobile and rigid fractions of the biofilm, and qualitatively determined dynamical changes by comparing CP buildup behaviors. Protein and polysaccharide signals were differentiated and identified by utilizing FapC protein signals as a template, a biofilm forming functional amyloid from Pseudomonas. We identified several biofilm polysaccharide species such as glucose, mannan, galactose, heptose, rhamnan, fucose and N-acylated mannuronic acid by using ¹H and ¹³C chemical shifts obtained from the 2D spectrum. To our knowledge, this study marks the first high-resolution multidimensional ssNMR characterization of a native bacterial biofilm. Our experimental pipeline can be readily applied to other in vitro biofilm model systems and natural biofilms and holds the promise of making a substantial impact on biofilm research, fostering new ideas and breakthroughs to aid in the development of strategic approaches to combat infections caused by biofilm-forming bacteria.

1. Introduction

Many species of bacteria form densely structured communities known as biofilms that confer protection to the resident cells against diverse environmental stressors. Biofilm protected bacteria cause ~ 80 % of all chronic infections, which are difficult to treat and augment antimicrobial-resistance (AMR) [1–3]. AMR results in ~ one million casualties per year and is estimated to cause more death than cancer by 2050 [4]. The structural integrity of biofilms is maintained by a complex array of extracellular polymeric compounds [5], including proteins in the form of amyloid fibrils and polysaccharides, but very little structural information exists about biofilms and their components. For the protein part of biofilms, currently, there is no high-resolution structure of a biofilm forming functional amyloid other than TasA from Bacillus determined from in vitro or purified preparations [6,7]. Similarly, only limited structural information exists on the polysaccharide fraction of biofilms, which is a complex environment for Pseudomonas species composed of several polysaccharides with different charges (for example, for P. aureginosa the polysaccharides are mainly Psl (neutral), Pel (cationic) and alginate (anionic)) [8,9]. This lack of structural understanding of biofilms, hampers development of therapeutic strategies. Novel approaches directed by structural insights are necessary for obtaining detailed information on biofilms and structure–activity relationship, that will be instrumental in advancing the fight against chronic infections and AMR.

Structural biology has celebrated many achievements in recent years, but studying proteins under native physiological conditions, such as in biofilms, remains a significant challenge. High-resolution techniques mostly rely on unnaturally high concentrations and quantities of proteins. For characterization of biological systems at their native environment, magic-angle spinning (MAS) solid-state NMR (ssNMR) and cryo-EM/tomography are very powerful tools, and could provide atomic or near-atomic resolution information, respectively [10–14]. Compared to other techniques, ssNMR is extremely powerful if/when the sensitivity problem is addressed. In particular, proton-detected (¹H-detected) ssNMR and DNP-enhanced ssNMR (DNP ssNMR here on) are potentially game-changing in studying protein structure in native biofilms due to the remarkable increase in ssNMR sensitivity that reduces sample requirements by several orders of magnitude compared to conventional ssNMR [15–17]. Methodological developments in sample preparation/optimization and technological breakthroughs in DNP-enhanced ssNMR have been demonstrated on complex systems in native environments, whole cells, and extracts [18–20]. We provided the first example of measuring the deuterated functional amyloid TasA in its native Bacillus bacterial biofilm and compared to in vitro preparations by using proton-detected high-resolution ssNMR [21].

Over the past decade, ssNMR spectroscopy has been widely used to characterize bacterial, plant, and fungal cell walls [22–24]. However, particularly in biofilm research, ssNMR studies thus far relied primarily on 1D ssNMR spectroscopy, which often poses challenges such as signal overlap and limited quantification due to lower-resolution and signal overlap. Nevertheless, structural and compositional information could be obtained and correlated to the nature of bacterial and fungal biofilms [22,23,25]. To overcome the limitations due to 1D ssNMR spectroscopy and differentiate signals from different biofilm parts, e.g. from proteins and polysaccharides, nD ssNMR experiments is crucial and proven powerful in plant and fungal cell wall studies [26,27]. To the best of our knowledge, there are only a few multidimensional ssNMR examples applied; by conventional ssNMR on cell-walls of S. aureus, high-resolution fast-MAS on E. coli-P. aureginosa cell-wall, and high-sensitivity DNP-enhanced ssNMR on B. subtilis cell-wall [28–30]. The ¹³C–¹³C DARR ssNMR was used to observe NMR spectrum on a fully isotope labeled S. aureus cell walls, and tentative assignments was used to understand structural details [29]. The DNP enhanced ssNMR experiments have been demonstrated utilizing 2D spectra on native systems like fungi and plants [24]. This high-sensitivity DNP approach has shown great utility in characterizing, e.g., intact or extracted bacterial cell walls as a sole demonstration till now [28]. Similarly, there are only few examples of utilizing high-resolution proton and carbon-detected ssNMR for bacterial cell-wall characterization [30,31]. In summary, a small number of multidimensional ssNMR results highlights that there is a knowledge gap in the ssNMR-based bacterial/biofilm research.

Consequently, there is a need to develop and employ robust advanced multidimensional ssNMR techniques for the structure and dynamics characterization of native biofilms [32]. Here, we aim to address this lack of structural information on bacterial biofilms by using Pseudomonas fluorescens Pf0-1 strain as a model organism. In this work, we utilize conventional ambient-temperature 1D and 2D MAS ssNMR to successfully detect, characterize, quantify, and better understand the structural details of a native Pseudomonas biofilm. To our knowledge, this study comprises the first high-resolution 2D ssNMR characterization of an intact Pseudomonas colony biofilm without chemical extraction and fractionation of individual biofilm components, and the quantification of the rigid and mobile fractions of the system.

2. Methods

2.1. Preparation of biofilm samples for NMR

Details of the Pseudomonas fluorescens colony biofilm preparation was described previously [33–35]. In brief, liquid cultures of Pseudomonas fluorescens Pf0-1 strain were spotted on solid Pseudomonas growth agar (PAF) and incubated at 30 °C for three days. Colony biofilms were gently scraped off the agar surface and stored at 4 °C in microfuge tubes for ssNMR analysis. For the ¹³C detected ssNMR experiments, we utilized thin-walled 3.2 mm rotors. Total amount of biofilm material in the rotor was ~ 40–60 mg. We used a benchtop centrifuge to transfer the biofilm material into the rotor, by using simple pipets attached directly onto the rotor and then centrifuged. A large fraction of the native hydrated biofilm material consists of water, and the effective solid material in the rotor is small as determined by a simple wet-biofilm drying test (data not shown). As a result, we analyzed both the wet and dried biofilm samples in fully packed rotors. Drying method increased the amount of total material in the NMR rotor up to ~ 10-fold, due to the removal of excess hydration. Drying of the wet biofilm was performed gently at ~ 50 °C in an oven under atmospheric conditions.

2.2. NMR spectroscopy

All the MAS ssNMR experiments were performed at a 750 MHz Bruker Avance III NMR spectrometer equipped with a low-temperature triple-resonance 3.2 mm probe. We employed 10 kHz MAS spinning for all the experiments, and the set temperature was 275 K for all measurements, which corresponds to a sample temperature of around ambient temperature. 3.3 μs and 5 μs pulses were used for ¹H and ¹³C. For the CP experiments, 1 ms contact time was used with a 70–100 % ramp on the proton channel. ~ 90 kHz proton dipolar decoupling was applied for all the spectra recorded. The ¹H chemical shifts were referenced directly to 0 ppm by using DSS as an internal standard added to the NMR samples, and the ¹³C chemical shifts were indirectly referenced by using the ¹H frequency [36].

For the wet biofilm samples, the 1D ¹³C INEPT and CP spectra were recorded with 82 k and 43 k scans to allow for good SNT for analysis by using 1 and 2 s of recycle delays, respectively. The direct-polarization (DP) spectra were recorded by 2048 and 1664 scans, respectively, with 5 s recycle delays. For the dry biofilm samples, the 1D ¹³C INEPT and CP spectra were recorded with 55 k and 8 k scans by using 1 s of recycle delay. These 1D ¹³C ssNMR experiment times correspond to 2.9, 23, 23 h for the DP, INEPT and CP experiments on wet biofilm, whereas, to 2.3, 15.4, 2.3 h for the DP, INEPT and CP experiments on dry biofilm. The S/N ratios were determined by using Topspin 3.6. The signal region was set to 90 to 10 ppm, and the noise region to −70 to −150 ppm for all spectra. This signal region overestimates the S/N, due to the presence of the sharp signals from the mobile species, particularly observed for the SNT numbers for the DP spectra. For the DP experiments the SNT values were also calculated by using the same noise region but also the signal regions 10–90 ppm, which results in an order of magnitude lower SNT values. The determined S/N values were used to calculate the SNT by dividing S/N by the square root of the time in terms of minutes. The corresponding SNT values are given in Fig. 2. The spectral fitting and peak deconvolution were performed by using the ssNake program package [37]. The 1D INEPT and CP spectra were processed with gaussian broadening window function (with 35 Hz at GB = 0.02 in Topspin).

Fig. 2.

(A,B) Representation of the wet native and dry compact bacterial biofilm. (C,D) 1D ¹³C ssNMR spectra recorded with these methods on wet and dry biofilm preparations. Direct polarization (DP) as well as CP and INEPT polarization transfer schemes were used to record the ¹³C ssNMR spectra. The signal to noise ratio per unit time (SNT) values in terms of minutes are given along with the total experiment times (determined by (S/N)/(minute)^0.5 for a direct sensitivity comparison). For the DP experiments the two SNT values correspond to the calculation with the same noise region but two different signal regions, 10–90 ppm versus 110–190 ppm. These experiment times correspond to 2.9, 23, 23 h for the DP, INEPT and CP experiments on wet biofilm, whereas, to 2.3, 15.4, 2.3 h for the DP, INEPT and CP experiments on dry biofilm. The CP buildup for different signals in the 1D ¹³C CPMAS spectra (given with the color-coded curves as a functional of contact-time) are shown for (E) wet and (F) dry biofilm samples. Each CP buildup curve was normalized to the maximum intensity within that same dataset and different curve intensities are not directly comparable. The five different chemical shift ranges for the integration of signals are given on the side in terms of ppm scale.

The 2D ¹H–¹³C INEPT ssNMR spectrum was recorded with 3 k and 1.5 k transients for wet and dry biofilm samples, respectively. 1 s recycle delay was used and a total of 78 indirect dimension data points were recorded with an increment of 100 μs. The total experiment time was 67 and 33 h for the wet and dry biofilm samples, respectively. The 2D ¹H–¹³C CP ssNMR spectrum was recorded with 1 k transients for dry biofilm sample in ~ 22 h. 1 s recycle delay was used and a total of 78 indirect dimension data points were recorded with an increment of 100 μs. The solid-state NMR spectra were recorded at 10 kHz MAS and 750 MHz Bruker Avance III NMR spectrometer utilizing a 3.2 mm probe at 275 K. The solution-state 2D ¹H–¹³C HSQC NMR spectrum of the monomeric soluble FapC was recorded at a 600 MHz Bruker Avance III spectrometer equipped with a 5 mm triple resonance TCI cryoprobe [38]. 5 mm NMR sample tube was used with a total volume of 550 μl sample and a protein concentration of ~ 100 μM at 274 K. The 2D spectrum was processed with gaussian broadening (with 35 Hz at GB = 0.025 in Topspin) for direct dimension and by a mixed sine squared for indirect dimension (SSB = 3 in Topspin).

3. Results

3.1. A general solid-state NMR approach to study native bacterial biofilms

Our conventional ssNMR characterization demonstrates the feasibility of detecting sufficient ¹³C NMR signals at natural abundance from the native biofilm samples. Fig. 1A depicts biofilm formation and the extracellular components that are crucial for maintaining its structural integrity, such as polysaccharides, fibrillar functional amyloid proteins, lipids and extracellular DNA (eDNA). Fig. 1B summarizes the two different approaches to record spectra for structural analysis, by conventional room-temperature or low-temperature hyperpolarized DNP-enhanced ssNMR [39–41]. In this work, we utilized the high-resolution conventional ssNMR approach performed at ambient temperature of ~ 300 K to characterize Pseudomonas biofilm. This allows the detection of biofilms close to their native in vivo conditions. By recording CP or INEPT based ssNMR experiments the rigid and dynamics parts of the biofilm can be quantified and analyzed, respectively, Fig. 1C. Moreover, the DP based ssNMR experiments allow the quantification of the signals and help to estimate the relative CP or INEPT fractions. The DNP-enhanced ssNMR spectra on the other hand, result in much larger sensitivity spectra due to hyperpolarization, however, due to the experimental conditions of ~ 100 K the mobile fraction of the signals is frozen and only a cumulative rigid spectrum is obtained. Both methods have advantages and disadvantages and could be utilized simultaneously or separately according to the information content needed.

Fig. 1.

Schematic of the workflow for MAS solid-state NMR study of bacterial biofilms. (A) A cartoon representation of the biofilm formation, its complex environment, and its components such as bacterial cells, polysaccharides, fibrillar functional amyloid proteins, lipids and extracellular DNA (eDNA). (B) Two complementary ssNMR approaches that can be utilized to characterize bacterial biofilms, Pseudomonas biofilm in this study. The conventional high-resolution ssNMR is usually performed at ~ 300 K and the DNP-enhanced high-sensitivity ssNMR is usually performed at ~ 100 K. (C) 1D or 2D ssNMR methods based on CP or INEPT polarization transfer schemes could be applied to record spectra by ¹H or ¹³C detection depending on the sensitivity, resolution, and isotope-labelling possibilities. DP can additionally be utilized to quantify total signal amount. The cartoon representation of the NMR spectrometer was obtained from www.wikipedia.com under nuclear magnetic resonance.

The ssNMR samples from Pseudomonas biofilm were prepared in two different ways, Fig. 2A,B. First, the native wet biofilms were packed into the ssNMR rotors without any further treatment to the close to the in vivo condition. In this hydrated biofilm material only a fraction of the total mass is solid due to the excess hydration. Second, we gently dried the biofilm at 50 °C to remove the excess hydration until a brittle solid is obtained. This drying step improves the sample packing efficiency up to ~ 10-fold, and we effectively packed more solid material into the rotor to obtain better signal-to-noise per unit time (SNT). The 1D ¹³C ssNMR spectra recorded with these methods are shown in Fig. 2C,D at room temperature by CP or INEPT polarization-transfer schemes, as well by DP spectra to compare the CP and INEPT signals to the total signal. Below, 1D and 2D ssNMR results are discussed in detail qualitatively and quantitatively and correlated to biofilm structure and dynamics.

3.2. 1D ¹³C MAS ssNMR to characterize natural-abundance native Pseudomonas biofilm

Conventional room-temperature ¹³C ssNMR spectroscopy on ~ 50 mg (in a fully packed 3.2 mm ssNMR rotor) of hydrated native Pseudomonas biofilm results in high-sensitivity 1D NMR spectra within ~ 1 day (1365 min) of total data acquisition, Fig. 1C. We recorded the spectra shown in Fig. 2 with ample number of transitions with high SNT to allow quantification more precisely, whereas spectra with sufficient SNT could be recorded in a few hours under these conditions. These spectra were recorded with CP and INEPT which selectively report on the rigid versus mobile/flexible biofilm components, respectively. To our knowledge, this is the first attempt to quantify rigid versus mobile signals/parts of bacterial biofilm, whereas this comparison previously revealed valuable information in a bacterial cell wall study [31]. Moreover, the DP spectra were recorded to obtain the total signal from the whole sample and contains signals from both the mobile and rigid fractions. We compared the DP spectra to the CP and INEPT spectra. The existence of the sharp signals resembling the INEPT spectra and broader signals resembling the CP spectra, represents that DP spectra is composed of a combination of the CP and the INEPT spectra. The sharp signals observed in the INEPT spectra are similarly detected in the DP spectra. The relative intensity of the ¹³C signals in the DP spectrum of wet sample follows a similar trend as the CP spectrum, where the CO intensity is similar to the rigid part of the aliphatic signal intensity. However, for the dried sample the relative intensity of the polysaccharide and aromatic signals in the ¹³C DP spectra is much less compared to the larger fraction of the aliphatic signals. This indicates that the reduced intensity signals from poly-saccharides and aromatics, may have increased ¹³C T₁ relaxation times due to drying, and as a result are observed with less efficiency/intensity.

These 1D ¹³C spectra recorded at room temperature have significantly different sensitivity and resolution. The SNT:16 for the mobile/flexible fraction of the NMR signal recorded by INEPT, whereas the SNT:2.3 for the rigid fraction recorded by CP. Remarkably, much more signal was observed for the mobile spectrum ~ 7-fold larger SNT compared to the rigid-fraction. This indicates that under the current experimental conditions the native wet Pseudomonas biofilm largely consists of flexible species due to the hydrated nature. Alternatively, the rigid signals could be broadened and observed with less efficiency as also observed by a ~ 3 broadening in FWHM in the CP and INEPT based signals. Despite this broadening the CP-based spectra are still observed with decent SNT. The DP spectra have a SNT of 8–9 for wet and dry biofilm samples, which falls between the CP and INEPT SNT values. This SNT values drop to ~ 1 for both wet and dry samples when they were calculated by the 110–190 ppm signal area that does not contain the intense signals.

The resolution of the ¹³C INEPT ssNMR spectrum recorded on the wet native biofilm sample contains narrow signals, Fig. 2C top and Fig. 3B. The average full-width at half maximum (FWHM) observed for the signals is ~ 185 Hz (~1 ppm), see Table 1 for the list of the deconvoluted NMR signals. On the other hand, the ¹³C CP ssNMR spectrum resolution of the same wet biofilm sample is much less with broader and overlapped signals with an average FWHM of ~ 490 Hz (~2.6 ppm), Fig. 2C bottom, Fig. 3A and Table 1. When the biofilm sample was dried for maximizing ssNMR rotor packing, the ¹³C INEPT ssNMR spectra changes dramatically and only a few signals remain. Predominantly lipid signals are observed as intense and narrow resonances, whereas the carbohydrate and protein signals are removed due to their reduced flexibility. Nevertheless, the ¹³C CP spectra are very similar for the wet and dry biofilm samples with the same overall spectral pattern. An increase at the polysaccharide signal intensity at ~ 75 ppm is observed for the dry biofilm sample. This indicates increased rigidity of the polysaccharide fraction in the biofilm upon drying, as a result more efficient CP polarization transfer and larger signal. The CP buildup curves for both biofilm samples are shown in Fig. 2E,F, which further supports changing CP dynamics upon drying as the curves maximizes at shorter contact times except the carbonyl resonance. This indicates an overall increased rigidity in the sample including protein and carbohydrate fractions. Previously changing of T_1rho relaxation times, have been correlated to dynamics of carbohydrates in fungal systems [42], similar to the current buildup curve changes in ¹³C CP ssNMR experiments. These buildup curves are similar to the behavior obtained on previous polysaccharide and whole cell bacteria work, where the maximum intensity signal was obtained at ~ 500–1000 μs contact time [43,44].

Fig. 3.

(A) Quantification of different chemical species in the bacterial biofilm 1D ¹³C MAS ssNMR spectra recorded with CP and INEPT polarization transfer by peak deconvolution. The peaks are determined and fitted by ssNake program. The list of these peaks along with their linewidths and ratios are given in Table 1. Different tentative group assignments are color coded in the spectra and labeled in A.

Table 1.

List of signals determined by deconvolution of the ¹³C MAS ssNMR recorded with (A) CP or (B) INEPT, between 13.0 and 181.8 ppm in the spectra shown in Fig. 3A,B. Chemical shifts in terms of ppm, percentage integral ratio normalized to the maximum integrals (bold numbers) within the CP or INEPT peaks, and the linewidths are given in the table. A gap in the list indicates that particular chemical shift was not observed in the CP versus INEPT, or vice versa. The chemical shifts of the peaks close to each other are assumed to be from a similar chemical site and listed in the same line in the table. The average FWHM for CP and INEPT spectra are ~ 490 (~2.6 ppm) and ~ 185 Hz (~1 ppm), respectively with the processing parameters given in the methods section. The spectra were recorded at 750 MHz ¹H Larmor frequency.

13C CP			13C INEPT			13C CP			13C INEPT
ppm	%	FWHM [Hz]	ppm	%	FWHM [Hz]	ppm	%	FWHM [Hz]	ppm	%	FWHM [Hz]
13.0	0.181	489	15.2	0.085	200	84.8	0.089	697	84.6	0.003	175
16.6	0.445	480							85.5	0.011	175
18.7	0.136	341	18.3	0.949	150	87.6	0.010	300	87.9	0.004	250
20.3	0.403	381	20.1	0.021	300				89.8	0.002	100
22.0	0.731	660	22.0	0.070	300	92.1	0.086	482	90.7	0.002	100
24.2	0.425	370	23.8	0.131	250				93.2	0.003	125
26.2	0.663	500	25.6	0.082	275	95.3	0.006	150	95.4	0.002	100
			26.5	0.021	100				97.5	0.006	150
28.4	0.426	411	28.0	0.070	200				98.4	0.007	150
31.2	0.885	396	30.8	0.264	350	100.4	0.003	150	101.1	0.017	175
33.6	0.638	579	32.9	0.067	225	101.8	0.017	200	102.1	0.006	225
			35.2	0.067	178	103.6	0.014	250	104.3	0.033	200
37.1	0.537	891	37.6	0.039	400				106.1	0.008	200
40.7	0.146	303	40.6	0.124	250	116.8	0.027	332	117.3	0.005	175
42.0	0.510	884	43.0	0.046	175	119.6	0.151	617	119.9	0.004	200
46.1	0.189	954	46.7	0.008	200	122.1	0.076	878
			47.6	0.012	200	123.8	0.013	369
			48.6	0.011	200	125.3	0.014	309	125.5	0.004	150
50.4	0.379	800	51.1	0.015	200	128.0	0.058	500
			52.1	0.030	200	128.0	0.058	500
53.8	1.000	800	54.9	0.047	200	129.4	0.033	300
56.6	0.425	427	56.2	0.056	200	131.0	0.085	300	130.7	0.042	150
58.8	0.239	400	58.9	1.000	150	133.1	0.021	400	132.0	0.009	150
61.2	0.720	1050	60.9	0.011	100	134.4	0.006	400
			62.1	0.116	200	136.2	0.017	300
			64.0	0.329	150	138.6	0.043	400
65.9	0.405	600	66.3	0.029	200	141.3	0.007	200	141.4	0.006	150
68.1	0.055	300	67.7	0.010	175				143.1	0.009	150
70.3	0.308	600	70.0	0.036	125				145.3	0.005	250
			70.8	0.034	125	153.0	0.019	400
72.1	0.116	615	72.0	0.051	200	154.7	0.018	400
73.7	0.158	500	73.6	0.168	175	157.4	0.043	500
			74.5	0.005	200	157.9	0.011	500
76.1	0.145	500	75.8	0.033	250	166.1	0.010	250
			77.1	0.061	225	170.9	0.187	650
79.1	0.087	485	78.4	0.003	150	173.3	0.720	600
79.9	0.025	305	79.7	0.004	125	176.6	0.499	800
81.4	0.066	340	82.8	0.013	150	181.8	0.084	600

Fig. 3 A,B depicts the deconvolution analysis and tentative group assignments of the carbon chemical shifts observed for the wet native biofilm sample by CP and INEPT ¹³C ssNMR spectra. We utilized ssNake fitting program package [37], and was able to fit the spectra to a very good agreement. This allowed us to semi-quantitatively analyze the observed chemical species already in the 1D ¹³C ssNMR spectra and to compare the CP and INEPT spectra. Similar deconvolution of NMR signals approach was shown previously as a valid approach by Schaefer, Cegelski, and coworkers [45,46]. We demonstrate here extensive deconvolution and fitting of all peaks in these 1D spectra, as a showcase of extracting detailed information from 1D spectra of complex biological environments. Since the carbon signals from polysaccharides, proteins and other biofilm components have different chemical structure, they are observed at different chemical shift ranges despite several overlapped regions. With this protocol ~ 80 unique chemical sites were identified and listed in Table 1. The resonance sets identified in the CP and INEPT spectra resemble each other, with still unique differences. The previous studies and growing database on polysaccharide/carbohydrate chemical shifts, help us greatly to tentatively assign these peaks, as well as our recent solution-state NMR assignment of the Pseudomonas biofilm forming functional amyloid FapC [38,45–47]. The 2D ¹H–¹³C INEPT ssNMR spectrum is the ultimate way to quantitatively separate and analyze these signals, as shown in Fig. 4A,B, and will be discussed below.

Fig. 4.

2D ¹H–¹³C ssNMR correlation spectra for the wet (in red) and dried (in blue) native and natural-abundance Pseudomonas biofilm samples recorded with INEPT and CP polarization transfer schemes. (A) Comparison of the 2D ¹H–¹³C ssNMR spectra of the dry biofilm recorded with CP (in blue), and of the wet biofilm recorded with INEPT (in red) (B) The 2D ¹H–¹³C INEPT ssNMR spectra of wet (in red) and dry (in blue) biofilms. Different signal areas are highlighted in the spectra. (C) The protein aliphatic / polysaccharide region is shown as a zoom out (in red). The solution-state HSQC NMR spectrum of the biofilm forming functional amyloid FapC as a soluble monomer is overlayed (in black) to indicate the protein signals in the unstructured protein conformations. The protein signals of the biofilm spectra were tentatively assigned by utilizing the FapC spectra and the BMRB database.

The observed chemical shifts are consistent with the polysaccharide, carbohydrate, bacteria, and biofilm studies reported previously [45–47]. Overall, the carbonyl signals are at ~ 165–190 ppm, aromatic signals (and/or nucleic acids) from proteins are at ~ 110–165 ppm, polysaccharides (and/or nucleic acids) are at ~ 65–110 ppm, protein C_α/β are at ~ 50–70 and glycines C_α are at ~ 45 ppm, and finally the aliphatic signals from proteins, carbohydrates and lipids are at ~ 10–50 ppm. The existence of the carbonyl NMR signal at ~ 175 ppm indicates the presence of proteins, as this peak disappears at the INEPT spectrum due to the lack of protons at carbonyl carbons. This is additionally supported by the NMR signals observed for alpha and aliphatic protein carbons at ~ 10–70 ppm. Moreover, the C_α carbon signals in the 2D ¹H–¹³C ssNMR spectrum correlate with the correct alpha proton shifts at ~ 4.5 ppm, Fig. 4. This is consistent with the previous 1D NMR studies [48]. Due to the presence of intense resonance from lipid signals at 18.3, 30.8 ppm and other signals at 58.9 and 64.0 ppm the percentage integral of the other peaks are apparently much smaller. For the ¹³C CP spectrum the spread of peak integrals is less broad since the spectrum doesn’t have as much of intensity difference as the INEPT spectrum.

3.3. 2D ¹H–¹³C MAS ssNMR spectroscopy for high-resolution native biofilm characterization

The high-resolution 2D INEPT ¹H–¹³C MAS ssNMR spectrum shown in Fig. 4A,B is recorded in ~ 67 and ~ 33 h for wet and dry biofilms, respectively. Moreover, the 2D CP ¹H–¹³C MAS ssNMR spectrum shown in Fig. 4A is recorded in ~ 22 h for the dry biofilm sample. The CP-based spectrum comprises signals from the rigid biofilm fractions, whereas INEPT-based spectrum comprises signals from the flexible and mobile wet/dry biofilm species at natural-abundance without isotope-labelling. The spectral sensitivity and resolution of the INEPT-based 2D are excellent with averaged ¹³C resonance linewidths of ~ 185 Hz (~1 ppm). To our knowledge this spectrum is the first demonstration of multidimensional ssNMR spectroscopy at high-resolution for bacterial biofilms. The CP based 2D spectrum has much less resolution, with an average ¹³C linewidth of ~ 490 Hz (~2.6 ppm), which hampers the sensitivity. Different spectral regions comprising signals from protein aliphatic and polysaccharide, polysaccharide and aromatic regions are highlighted in Fig. 4B. The zoomed out aliphatic region is shown at the bottom for clarity, Fig. 4C. In Fig. 4B, the 2D spectra of the wet and dried biofilm samples are shown in red and blue colors, respectively. The dried biofilm sample yielded only a few sharp resonances as also shown in the 1D comparison in Fig. 2, nevertheless overlapping perfectly with the wet biofilm spectrum. To tentatively assign the protein resonances and differentiate them from the polysaccharide signals for example, we utilized solution-state 2D ¹H–¹³C HSQC spectrum of the soluble monomeric FapC protein, in black color in Fig. 4C. FapC is a biofilm forming functional amyloid from Pseudomonas and is an intrinsically disordered protein (IDP) in solution [38]. The signals from FapC matches the signals of the biofilm sample to a high degree, and we used FapC assignment as the basis of the tentative signal assignment in the biofilm spectrum, along with the BMRB database. In this manner, we accomplished the differentiation of the overlapped regions, where the protein and polysaccharide signals coexist at ~ 10–70 ppm. The signals observed from the flexible fraction of the biofilm sample (red spectrum) appear to be originating from the less ordered proteins, peptides, or amino acids due to the striking resemblance to the IDP FapC spectrum. There are additional signals in the biofilm spectrum, which we at this point suffice by referring to those as signals from the non-protein fraction of the biofilm. Extracting structural information from the 2D spectrum of the rigid biofilm fraction via CP remains as a challenge due to low resolution, requiring further optimizations of experimental conditions such as MAS frequency.

Taking advantage of the high-resolution 2D ¹H–¹³C INEPT ssNMR spectra shown in Fig. 4 and utilizing the complex carbohydrate magnetic resonance database (CCMRD), we attempted to determine the types of polysaccharide signals, Fig. 5. The Pseudomonas biofilm polysaccharides include diverse and complex polymeric substances, including alginate which is widespread across Pseudomonas species. The well-studied P. aeruginosa strain polysaccharides are here utilized a guide for signal identification [8]. In addition, the soluble extracellular polysaccharide composition of our model organism, P. fluorescens Pf0-1 strain, was previously quantified by chemical composition analysis [5,49]. All in all, by utilizing the ¹H and ¹³C chemical shifts observed in the 2D INEPT spectrum, we identified polysaccharide signals from glucose, mannan, galactose, heptose, rhamnan, fucose, and N-acylated mannuronic acid, [8,25,50] which agrees with the previous analysis [49].

Fig. 5.

Representation of the tentative polysaccharide ¹H/¹³C resonance assignments in the INEPT-based 2D ¹H–¹³C ssNMR spectrum for the wet Pseudomonas biofilm sample. We utilized the chemical shift values and the CCMRD database for identification of species. We only judge the presence of a certain polysaccharide species with four to six cross peak pairs identified from a specific candidate. These assignments comprise species such as glucose (CCMRD entry #318), mannan (poly-mannose, #33), galactose (#319), heptose (#327), rhamnan (poly-rhamnose, #329), fucose (CCMRD entry #333), and N-acylated mannuronic acid (ManpNAc #335). To unambiguously assign the resonances and identify the remaining signals, additional experiments are required and will be presented in the future.

4. Discussion and conclusions

We present in this work a high-resolution one- and two-dimensional ssNMR study of the Pseudomonas fluorescens biofilm, which is amenable to studying biofilms of any given microbial species including pathogenic P. aeruginosa. 1D/2D MAS ¹³C-detected ssNMR spectra was utilized to obtain signals from different chemical species in the biofilm and to characterize its molecular composition. This adds valuable structural and dynamics information to the growing and important research area in structural biology of biofilms, which have been predominantly depended on 1D ssNMR and limit the extend of information that could be extracted from these complex systems.

We propose multidimensional ssNMR spectroscopy to study biofilms by selectively obtaining signal from the rigid or mobile fractions of native biofilms. With the presented tentative ssNMR signal assignments, we were able to characterize the bacterial biofilm composition. 1D ssNMR spectra were recorded within several hours to a day with varying SNT and provide insights into the chemical species that exist in the biofilm. The high SNT INEPT-based 2D ¹H–¹³C ssNMR spectrum was recorded for the wet biofilm sample within 67 h and provides resonance identification. We anticipate that a 2D INEPT ssNMR spectrum of a lower, but sufficient SNT could be obtained in less than a day, which will provide a fast and high resolution detection and differentiation of different carbon pools. This will open new possibilities for understanding the interaction of proteins and other components (polysaccharides, lipids, and cell-wall) in Pseudomonas and other biofilms without the need of isotope labelling or isolating biofilm parts [28,48]. We confidently identified several polysaccharide species by analyzing the cross-peaks observed in the 2D spectrum and finding matches in the CCMRD polysaccharide database.

The challenge for future studies to overcome is the sensitivity limitation to characterize the rigid fraction of the biofilm, and similarly, to characterize the dried biofilm with sufficient resolution by multidimensional ssNMR spectroscopy. For the wet biofilm preparation, as shown in Fig. 2, the sensitivity for the rigid fraction of the biofilm is ~ 7 less compared to the mobile fraction. This indicates that a 2D spectrum similar to the one in Fig. 4A,B in terms of sensitivity would need ~ 3300 h (67 * 72 h), which is beyond possibility. In addition, since the resolution of the rigid fraction is much less compared to the mobile fraction, Fig. 4A, the sensitivity is further hampered due to broader line widths at the current moderate MAS frequency conditions. Nevertheless, by using the dry biofilm approach, we increased the amount of biofilm material in the ssNMR rotor and managed to record a CP-based 2D ¹H–¹³C spectrum, Fig. 4A. Despite lower resolution, the spectrum overlays well with the INEPT-based 2D spectrum. More optimizations of sample preparation and ssNMR spectroscopy are required to utilize the dry biofilm preparations.

DNP enhanced ssNMR would be very beneficial for increasing sensitivity of these experiments to record 2D or even 3D spectra of biofilms [39,40]. Such DNP ssNMR approach on fungal systems have been shown successful [16]. Secondly, ultra-fast MAS at > 100 kHz would also increase the sensitivity and resolution remarkably and could make it possible to characterize these systems within reasonable time and sufficient resolution. Such a fast MAS approach was shown for bacterial cell wall system [31]. Finally, a third approach could be isotope-labeling the biofilm systems to increase sensitivity by a factor of ~ 100 for ¹³C detection. This method has been used with selectively or fully isotope labeled systems [44,45], however, limits the application of ssNMR to the characterization of lab-grown systems and is not suitable for studying patient derived systems.

In summary, we presented a high-resolution multidimensional ssNMR study to characterize native Pseudomonas fluorescens colony biofilm, which could be readily applied to other in vitro biofilm model systems and natural biofilms. Wet native and dried compact biofilm sample preparation methods yielded valuable structural insights and up to ~ 80/134 distinct chemical sites were identified in the 1D by peak deconvolution and in the high-resolution 2D spectra. Based on the CP buildup behavior of the wet and dry biofilm samples, dynamical changes were qualitatively determined. A unique comparison of CP and INEPT ¹³C ssNMR spectra allowed the differentiation of signals from the mobile versus rigid fractions of the biofilm. These two fractions were compared to the spectrum obtained by DP. The high-resolution INEPT-based 2D ¹H–¹³C ssNMR spectrum along with the deconvoluted resonances from the 1D spectra resolved many cross peaks. This allowed the identification of protein signals from the polysaccharides signals by utilizing unstructured FapC protein as a template. This study represents the first demonstration of high-resolution 2D ssNMR to characterize a native biofilm sample at natural-abundance and without any sample treatment and extends the technical capacity of ssNMR in biofilm research. We demonstrate the potential of multidimensional ssNMR in structural studies for identifying chemical sites with reduced signal overlap. We provide the basis for multidimensional ssNMR-based structure characterization of biofilms and their components. Our experimental pipeline promises to carry broad impact on related research directions to stimulate new ideas and breakthroughs.

Data availability

Data will be made available on request.