*Article* **Insights into a Protein-Nanoparticle System by Paramagnetic Perturbation NMR Spectroscopy**

**Yamanappa Hunashal 1,2 , Cristina Cantarutti <sup>3</sup> , Sofia Giorgetti <sup>4</sup> , Loredana Marchese <sup>4</sup> , Federico Fogolari 5,6 and Gennaro Esposito 1,6,\***


Academic Editors: Michael Assfalg, Roberto Fattorusso and Mark von Itzstein Received: 6 July 2020; Accepted: 22 October 2020; Published: 7 November 2020

**Abstract:** Background: The interaction between proteins and nanoparticles is a very relevant subject because of the potential applications in medicine and material science in general. Further interest derives from the amyloidogenic character of the considered protein, β2-microglobulin (β2m), which may be regarded as a paradigmatic system for possible therapeutic strategies. Previous evidence showed in fact that gold nanoparticles (AuNPs) are able to inhibit β2m fibril formation in vitro. Methods: NMR (Nuclear Magnetic Resonance) and ESR (Electron Spin Resonance) spectroscopy are employed to characterize the paramagnetic perturbation of the extrinsic nitroxide probe Tempol on β2m in the absence and presence of AuNPs to determine the surface accessibility properties and the occurrence of chemical or conformational exchange, based on measurements conducted under magnetization equilibrium and non-equilibrium conditions. Results: The nitroxide perturbation analysis successfully identifies the protein regions where protein-protein or protein-AuNPs interactions hinder accessibility or/and establish exchange contacts. These information give interesting clues to recognize the fibrillation interface of β2m and hypothesize a mechanism for AuNPs fibrillogenesis inhibition. Conclusions: The presented approach can be advantageously applied to the characterization of the interface in protein-protein and protein-nanoparticles interactions.

**Keywords:** protein-nanoparticle interactions; protein NMR; amyloidogenic proteins; nitroxide paramagnetic perturbation; spin label extrinsic probes; Tempol; β2-microglobulin

#### **1. Introduction**

The interaction of proteins with nanoparticle (NP) systems is a very challenging issue that has many implications in physical chemistry as well as in biomedical and biochemical applications [1]. Depending on the NP charge, size, shape, and chemical functions on the surface, proteins may be adsorbed onto that surface, concentrated in a layer named corona, or experience labile interactions and exchange with bulk solution. There are several experimental strategies that can be employed to assess the interaction between NP and proteins, ranging from direct inspection by microscopy or suitable spectroscopic techniques, e.g., UV, fluorescence, surface-enhanced Raman spectroscopy, etc., to the indirect inference based on the assay of the protein function through the related biological or cellular

response. More detailed information can be obtained also by other techniques such as NMR, although the application viability is restricted to those systems where the protein exchange can be exploited to gain information on the transiently bound states that are unobservable by NMR because of the large sizes. However, magnetic resonance can be employed over a wider scale range if electron resonance is considered, provided the systems under consideration respond to free paramagnetic species or can host suitable paramagnetic probes on the proteins or the NPs. For NMR applications, the most convenient NP size window to modulate the protein interaction is the medium-size range, i.e., 5–20 nm, where the effects of the NP shape, charge, and surface chemistry can be tuned for the scopes of interest.

Over the last few years, we have carried out systematic investigations on the interaction of amyloidogenic protein models and citrate-coated or alkanethiolate-coated gold NPs (AuNPs) with diameters of 3.6, 5, and 7.5 nm [2–7]. We worked in particular on β2-microglobulin (β2m) and variants thereof that represent a paradigmatic example of amyloidogenic protein misfolding [8,9]. β2m naturally occurs in class I major histocompatibility complex on the surface of antigen presenting cells, in conjunction with a larger domain. Due to renal failure and consequent high concentration from impaired clearance [8], or because of a mutation [9], a pathologic fibrillar aggregation of β2m takes place, leading to amyloid deposition in patients undergoing long-term hemodialysis or aged individuals with genetically inherited mutation.

NP interaction studies were conducted on wild-type β2m [10–12], the naturally occurring amyloidogenic mutant D76N β2m [9,13] and ∆N6 β2m, which is a variant devoid of the first six residues that is found only in natural fibrils of the wild-type species [14]. Contrary to the expectations based on earlier results [15], the protein solutions with AuNPs were stable for several months, and no evidence of increased aggregation or partial unfolding was observed. The occurrence of uneven patterns of signal attenuation was indicative of a preferential interface of fast exchange with AuNPs [2–4]. With D76N β2m, the most amyloidogenic variant of β2m that fibrillates by agitation at neutral pH, the presence of citrate-stabilized AuNPs inhibited fibrillogenesis by interfering with the early aggregation steps of the protein that are crucial for the protofibril nucleus formation, as inferred from NMR, QCMD (Quartz Crystal Microbalance with Dissipation monitoring), and MD (Molecular Dynamics) [3,4,6].

Recently, we have revived the use of water-soluble nitroxides such as Tempol to explore the exchange dynamics of β2m [16]. Tempol and similar stable free radicals had been formerly employed as extrinsic probes for identifying the protein exposed locations, based on the paramagnetic perturbation of the NMR signals induced by the unpaired electron of the radical over accessible molecular surfaces [17–20]. The same paramagnetic perturbation measured under non-equilibrium conditions of the NMR magnetization determines an attenuation pattern that differs from the corresponding profile obtained under equilibrium conditions, i.e., with fully relaxed NMR magnetization. While in the latter conditions, the extent of NMR attenuation reflects the proximity to the unpaired electron and therefore the accessibility of or the distance from the molecular surface, the attenuation retrieved under non-equilibrium conditions of magnetization recovery can map also the locations of hindered accessibility or µs-to-ms exchange events, by identifying slower or faster relaxing nuclei, respectively, with respect to the average relaxation rate enhancement brought about by the nitroxide probe [16].

Here, we show the application of this novel use of Tempol attenuation to gain insights into the interactions that wild-type β2m establishes with citrate-coated AuNPs. The analysis of the ternary system protein/AuNP/spin-label probe is conducted with respect to all the NMR- (Nuclear Magnetic Resonance) and ESR- (Electron Spin Resonance) accessible controls involving only two components of the system, which are based also on the previously reported evidence [2–7,16].

#### **2. Results**

Using extrinsic spin labels such as nitroxides to extract structural information requires testing the reliability of their non-specific probe behavior [17–20]. ESR spectra of Tempol in the absence and presence of β2m had previously shown that only statistical encounters occur between the free radical and the protein, as inferred from the invariance of linewidths and amplitudes of the superimposed spectra [16]. For the ternary system protein + AuNPs + Tempol, the ESR trace superposition for the three controls (Tempol, Tempol + AuNPs, and Tempol + β2m) and the ternary system shows substantial coincidence with some small amplitude deviations (Figure S1). The degree of meaningfulness of these deviations was assessed by calculating the rotational correlation time (τ*c*) of the nitroxide in the different tested conditions, according to Equations (4) and (5) (see Section 4). Table 1 lists the corresponding values. Under any tested condition, the tempol τ*<sup>c</sup>* value remains around an average of 31.6 ps (the standard deviation is 1.3 ps). This indicates that, within the experimental error, no detectable effect arises from β2m, or AuNPs, or both on the tumbling rate of the free radical. On the other hand, that average τ*<sup>c</sup>* value is consistent with those reported for 2 mM Tempo (91.9 ps) and Tempone (14.9 ps) in water at 300 K [21]. Therefore, the occurrence of Tempol interactions other than the statistical collision in the binary and ternary systems here considered should be ruled out.

**Table 1.** Rotational correlation time (τ*c*/10−<sup>11</sup> s) of Tempol at the indicated concentrations and different solution compositions obtained from ESR measurements at 298 K.


Figure 1 shows the pattern of the normalized attenuation (*AN*) values observed with 8 µM β2m and 0.8 mM Tempol with respect to the same protein solution without the nitroxide. The numerical values are listed in Table S1, along with the corresponding errors. The graph of Figure 1 depicts in red the backbone amide signal *A<sup>N</sup>* values extracted from data collected under magnetization equilibrium conditions (*AN*[eq]), and in blue the analogous *A<sup>N</sup>* values extracted from data collected under magnetization off-equilibrium conditions (*AN*[off-eq]), which were respectively obtained from pairs of <sup>15</sup>N-1H HSQC spectra acquired with relaxation delays of 5 s and 0.5 s. According to our previous interpretation [16–18], *AN*[eq] values larger or smaller than unity indicate amide signals attenuated above or below the average attenuation, respectively, and therefore, they identify molecular locations more or less accessible to the nitroxide probe, depending on the specific surface exposure. Instead, the interpretation of the *AN*[off-eq] values is related to their relationship with the corresponding *AN*[eq] figures [16]. In particular, the pattern *AN*[off-eq] > *AN*[eq] identifies amide positions with either locally hindered accessibility on the molecular surface or true structurally buried positions. As such, this pattern, which was named the Type I deviation of *AN*[off-eq], is typically, though not exclusively, associated with *AN*[eq] < 1, i.e., locations with accessibility lower than average [16]. As a matter of fact, when Type I deviation occurs at exposed locations, the corresponding *AN*[eq] value is only slightly larger than unity. On the other hand, the pattern *AN*[off-eq] < *AN*[eq], named Type II deviation of *AN*[off-eq], identifies those amide positions whose recovery is faster than the average off-equilibrium signal recovery, thereby proving less attenuated than that average. In the absence of specific interactions of the spin probe with the protein and/or structural transitions of the latter induced by the former, as verifiable by the invariance of the amide signal chemical shifts (Figure S2), Type II deviation of *AN*[off-eq] can be associated to local chemical or conformational exchange occurring on a ms-to-µs time scale that introduces additional relaxation increments affecting both *T*1*<sup>p</sup>* and *T*2*p*, i.e., the paramagnetic contribution to longitudinal and transverse relaxation times [16].

**Figure 1.** Overlay of the *A<sup>N</sup>* values obtained from <sup>1</sup>H-15N HSQC spectra of 8 µM β2m in the presence of 0.8 mM Tempol, with a relaxation delay of 0.5 s (**blue**) or 5 s (**red**). The β-strand location and naming along the sequence is reported with yellow strips. The cartoon on the right highlights the positions of the accessible amides (red), i.e., exhibiting *AN*[eq] > 1, and the amides with Type II deviation of *AN*[off-eq] (**blue**), i.e., displaying *AN*[off-eq] < *AN*[eq]. The magenta color denotes sites where both *AN*[eq] > 1 and Type II deviation occur simultaneously. Here and elsewhere, the reproduced structure is the NMR solution structure of β2m [10] (Protein Data Bank or PDB code 1JNJ). The secondary structure elements of β2m are indicated according to the crystallographic naming scheme (PDB code 3HLA). Structures are always drawn with PyMOL (Schrödinger, Inc., version 2.3.5, New York, NY, USA).

The pattern of Figure 1 is different with respect to the corresponding one previously observed in diluted conditions, precisely at β2m concentration of 50 µM probed with 250 µM Tempol [16]. Apart from the larger error affecting the previous data, there are two important differences to point out. First, the former tempol/protein ratio was 5:1, whereas here, we consider a ratio of 100:1. These ratios and the absolute concentrations affect directly the collision probability [22]. We reasoned that a high tempol/protein ratio is required to balance a low absolute concentration of β2m and measure the paramagnetic perturbation. This is confirmed by the slight increase above the unity of the average relative intensity, RIav, (for RI definition see Section 4) under off-equilibrium conditions [16] (Figure S3).

Second, at 8 µM concentration, the extent of β2m dimerization and higher oligomerization should be further reduced compared to 50 µM [14,23,24]. Hence, it could be possible to observe features related to the protein association interface. Table 2 lists the details of the pattern observed in Figure 1 compared to the earlier results at 50 µM [16]. The most relevant differences concern the higher exposure in the 8 µM solution of strands C, D, F, and G and the intensification in local conformational or chemical exchange at strands C and F, with a simultaneous loss of accessibility at strand A and loop AB. In addition to the relevance for the involvement in the association interface, these features are also important to delineate a starting point and thus appreciate the interactions and structural effects that the presence of AuNPs may induce.


**Table 2.** Paramagnetic perturbation induced by 0.8 mM Tempol on the amide NMR signals of 8 µM β2m. Equilibrium (column 2) and off-equilibrium (columns 3 and 4) data are compared to the corresponding data obtained at 50 µM β2m and 5:1 Tempol:protein ratio [16] and reported below in italic fonts.

The same paramagnetic perturbation analysis as done with isolated β2m can be performed for the system protein + AuNPs, because a statistical collision model can still be adopted, according to the ESR-based determinations of the Tempol τ*<sup>c</sup>* values under different experimental conditions. Moreover, AuNPs are known to essentially preserve the chemical shifts, and therefore the structure, of β2m and D76N β2m [2,4,6], although at concentrations and NP/protein ratios as low as 4–8 µM and 1/100–1/200, small chemical shift deviations have been detected for both variants [2,6]. Most of these deviations were observed with synthetic AuNPs with an average diameter of 7.5 ± 1 nm, and therefore, some difference can be expected upon decreasing the NP diameter to 5 nm. With the commercial AuNPs here employed, minor chemical shift perturbations [25] are measured at Q2, N17, S33, D38, and S61 NHs of β2m, which in two cases (N17, D38) decrease below the resolution significance in the presence of Tempol (Figure S2). A possible explanation for those chemical shift perturbations may be related to NP-induced alterations of the intra-residue interaction between the side-chain polar group and the backbone amides. From the results reported in Figure 2 and Table 3, it can be seen that N17 and D38 become accessible to Tempol in the presence of AuNPs. Therefore, the reduction of their chemical shift perturbation could be related to the interaction with Tempol that, albeit non-specific, competes with the intra-residue interaction. On the other hand, the conserved chemical shift deviations of Q2, S33, and S61 after Tempol addition match with a hindered accessibility in the presence of AuNPs (Figure 2, Table 3). Therefore, all of the observed chemical shift perturbations can be attributed exclusively to the protein interaction with the NPs. However, given the substantial invariance of the fingerprint pattern in the <sup>15</sup>N-1H HSQC spectra and the limited amounts of the frequency changes, all the mentioned deviations do not impair the assumptions of protein structure conservation and statistical character for the nitroxide probing. However, all the previous evidence indicates that AuNPs unevenly affect

the intensity of the <sup>15</sup>N-1H HSQC peaks of β2m and variants thereof [2–7]. Although this effect is precious to interpret the molecular details of the protein interaction with nanoparticles, it may prove detrimental when evaluating the paramagnetic attenuation contributed by the nitroxide probe, under magnetization equilibrium or off-equilibrium conditions.

**Figure 2.** Overlay of the *A<sup>N</sup>* values obtained from <sup>1</sup>H-15N HSQC spectra of 8 µM β2m + 60 nM gold nanoparticles (AuNPs) in the presence of 0.8 mM Tempol, with a relaxation delay of 0.5 s (**blue**) or 5 s (**red**). The β-strand location and naming along the sequence is reported with yellow strips. The cartoon on the right highlights the positions of the accessible amides (**red**), i.e., exhibiting *AN*[eq] > 1, and the amides with Type II deviation of *AN*[off-eq] (**blue**), i.e., displaying *AN*[off-eq] < *AN*[eq]. The magenta color denotes sites where both *AN*[eq] > 1 and Type II deviation occur simultaneously. The secondary structure elements of β2m are indicated according to the crystallographic naming scheme (PDB code 3HLA).




**Table 3.** *Cont*.

A preliminary control of the <sup>15</sup>N longitudinal relaxation rates is useful to estimate the entities of the effects brought about by AuNPs and Tempol on β2m signals (Figure S4). Indeed, the average *T*<sup>1</sup> value of 8 µM β2m decreases by the same extent, i.e., 4%, with AuNPs or Tempol, whereas the addition of the nitroxide to β2m in AuNPs suspension shortens the average *T*<sup>1</sup> only by 0.7%. This means that the Tempol paramagnetic attenuation probing the statistical sampling differences would be extensively masked by the general and specific dipolar attenuation AuNPs inflict to β2m signals, if the *A<sup>N</sup>* values were computed with respect to the isolated protein intensities. In conclusion, proper control intensities should be obtained from the protein sample in the presence of AuNPs rather than from the protein alone. Figure 2 shows the attenuation pattern of the backbone amide signals observed with 8 µM β2m and 60 nM AuNPs due to 0.8 mM Tempol. The numerical values are listed in Table S2, along with the corresponding errors. According to the preliminary *T*<sup>1</sup> control measurements, the *A<sup>N</sup>* values are calculated with respect to the control solution, i.e., the sample with the same composition without the nitroxide. Compared to the *A<sup>N</sup>* profile of the isolated protein (Figure 1), evident differences emerge in the *A<sup>N</sup>* plot of Figure 2 concerning exposed or poorly accessible regions, as well as locations undergoing chemical or conformational exchange. In particular, AuNPs hinder accessibility at the N-terminal, end of strand B, loop CC' and subsequent strand C', start of strand D, strand F, and C-terminal of β2m, while increasing strand A exposure. The details of these differences can be appreciated from Table 3, which lists the equilibrium and off-equilibrium attenuation data with (bold fonts) and without (plain fonts) AuNPs, thereby enabling a direct comparison.

#### **3. Discussion**

The results above described must be analyzed from two viewpoints. On one side, the nitroxide-based screening on the free protein helps to gather elements on the association processes β2m undergoes in solution. These elements may reveal features related to the fibrillogenic propensity of β2m. On the other side, the same paramagnetic probing of the protein in the AuNPs suspension provides the elements that enable outlining the β2m interaction with those nanoparticles. Joining these two lines of evidence is particularly tempting, because it is thus possible to focus the mechanism of fibrillogenesis inhibition experimentally observed with AuNPs and the β2m variant D76N [4].

As pointed out in the previous section, the most relevant differences highlighted by equilibrium and off-equilibrium nitroxide collisional labeling when comparing the 8 µM and the 50 µM β2m solutions concern the higher exposure in the former of strands C, D, F, and G and the intensification in local conformational or chemical exchange processes at strands C and F, with a simultaneous loss of accessibility at strand A and loop AB (Table 2). This result is consistent with the previous inference and evidence addressing respectively the intermolecular interface in ∆N6 β2m, the variant devoid of the N-terminal hexapeptide fibrils [14], and the H-D exchanged β2m fibrils dissolved in DMSO [26]. In particular, it is very meaningful that the increased accessibility at strands C and F observed at lower

concentration is paralleled by the onset, at the same positions, of exchange events (*AN*[off-eq] Type II deviations, Table 2) that must witness the remnant of the intermolecular interaction propensity at those locations. At higher concentration, the pattern changes into significantly lower exposure and exchange due to the shift of the association dynamics toward oligomeric adducts.

The statistical sampling of Tempol shows that the presence of AuNPs exposes strand A and limits accessibility at the very N-terminal segment, end of strand B, turn CC' and strand C', start of strand D, strand F, and C-terminal region of β2m. Again, the gain or loss of accessibility at strands A, B, C', and F, and turn CC' are accompanied, at the same locations, by a gain or loss of Type II deviation of *AN*[off-eq]. However, except for the N-terminal fragment (Q2, R3), the interaction pattern obtained by paramagnetic perturbation does not seem to match the previously reported picture based on relative intensity losses affecting essentially the end of strands B and D, and loops BC and DE [2]. The differences in β2m concentrations (8 vs. 26 µM) and protein/AuNPs ratios (100 vs. 200) between present and previous experiments could certainly play a role in determining some deviations. However, the relative intensity changes marking the difference between the absence and presence of AuNPs, and the paramagnetic perturbation measured in the presence of AuNPs, monitor different aspects of β2m interaction with AuNPs, and thus, the mismatch may be only apparent. In fact, similar to the previously reported low relative intensity of Q2 and R3 induced by AuNPs that translates into a Type I deviation of *AN*[off-eq], i.e., hindered accessibility to the nitroxide probe, a similar effect with the paramagnetic perturbation is seen also for S28 (strand B, see Table 3). More frequently, instead, the β2m residues reported to lower the relative intensities due to AuNP interaction [2] exhibit Type II deviation of *AN*[off-eq], i.e., local exchange, with nitroxide probing in the presence of AuNPs, which are typically coupled to above average accessibility (*AN*[eq] > 1). This is the case with the ends of strands B and D, and with the involved subsequent locations at BC and DE loops (Table 3). Therefore, the picture emerging from the interpretation of the results obtained by equilibrium and off-equilibrium paramagnetic mapping does not conflict with the previous evidence but rather provides a more detailed and enriched characterization of the interaction between β2m and AuNPs. This appears quite clearly by the inspection of Figure 3 where the surface of the protein according to the equilibrium and off-equilibrium attenuation pattern induced by Tempol in the absence (upper structures) and presence (lower structures) of AuNPs is highlighted.

Hindered accessibility is marked by orange surfaces that change their distribution on moving from the isolated protein to the presence of AuNPs. In either conditions, those surfaces could be associated to regions that become screened by the relevant interaction, namely the residual protein-protein or the protein-nanoparticle one. The occurrence of exchange processes at the blue and magenta locations can be considered the consequence of an interaction that takes place over the ms-to-µs time scale and may represent a further interface with different dynamic properties with respect to the hindered accessibility surface, provided that the occurrence of a local conformational exchange process is ruled out [16]. Finally, the highly accessible positions that are identified in red indicate the surface that is not involved in any protein-protein nor protein-nanoparticle contact. Based on the discussed evidence for β2m alone [14,26] and with AuNPs [2,3], the results listed in Tables 2 and 3 and depicted in Figure 3 suggest that the AuNP interference leading to the inhibition of fibrillogenesis [4,6] could occur via interaction of the nanoparticles with the N-terminal and strands D and F of β2m, in addition to other contacts at the end of strand B, turn CC', and strand C'. The suggested hypothesis is that these interactions in which AuNPs engage with the protein surface prevent the protein-protein contacts at the same locations that are necessary for fibrillogenic aggregation.

**Figure 3.** Cartoon representation of β2m surface sampled by Tempol in the absence (**upper pair**) and presence (**lower pair**) of AuNPs. The positions of the accessible backbone amides, i.e., exhibiting *AN*[eq] > 1, are marked in red. The locations of the amides with Type I or Type II deviation of *AN*[off-eq], i.e., displaying *AN*[off-eq] > *AN*[eq] or *AN*[off-eq] < *AN*[eq], are highlighted in orange or blue, respectively. The magenta color denotes sites where both *AN*[eq] > 1 and Type II deviation occur simultaneously. The very few positions where *AN*[eq] > 1 and Type I deviation coincide were left in orange. The secondary structure elements of β2m are indicated according to the crystallographic naming scheme (PDB code 3HLA).

#### **4. Materials and Methods**

#### *4.1. Chemicals*

Sodium Citrate, <sup>2</sup>H2O, Tempol (4-hydroxy-2,2,6,6-tetramethyl-piperidine-l-oxyl), and HEPES (*N*-(2-Hydroxyethyl)piperazine-*N*'-(2-ethanesulfonic acid) were all from Sigma Aldrich (St. Louis, MO, USA). From the same source were also the citrate-stabilized Au nanoparticles, here referred to as AuNPs. The average AuNP diameter was 5 nm, and the supplied suspension concentration was 91 nM.

#### *4.2. Sample Preparation*

The uniformly <sup>15</sup>N-labeled wild-type human β2m was expressed with an additional methionine at the N-terminus (Met-0) and purified as previously reported [14]. The protein samples in the absence of AuNPs were prepared in H2O/D2O 95/5, 1.5 mM sodium citrate, 20 mM HEPES buffer, pH 7. The protein concentration was 8.0 µM, as determined by UV absorption at 280 nm. For solutions with AuNPs, proper amounts of D2O and concentrated HEPES and β2m solutions were added to the mother NP suspension containing already citrate to reproduce the above-mentioned composition. Following dilution, the final AuNP concentration was 60 nM. A few microliters of concentrated Tempol solution were added when necessary to the NMR tube containing 0.550 mL of β2m, with or without AuNPs, to reach the desired Tempol/protein concentration ratio. For NMR samples, Tempol concentration was always 0.8 mM. For ESR samples, solutions at variable Tempol concentrations were prepared, i.e., 0.4, 0.8, and 1.6 mM, in aqueous buffer (20 mM HEPES, 1.5 mM sodium citrate, pH = 7), either alone or in the presence of 8 µM β2m, or 60 nM AuNP, or 8 µM β2m + 60 nM AuNP.

#### *4.3. Spectroscopy*

All the spectra were acquired at 298 K. The NMR experiments were collected at 14.0 T (1H at 600.19 MHz, <sup>15</sup>N at 60.82 MHz) on a Bruker Avance III NMR system equipped with triple resonance cryoprobe. Two-dimensional <sup>15</sup>N-1H HSQC experiments [27] carried out using sensitivity-improved Echo/Antiecho-TPPI pure phase detection in F1, gradient coherence selection, and flip-back pulse for solvent suppression [28–30] were acquired over spectral widths of 40 ppm and 14 ppm in F1 and F2 dimensions, respectively, with 64 time-domain points in t1, 256 or 512 scans × 2048 points in t2, and 64 dummy scans to achieve steady state. After a reproducibility check, relaxation delays were set to 0.5 and 5 s, respectively for off- and on-equilibrium conditions of magnetization recovery, following the guidelines previously reported [16]. The contour plots are reported in Figure S5, along with the corresponding signal-to-noise values (Table S3). The <sup>15</sup>N longitudinal relaxation times were measured using the sequence proposed by Kay and colleagues [31] with the modifications for sensitivity enhancement and flip-back pulse for solvent suppression [28–30]. The spectra with eight different relaxation intervals were acquired (10, 30, 60, 100, 140, 200, 400, and 1200 ms). All NMR data were processed with TOPSPIN version 4.0.2. Prior to Fourier transformation, linear prediction in t1 (up to 128 points) and zero filling were applied to yield a final data set of 2 K × 1 K points. For longitudinal relaxation analysis, the Bruker Dynamics Center 2.5.3 routine was used.

ESR spectroscopy experiments were collected with a Bruker EMXnano spectrometer operating in the X band. Capillaries filled with 50 µL of sample solution were placed in standard 4 mm tubes and submitted to acquisition (1 scan). The ESR operating parameters were as follows: frequency = 9.6 GHz; microwave power = 0.316 mW; modulation amplitude = 1 Gauss; modulation frequency = 100 kHz; center field = 3429.8 Gauss; sweep width = 200 Gauss; time constant = 1.28 ms. The data were processed using the software package Xenon (version 1.1b50, Bruker, Billerica, MA, USA).

#### *4.4. Spectroscopic Data Treatment*

Amide cross-peak intensities in <sup>15</sup>N-1H HSQC spectra of β2m in the absence (*Id*) and in the presence (*Ip*) of Tempol or/and AuNPs were measured by SPARKY software (version 3.133, T.D. Goddard and D.G. Kneller, University of California, San Francisco CA, USA). Normalized attenuation, *AN*, was calculated according to Equation (1) [18]

$$A\_N^k = \left(2 - \frac{t\_p^k}{t\_d^k}\right) \tag{1}$$

where the running index *k* refers to the kth residue amide cross-peak and the ι *k p*,*d* values are the corresponding auto-scaled intensities of the peaks in the presence (subscript *p*) and absence (subscript *d*) of nitroxide or/and AuNPs, which are defined as

$$\mu\_{p,d}^k = \frac{I\_{p,d}^k}{\frac{1}{n}\sum\_{k=1}^n I\_{p,d}^k} \tag{2}$$

with *n* representing the total number of measured peaks. From the above equation, it is seen that the scaling factor is simply the mean value over the *n* molecular locations for which the corresponding peak intensity can be estimated (*I av p*,*d* ), the mean value of the individual auto-scaled intensities being unitary, by definition [16,18]. Therefore, values of *A<sup>N</sup>* above or below unity indicate larger or smaller attenuations, respectively, with respect to the average absolute signal attenuation. From the definitions, the error on the individual *A<sup>N</sup>* values can be calculated as [16]

$$
\Delta A\_N^k = A\_N^k \times \sqrt{\left[\frac{\Delta I\_p^k}{I\_p^k}\right]^2 + \left[\frac{\Delta I\_d^k}{I\_d^k}\right]^2 + \left[\frac{\frac{1}{n^2}\sum\left(\Delta I\_p^k\right)^2}{\left(I\_p^{av}\right)^2}\right] + \left[\frac{\frac{1}{n^2}\sum\left(\Delta I\_d^k\right)^2}{\left(I\_d^{av}\right)^2}\right]}\tag{3}
$$

where the first two terms under the square root sign represent the error on the relative intensity (*RI*) of the *k*th residue signal, i.e., the signal intensity ratio in the presence and absence of nitroxide or/and AuNPs, and the ∆*I* are the experimental intensity uncertainties obtained from the individual peak signal-to-noise figure.

The ESR spectra were employed to extract the rotational correlation time (τ*c*) of Tempol in absence or presence of β2m or/and AuNPs. Based on the method of Knowles and colleagues [32] and Kivelson's theoretical analysis [33], the τ*<sup>c</sup>* values were estimated from:

$$
\pi\_{\varepsilon} = 6.5 \times 10^{-10} \,\,\Delta B\_0 [(h\_0/h\_{-1})^{1/2 - 1}] \tag{4}
$$

where ∆*B*<sup>0</sup> is the linewidth of the of the central line of the nitroxide ESR signal (a triplet because of the hyperfine coupling with the <sup>14</sup>N nuclear spin), and *h*0/−<sup>1</sup> are the amplitudes of the central and upfield lines. The corresponding error was calculated from Equation (4) by error propagation of the experimental uncertainties on linewidth (∆∆*B*0) and amplitudes (∆*h*0/−1), according to:

$$
\Delta \tau\_{\mathcal{E}} = 6.5 \cdot 10^{-10} \left\{ \left[ \left( \frac{h\_0}{h\_{-1}} \right)^{\frac{1}{2}} - 1 \right] \Delta \Delta B\_0 + \frac{\Delta B\_0}{2 \left( h\_{-1} \right)^{\frac{1}{2}} \left( h\_0 \right)^{\frac{1}{2}}} \Delta h\_0 + \frac{\Delta B\_0 (h\_0)^{\frac{1}{2}}}{2 \left( h\_{-1} \right)^{\frac{3}{2}}} \Delta h\_{-1} \right\} \tag{5}
$$

#### **5. Conclusions**

The results here described demonstrate that the paramagnetic perturbation methodology can be successfully applied to study protein-nanoparticle interactions. In addition to the surface accessibility mapping, extrinsic paramagnetic probes can provide valuable information on hindered accessibility and exchange processes by means of off-equilibrium attenuation analysis [16]. This methodology represents an additional tool that enriches the NMR relaxation approach to the characterization of protein interaction with the nanoparticle surface [34]. The delineation of the contact interface between protein monomers and between protein and nanoparticles is important not only for the comprehension of the mechanisms of protein aggregation and the elaboration of contrast strategies that bear particular relevance in amyloidogenic systems, but also for the characterization of that ensemble of labile contacts that is involved in the build-up of the so-called soft corona, i.e., the coating layer of weakly bound protein molecules with short residence times that can affect the nanoparticle targeting [35].

**Supplementary Materials:** The following are available online, Figure S1: overlay of ESR spectra; Figure S2: chemical shift perturbation values of β2m amide signals under different conditions; Figure S3: relative intensity plots of 8 µM + 0.8 mM Tempol under magnetization equilibrium and off-equilibrium conditions; Figure S4: <sup>15</sup>N longitudinal relaxation times of β2m amide signals under different conditions. Figure S5: <sup>15</sup>N-1H HSQC maps; Table S1: *A<sup>N</sup>* values plotted in Figure 1 and corresponding errors; Table S2: *A<sup>N</sup>* values plotted in Figure 2 and corresponding errors; Table S3: individual peak signal-to-noise ratios of <sup>15</sup>N-1H HSQC spectra.

**Author Contributions:** Conceptualization: G.E., Y.H., C.C., F.F.; Methodology: G.E., Y.H., C.C., F.F.; Formal analysis: Y.H., G.E., F.F.; Investigation: Y.H., G.E., F.F.; Resources: C.C., S.G., L.M.; Writing—Original Draft Preparation: G.E., Y.H.; Writing—Review and Editing: S.G., L.M., C.C., F.F., Y.H., G.E.; Supervision: G.E., F.F.; Project Administration: G.E., Y.H., F.F.; Funding Acquisition: G.E. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by NYUAD (grant No. 76 71260 ADHPG VP046).

**Acknowledgments:** We thank H. Molinari for stimulating the interest in the subject and the Core Technology Platform of NYUAD for the instrumentation. The assistance of Makek, A. is also gratefully acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest. The funder had no role in the design of the project, in the collection, analysis and interpretation of the data, in the writing of the paper, or in the decision to publish the results.

#### **References**


**Sample Availability:** Samples of Gold nanoparticles are commercially available from Sigma. Protein samples must be expressed and are currently not available from the authors.

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Protein Adsorption on Solid Supported Membranes: Monitoring the Transport Activity of P-Type ATPases**

#### **Francesco Tadini-Buoninsegni**

Department of Chemistry "Ugo Schiff", University of Florence, 50019 Sesto Fiorentino, Italy; francesco.tadini@unifi.it; Tel.: +39-055-4573239

Academic Editor: Michael Assfalg Received: 6 July 2020; Accepted: 2 September 2020; Published: 11 September 2020

**Abstract:** P-type ATPases are a large family of membrane transporters that are found in all forms of life. These enzymes couple ATP hydrolysis to the transport of various ions or phospholipids across cellular membranes, thereby generating and maintaining crucial electrochemical potential gradients. P-type ATPases have been studied by a variety of methods that have provided a wealth of information about the structure, function, and regulation of this class of enzymes. Among the many techniques used to investigate P-type ATPases, the electrical method based on solid supported membranes (SSM) was employed to investigate the transport mechanism of various ion pumps. In particular, the SSM method allows the direct measurement of charge movements generated by the ATPase following adsorption of the membrane-bound enzyme on the SSM surface and chemical activation by a substrate concentration jump. This kind of measurement was useful to identify electrogenic partial reactions and localize ion translocation in the reaction cycle of the membrane transporter. In the present review, we discuss how the SSM method has contributed to investigate some key features of the transport mechanism of P-type ATPases, with a special focus on sarcoplasmic reticulum Ca2+-ATPase, mammalian Cu+-ATPases (ATP7A and ATP7B), and phospholipid flippase ATP8A2.

**Keywords:** sarcoplasmic reticulum Ca2+-ATPase; Cu+-ATPase; phospholipid flippase; charge displacement; concentration jump; solid supported membrane; conformational transition; electrogenicity; ion translocation; phospholipid flipping

#### **1. Introduction**

P-type ATPases constitute a superfamily of membrane transporters that are present in all forms of life and are located in various membrane types, such as the plasma or cellular organelle membranes. The superfamily of P-type ATPases is classified into five distinct subfamilies (P1–P5), which are specific to different substrates [1–3]. These enzymes use the energy provided by ATP hydrolysis to transport various ions or phospholipids across cellular membranes, thereby generating and maintaining essential electrochemical potential gradients.

P-type ATPases share a similar molecular architecture, which comprises three distinct cytosolic domains, i.e., the actuator (A), nucleotide binding (N) and phosphorylation (P) domains, and two transmembrane domains, the transport domain of six helical segments (TM1 to TM6), which contains the ion binding sites located halfway through the membrane, and a class-specific support domain of four helical segments (TM7 to TM10). Moreover, in many P-type ATPases, the N- or C-terminal extensions at the cytosolic side act as regulatory (R) domains, which are autoinhibitory or can function as sensors for the transported cations [3,4]. Interestingly, the R domains of P-type ATPases have the characteristics of disordered proteins and are therefore highly variable and flexible. The disordered structure of the R domains is likely to facilitate their regulatory function favoring interaction with binding partners and helping to stabilize particular enzyme conformations [4,5].

P-type ATPases couple ion transport and ATP hydrolysis in a cyclic sequence of partial reactions that constitute the catalytic cycle. During catalysis, a transient phosphorylated intermediate is formed by the interaction of ATP with a conserved aspartate residue in the P domain, which is a specific feature of P-type ATPases. The Albers–Post or E1–E<sup>2</sup> scheme [6,7] is the generally accepted model of the catalytic cycle of P-type ATPases. According to this model, the ATPase protein can assume two main conformational states, denoted E<sup>1</sup> and E2, with different affinity for the transported ions and accessibility of the ion binding sites to the cytoplasmic and extracellular/luminal side. During the catalytic cycle, the ATPase undergoes structural rearrangements and conformational transitions between E<sup>1</sup> and E<sup>2</sup> states to perform ATP-driven transport of ions or phospholipids across the membrane [3,4].

The molecular mechanism of transport by P-type ATPases has been described in several reviews, see e.g., [1,3,4,8–10]. Figure 1 shows a simplified diagram of sequential reactions in the catalytic cycle of sarcoplasmic reticulum Ca2+-ATPase (SERCA) [11]. Starting at the E<sup>1</sup> conformation, the SERCA cycle includes initial enzyme activation by high-affinity binding of two Ca2<sup>+</sup> ions from the cytoplasmic side, followed by enzyme phosphorylation by ATP and the formation of a high-energy phosphorylated state E1~P (an ADP-sensitive phosphorylated intermediate that retains sufficient chemical energy to be able to transfer the phosphate to ADP, thus forming ATP). A conformational transition from the E1~P state to the lower energy phosphoenzyme intermediate E2P (an ADP-insensitive phosphorylated intermediate whose relatively low energy is suggested by its non-reactivity with ADP) favors the translocation of Ca2<sup>+</sup> ions across the membrane and their release into the sarcoplasmic reticulum (SR) lumen in exchange for two luminal protons. Hydrolytic cleavage of the phosphoenzyme E2P (dephosphorylation) is followed by proton translocation and release to the cytosolic side, thus accelerating the E<sup>2</sup> to E<sup>1</sup> conformational transition, which completes the catalytic and transport cycle. Following the first high-resolution crystal structure of SERCA with bound Ca2<sup>+</sup> (i.e., the E1·Ca<sup>2</sup> state) [12], several crystal structures of SERCA in different conformational states in the transport cycle have been determined at atomic resolution, as reviewed in e.g., [10,13–19].

**Figure 1.** Simplified diagram of sequential reactions in the transport cycle of sarcoplasmic reticulum Ca2+-ATPase (SERCA).


#### **2. Current Measurements on Solid Supported Membranes**

The SSM represents a convenient model system for a lipid bilayer membrane. In particular, the SSM consists of a hybrid alkanethiol/phospholipid bilayer supported by a gold electrode. The SSM is formed by covering the gold surface with an alkanethiol monolayer, usually an octadecanethiol

monolayer, and then by self-assembling a phospholipid monolayer on top of the gold-supported thiol layer [21,22]. The so-formed hybrid bilayer (Figure 2) is characterized by a high mechanical stability so that fast solution exchange can be performed at the SSM surface. The exchange of solutions provides the substrate or ligand and activates the membrane transporter adsorbed on the SSM [22].

Δ **Figure 2.** Schematic diagram of an sarcoplasmic reticulum (SR) vesicle containing Ca2+-ATPase and of a membrane fragment incorporating Na+,K+-ATPase adsorbed on a solid supported membranes (SSM) and subjected to an ATP concentration jump. If the ATP jump induces charge movement across the ATPase, a compensating electrical current flows along the external circuit (the red spheres represent electrons) if the potential difference (∆V) applied across the whole system is kept constant. RE, reference electrode. Adapted with permission from [23]. Copyright 2009 American Chemical Society.

Various membrane preparations containing the transport protein of interest, i.e., native membrane vesicles, purified membrane fragments, and proteoliposomes with reconstituted proteins, can be physically adsorbed on the SSM (Figure 2). Adsorption of such membrane preparations allows a variety of transport proteins to be immobilized on the SSM surface in a simple spontaneous process. This experimental approach is much easier and more effective than direct incorporation of the membrane transporter in a free-standing planar lipid bilayer, such as the black lipid membrane, which requires complicated incorporation procedures, leading to a superior signal-to-noise ratio and time resolution of the electrical measurement.

Following stable adsorption of the membrane sample on the SSM, the membrane transporter is subjected to a substrate concentration jump through the solution exchange technique. A rapid exchange from a solution with no substrate for the membrane transporter to one containing a specific substrate, e.g., ATP for P-type ATPases, activates the transport protein. If the substrate concentration jump induces charge displacement across the protein, an electrical current is measured due to capacitive coupling between the membrane sample and the SSM [20,24,25]. In particular, movement of a net charge across the activated protein is compensated by a flow of electrons along the external circuit toward the electrode surface, to keep constant the potential difference (∆V) applied across the whole metal/solution interphase (Figure 2) [20]. This flow of electrons corresponds to the measured capacitive current, which is strictly correlated with the transporter-generated current and is recorded as a transient current signal [20,24,25]. The SSM method allows the measurement of charge displacement under pre-steady state conditions, while steady-state currents are not recorded. We point out that the electrical

behavior of the system is essentially the same whether membrane vesicles or membrane fragments are adsorbed on the SSM.

The transport mechanism of various P-type ATPases belonging to different subfamilies was characterized using the SSM technique, such as in the case of Na+,K+-ATPase [22,26], SERCA [27,28], and H+,K+-ATPase [29], belonging to the P2-ATPase subfamily, and more recently bacterial and mammalian Cu+-ATPases of subclass P1B [30,31] and P4-ATPase phospholipid flippase [32]. On the other hand, the P3-ATPase subfamily, which comprises plasma membrane H+-ATPases of fungal and plant cells, has not yet been investigated by the SSM method.

SSM measurements on P-type ATPases were useful to identify electrogenic steps, i.e., reaction steps associated with a net charge transfer, and to assign time constants to partial reactions in the ATPase transport cycle. However, slow transport processes with time constants greater than 200 ms can be hardly recorded in SSM-based current measurements [20].

Finally, the SSM technique has been successfully employed to evaluate the effects of pharmacologically relevant compounds, such as anti-cancer drugs [33], on the transport activity of P-type ATPases and to characterize the interaction of specific ATPase inhibitors, thereby providing a quantitative estimate of inhibition potency (IC<sup>50</sup> values).

Analysis systems for SSM-based electrophysiology are commercially available and are based on the SURFE2R (Surface Electrogenic Event Reader) technology, as described in [29,34–36]. When higher throughput is required as in the case of drug screening, a fully automated device allows measuring electrical currents simultaneously from 96 individual SSM sensors in a parallel mode.

#### **3. P-Type ATPases Investigated on Solid Supported Membranes**

As mentioned above, the SSM technique was used to investigate net charge translocation (electrogenic transport) in P-type ATPases. Charge displacement associated with specific steps, i.e., ion binding/release, ion translocation, and exchange was measured in the ATPase transport cycle and the electrogenicity of partial reactions was determined, thereby providing mechanistic insights in the transport mechanism of different P-type ATPases. For example, a direct proof for the electrogenicity of cytoplasmic Na<sup>+</sup> binding to the Na+,K+-ATPase was obtained with Na<sup>+</sup> concentration jump experiments performed on membrane fragments containing Na+,K+-ATPase adsorbed on the SSM [26]. It was found that the charge associated with the Na<sup>+</sup> binding step is about 30% of the displaced charge related to Na<sup>+</sup> translocation and release, indicating that cytoplasmic Na<sup>+</sup> binding is a minor electrogenic event in the reaction cycle of Na+,K+-ATPase [26].

In the next sections, we will discuss the contribution of the SSM technique to unravel key features of the electrogenic transport activity of some prominent members of the P-type ATPase family. In particular, the focus of the present review is on SERCA, Cu+-ATPases ATP7A and ATP7B, and P4-ATPase (phospholipid flippase) ATP8A2. SERCA has been characterized in detail by the SSM technique, providing useful information on the enzyme's transport mechanism. This information was used for a comparative analysis of the transport properties of the Cu+-ATPases and phospholipid flippase, which were recently investigated by the SSM method.

#### *3.1. Sarcoplasmic Reticulum Ca2*+*-ATPase*

The SERCA enzyme is one of the most investigated P-type ATPase [15,16,37–39]. In muscle cells, SERCA couples the energy gained by the hydrolysis of one ATP molecule to the transport of two Ca2<sup>+</sup> ions against their electrochemical potential gradient from the cytoplasm into the lumen of SR, which is the main intracellular Ca2<sup>+</sup> storage organelle. Ca2<sup>+</sup> uptake in the SR lumen by SERCA plays an essential role in regulating cytoplasmic Ca2<sup>+</sup> concentration, which is kept at or below 0.1 µM; in this manner, SERCA induces muscle relaxation and contributes to intracellular Ca2<sup>+</sup> homeostasis. Modified SERCA expression and impaired pumping activity have been associated with pathological conditions and several diseases with a wide range of severity [39,40].

SERCA (approximately 110 KDa) belongs to the P2A-ATPase subfamily. In mammals, SERCA is encoded by three different genes, ATP2A1-3, but isoform diversity is increased by alternative splicing of the transcripts, which raises the number of possible SERCA isoforms to more than 10 [41,42]. A very convenient experimental system for functional and structural studies of SERCA is provided by vesicular fragments of longitudinal SR, where SERCA1a is the predominant isoform. SR vesicles contain a high amount of SERCA, which accounts for approximately 50% of the total protein and which reaches a density in the SR membrane of about 30,000 µm−<sup>2</sup> [43]. −

Electrical currents generated by SERCA were measured by adsorbing native SR vesicles containing SERCA1a from rabbit skeletal muscle on the SSM and by activating the calcium pumps with substrate, i.e., Ca2<sup>+</sup> and ATP concentration jumps. The observed current signals allow the direct measurement of charge translocation by SERCA under different activation conditions. In particular, charge movements related to different electrogenic partial reactions in the SERCA transport cycle were detected. It was shown that a Ca2<sup>+</sup> concentration jump in the absence of ATP induces a transient current (dotted line in Figure 3A), which is associated with an electrogenic event corresponding to enzyme activation by the initial binding of Ca2<sup>+</sup> to the cytoplasmic side of the ATPase (the exterior of the SR vesicle, see Figure 2) [27,28,44]. When an ATP concentration jump was performed in the presence of Ca2<sup>+</sup> ions, a current signal was detected (solid line in Figure 3A), which is associated with a further electrogenic step corresponding to ATP-dependent calcium translocation by the enzyme [20,27]. In particular, ATP concentration jump experiments on SR vesicles in the presence and absence of a calcium ionophore at different pH values [27] indicated that the ATP-induced electrical current is related to displacement and release of pre-bound Ca2<sup>+</sup> at the luminal side of the pump (the interior of the SR vesicle, see Figure 2) after phosphorylation of the enzyme by ATP. The transient currents measured after a Ca2<sup>+</sup> jump in the absence of ATP and an ATP jump in the presence of Ca2<sup>+</sup> were both fully suppressed by thapsigargin [44], which is a highly specific and potent SERCA inhibitor [45,46]. We point out that to perform ATP hydrolysis and active Ca2<sup>+</sup> transport SERCA undergoes large domain movements enabled by dynamic fluctuations and conformational transitions that are not random but instead are driven by the availability of specific substrates [47].

**Figure 3.** Transient currents generated by SERCA adsorbed on an SSM. (**A**) Transient current after a 10 µM free Ca2<sup>+</sup> concentration jump in the absence of ATP (dotted line) and a 100 µM ATP concentration jump in the presence of 10 µM free Ca2<sup>+</sup> (solid line). Reprinted from [44] with permission. (**B**) Current signals after 100 µM ATP concentration jumps in the presence of 10 µM free Ca2<sup>+</sup> and 100 mM KCl at pH 7 (black line) and 7.8 (red line). The inset shows the dependence of the normalized charge (QN) after 100 µM ATP concentration jumps on pH. The charges were normalized with respect to the maximum charge measured at pH 7. S.E. are given by error bars. Adapted from [48].

It is interesting to observe that the amplitude of the signal related to ATP-dependent Ca2<sup>+</sup> translocation decreases as the pH is raised from 7 to 8 (Figure 3B). It is known that exchange of Ca2<sup>+</sup> with H<sup>+</sup> is a specific feature of SERCA [37,48], which favors Ca2<sup>+</sup> release at the luminal side [17,49]. Useful information was provided by previous measurements on reconstituted proteoliposomes containing SERCA [49–52]. In particular, it was shown that the stoichiometry of the Ca2+/H<sup>+</sup> countertransport is about 1/1 when the luminal and medium pH is near neutrality [49,52]. The importance of Ca2+/H<sup>+</sup> exchange in determining the release of bound Ca2<sup>+</sup> from the phosphoenzyme E2P was demonstrated in steady-state experiments on native SR vesicles [37]. It was reported that the maximal levels of accumulated Ca2<sup>+</sup> are significantly reduced if the pH is raised above 7. This result shows that if exchange is limited due to low H<sup>+</sup> concentration, Ca2<sup>+</sup> is less likely to dissociate from the phosphoenzyme. Thus, the pH dependence of the current signals obtained with ATP concentration jumps (inset of Figure 3B) also indicates that when a lack of H<sup>+</sup> limits Ca2+/H<sup>+</sup> exchange, i.e., alkaline pH, the translocation of bound Ca2<sup>+</sup> is prevented, even though K<sup>+</sup> is present in high concentration and may neutralize acid residues at alkaline pH [48]. This suggests a requirement for specific H<sup>+</sup> binding at the Ca2<sup>+</sup> transport sites in order to obtain Ca2<sup>+</sup> release.

The SSM method has also been used to investigate a very interesting research topic, which is currently receiving much attention, i.e., the molecular mechanisms of SERCA regulation. In muscle cells, SERCA transport activity is regulated by two analogous transmembrane proteins: phospholamban (PLN, 52 amino acids), which is primarily expressed in cardiac muscle where it regulates the SERCA2a isoform [53], and sarcolipin (SLN, 31 amino acids), which is mainly expressed in skeletal muscle where it regulates the SERCA1a isoform [54]. In particular, PLN inhibits pump activity by lowering the apparent Ca2<sup>+</sup> affinity of SERCA, and the phosphorylation of PLN by protein kinases relieves SERCA inhibition [53]. There is general consensus that the PLN inhibition of SERCA involves the reversible physical interaction of a PLN monomer under calcium-free conditions. However, experimental evidence was provided that a PLN pentamer, which has been described as an inactive storage form, can also interact with SERCA [55,56].

To investigate the PLN effect on ATP-dependent Ca2<sup>+</sup> translocation by SERCA, SSM-based current measurements were carried out on co-reconstituted proteoliposomes containing SERCA and PLN [57]. The proteoliposomes were adsorbed on the SSM and activated by Ca2<sup>+</sup> and/or ATP concentration jumps. In particular, substrate conditions (various Ca2<sup>+</sup> and ATP concentrations) were chosen that promoted specific conformational states of SERCA, from which calcium transport could be initiated. The results from pre-steady state charge (calcium) translocation experiments were compared with steady-state measurements of ATPase hydrolytic activity. It was found that the PLN effect on SERCA transport activity depends on substrate conditions, and PLN can establish an inhibitory interaction with multiple conformational states of SERCA (a calcium-free E<sup>2</sup> state, a E1-like state promoted by Ca2+, and a E2-like state promoted by ATP, shown in red in Figure 4) with distinct effects on SERCA's kinetic properties [57]. It was also noted that once a particular SERCA–PLN inhibitory interaction is established, it remains throughout the SERCA transport and catalytic cycle. These findings were interpreted on the basis of a conformational memory [58,59] in the interaction of PLN with SERCA, whereby a defined structural state of the SERCA/PLN regulatory complex, which depends on substrate conditions, is retained during SERCA turnover and conformational cycling.

In addition to PLN and SLN, single-span transmembrane proteins have recently been discovered that act as regulators of SERCA activity: dwarf open reading frame (DWORF), myoregulin (MLN), endoregulin (ELN), and another-regulin (ALN) [60–62]. While MLN, ELN, and ALN have been identified as inhibitors of SERCA activity, it was shown that DWORF does not inhibit the SERCA pump [62], enhancing Ca2<sup>+</sup> uptake by displacing PLN. The oligomerization of these new SERCA regulators and the binding interaction of the monomeric form with the calcium pump were very recently investigated [63], thus providing a useful contribution in the characterization of the complexity of SERCA regulatory mechanisms. In this respect, it appears that the above-mentioned transmembrane peptides could be conveniently investigated by the SSM technique upon their reconstitution in proteoliposomes containing SERCA. This would help to elucidate the inhibitory or activation effects of the recently discovered SERCA regulators.

**Figure 4.** The SERCA transport cycle with relevant conformational states. The pre-incubation and concentration jump conditions used [57] are indicated. Shown in red are the calcium-free E<sup>2</sup> state, an E<sup>1</sup> -like state promoted by calcium, and an E<sup>2</sup> -like state promoted by ATP. Adapted from [57].

#### *3.2. Cu*+*-ATPases ATP7A and ATP7B*

The mammalian copper ATPases ATP7A and ATP7B are 165–170 KDa membrane proteins belonging to subclass IB of the P-type ATPase superfamily. At normal copper levels in the cell, ATP7A and ATP7B are found in the trans-Golgi network (TGN), and these enzymes translocate copper across the membrane from the cytoplasm into the TGN lumen using ATP hydrolysis [64–68]. ATP7A and ATP7B contribute to intracellular copper homeostasis by delivering copper to newly synthesized copper-containing proteins in the TGN and by removing copper excess from the cell [64]. ATP7A is expressed in most tissues but not in the liver, whereas ATP7B is mainly found in this organ [64]. The malfunction of either ATP7A or ATP7B is the cause of severe diseases, which are known as Menkes (ATP7A) and Wilson (ATP7B) diseases.

ATP7A and ATP7B show high sequence homology (about 60% identity). Their structure comprises eight transmembrane helices, which include a copper binding site (transmembrane metal binding site, TMBS), and the A, N and P cytoplasmic domains, which are common for P-type ATPases. A unique structural feature of ATP7A and ATP7B is the highly mobile N-terminal chain of six copper binding domains (N-terminal metal binding domain) that are involved in the copper-dependent regulation and intracellular localization of these enzymes [69].

As described in several reviews (e.g., [64,65,70–74]), Cu<sup>+</sup> transfer by ATP7A and ATP7B involves copper acquisition from donor proteins on the cytoplasmic side of the membrane and copper delivery to acceptor proteins on the luminal side, without establishing a free Cu<sup>+</sup> gradient. In conformity with other P-type ATPases, ATP7A and ATP7B hydrolyze ATP to form a transient phosphorylated intermediate, and they undergo conformational transitions that favor Cu<sup>+</sup> transfer to/from the TMBS. From high-resolution crystal structures and molecular dynamics simulations on a bacterial Cu+-ATPase (*Legionella pneumophila* Cu+-ATPase, LpCopA) [75,76], it appears that copper ATPases have a unique copper release mechanism that is likely to be involved in specific and controlled Cu<sup>+</sup> delivery to acceptor proteins.

Electrogenic copper movement within mammalian copper ATPases was demonstrated by current measurements on COS-1 microsomes expressing recombinant Cu+-ATPases (ATP7A and ATP7B) adsorbed on an SSM [31,48]. When an ATP concentration jump was performed on microsomes containing ATP7B (or ATP7A) in the presence of CuCl<sup>2</sup> and dithiothreitol to reduce Cu2<sup>+</sup> to

Cu+, a current signal was obtained (solid line in Figure 5A), which was not observed when bathocuproinedisulfonate (BCS), acting as Cu<sup>+</sup> chelator, was added to the reaction buffer (dotted line in Figure 5A). These experiments indicate that the copper-related current signal is associated with an electrogenic event corresponding to Cu<sup>+</sup> movement within ATP7B upon phosphorylation by ATP [31,48], which is consistent with copper displacement from the TMBS to the luminal side of the enzyme.

**Figure 5.** Transient currents generated by ATP7B adsorbed on an SSM. (**A**) Transient currents after 100 µM ATP concentration jumps in the presence of 5 µM CuCl<sup>2</sup> (solid line) or 1 mM bathocuproinedisulfonate (BCS) (dotted line). The inset shows current signals after 100 µM ATP concentration jumps on ATP7B (red line) and SERCA (black line). Adapted from [31] with permission from Wiley. (**B**) Current signals after 100 µM ATP concentration jumps in the presence of 5 µM CuCl<sup>2</sup> at pH 6 (black line) and 7.8 (red line). Adapted from [48].

By fitting the decay phase of the transient current with a first-order exponential decay function, a charge transfer decay time constant (τ) of 140 ms was determined for ATP7B, which is within the time frame of aspartate phosphorylation by ATP [31], suggesting that copper displacement in ATP7B is correlated to formation of the phosphorylated intermediate and precedes phosphoenzyme hydrolytic cleavage. This conclusion was also supported by SSM-based current measurements on the D1044A mutant of ATP7A. Asp1044 is the conserved aspartate residue in the P-domain of ATP7A that interacts with ATP to form the aspartyl phosphorylated intermediate. It was shown that the D1044A mutant yielded no current signal upon an ATP concentration jump in the presence of Cu<sup>+</sup> [77]. This result further indicated that ATP-dependent copper movement through the ATPase is directly correlated to formation of the aspartyl phosphorylated intermediate by ATP consumption.

It is interesting to observe that ATP-induced copper movement in mammalian Cu+-ATPases is significantly slower than ATP-dependent Ca2<sup>+</sup> translocation in SERCA [31], as shown by the different decay time constants τ for charge displacement following ATP jumps (inset of Figure 5A) on ATP7B (red line, τ = 140 ms) and SERCA (black line, τ = 25 ms). It is worth mentioning that the τ values for charge movements in ATP7B and SERCA are consistent with a slower phosphoenzyme formation in the copper ATPase [31] with respect to SERCA [78]. It should be noted that these decay time constants are attributed to initial partial reactions of the pump transport cycle and are not equivalent to steady-state turnover [31].

SSM measurements on ATP7A and ATP7B revealed that ATP-induced charge movement in these enzymes is not changed by alkaline or acid pH [48], as shown by charge transfer measurements at different pH values (Figure 5B). This finding indicated that copper displacement in ATP7A and ATP7B is pH independent, and it highlights a significant difference in the transport mechanisms of ATP7A/B and SERCA. It was proposed that in ATP7A/B, Cu+/H<sup>+</sup> exchange may not be required for luminal copper release [48], as opposed to the strict requirement of Ca2+/H<sup>+</sup> exchange in SERCA as discussed above. It is worth mentioning that carboxylate residues are absent in the ion-binding cluster located in the transmembrane region of the bacterial *Archaeoglobus fulgidus* CopA [79] and LpCopA [30], while crucial aspartate and glutamate residues are present in the equivalent transmembrane domain of SERCA [12,16,80] that are directly involved in Ca2+/H<sup>+</sup> exchange. Thus, SSM measurements on ATP7A/B supported the hypothesis that Cu<sup>+</sup> release in these enzymes may not be coupled to a net proton countertransport, which has not been observed for PIB-type ATPases [72,73,81]. Interestingly, a very recent study reported real-time fluorescence measurements on *E.coli* Cu+-ATPase (EcCopA) reconstituted in small unilamellar vesicles encapsulating a set of fluorescence probes that are selective for Cu+, pH, and membrane potential [82]. The results of this study demonstrated the absence of H<sup>+</sup> countertransport in the Cu<sup>+</sup> translocation cycle of EcCopA, qualifying EcCopA as an electrogenic uniporter.

#### *3.3. P4-ATPase ATP8A2*

A characteristic feature of eukaryotic cell membranes is the asymmetrical distribution of different lipids across the membrane bilayer. This is particularly evident in the plasma membrane, where phosphatidylcholine (PC) and sphingolipids, i.e., sphingomyelin and glycosphingolipids, are concentrated in the exoplasmic leaflet of the membrane, whereas phosphatidylserine (PS) and phosphatidylethanolamine (PE) are mainly restricted in the cytoplasmic leaflet [83–86]. Membrane lipid asymmetry is essential for a variety of cellular processes that include, e.g., cell and organelle shape determination, membrane stability and impermeability, membrane curvature, vesicle formation and trafficking, host–virus interactions, membrane protein regulation, blood coagulation, and apoptosis [86–90].

Phospholipid flippases, belonging to the P4-ATPase subfamily, couple ATP hydrolysis to the translocation of specific phospholipids from the exoplasmic to the cytoplasmic leaflet of biological membranes in order to generate and maintain transmembrane lipid asymmetry [89,91–95]. P4-ATPases are only found in eukaryotes and constitute the largest P-type ATPase subfamily. In mammals, at least 14 P4-ATPases are known, which are divided into five classes [89]. P4-ATPase dysfunction has been associated with severe neurological disorders and liver diseases in humans [92]. These lipid transporters consist of a large polypeptide with a molecular mass of about 120 kDa, which shares the general architecture of P-type ATPases. Most P4-ATPases form a heterodimeric complex with an accessory β-subunit of about 50 kDa belonging to the CDC50/LEM3 family [89,96,97]. High-resolution structures of yeast [98,99] and human [100] lipid flippases were determined by cryo-electron microscopy, as reviewed in [101], and very recently, the crystal structures of a human plasma membrane flippase were also reported [102].

The transport mechanism of P4-ATPases is the subject of intensive research, and various models have been proposed for the phospholipid translocation pathway in P4-ATPases [103–107]. The recent atomic resolution structures of yeast and human P4-ATPases have provided valuable information on different conformational states in the flippase transport cycle, which is depicted by the Albers–Post or E1–E<sup>2</sup> scheme commonly used to describe the mechanism of ion transporting P2-type ATPases. The P4-ATPase reaction cycle (see the simplified diagram in Figure 6A) has been examined in some detail for the mammalian flippase ATP8A2 [108], which is highly expressed in the retina, brain, testis, and spinal cord. It was shown that ATP8A2 is phosphorylated by ATP at the aspartate conserved in all P-type ATPases, and the phosphoenzyme exists in E1P and E2P states [108]. Dephosphorylation of the E2P state is activated by binding of the two known substrates PS and PE, but not by binding of PC that is not a substrate of ATP8A2 [109], and dephosphorylation is associated with lipid translocation from the exoplasmic to the cytoplasmic leaflet of the membrane bilayer. Although significant progress has been made in our understanding of phospholipid flipping by P4-ATPases, several aspects of the flippase transport mechanism remain to be explored, such as the stoichiometry of phospholipid molecules translocated per ATP hydrolyzed, the mechanisms underlying lipid binding and release, the electrogenicity of phospholipid transport, and the related issue of countertransport, i.e., countertransport of an ion or other charged substrate from the cytoplasm to the exoplasm in connection with the E<sup>1</sup> → E1P → E2P reaction sequence as observed for P2-type ATPases.

→ →

**Figure 6.** Simplified diagram of the ATP8A2 reaction cycle and transient currents generated by ATP8A2 adsorbed on a SSM. (**A**) E<sup>1</sup> , E1P, E2P, and E<sup>2</sup> are different enzyme conformational states, where "P" indicates covalently bound phosphate. The phospholipid substrate, PL (phosphatidylserine (PS) or phosphatidylethanolamine (PE)), enters the cycle from the exoplasmic leaflet of the lipid bilayer by binding to the E2P phosphoenzyme, thereby stimulating the dephosphorylation and release of the lipid substrate toward the cytoplasmic leaflet as a consequence of the E<sup>2</sup> to E<sup>1</sup> conformational change. (**B**) Current transients observed upon 100 µM ATP concentration jumps on ATP8A2 reconstituted in proteoliposomes containing a mixture of 90% PC and 10% PS, in the absence (black line) or in the presence (red line) of 50 µM orthovanadate. The inset shows the current signal induced by a 100 µM ATP concentration jump on native SR vesicles containing SERCA. Reprinted from [32].

To address unexplored key aspects of the flipping mechanism of P4-ATPases, in particular the electrogenicity of phospholipid flippases and ion countertransport, the SSM method was very recently used in a study of the elctrogenic properties of wild-type and mutant forms of the flippase ATP8A2 [32]. Purified ATP8A2 and its accessory CDC50A protein were reconstituted in proteoliposomes of different lipid compositions that were adsorbed on the SSM surface and subjected to ATP concentration jumps. It was shown that an ATP jump on ATP8A2 reconstituted into proteoliposomes consisting of a mixture of 90% PC and 10% PS induced a current signal (black line in Figure 6B) that was completely suppressed in the presence of the ATPase inhibitor orthovanadate (red line Figure 6B). Since orthovanadate binds to the ATPase from the cytoplasmic side, it was concluded that the ATPase molecules with the cytoplasmic side facing the external aqueous solution generated the ATP-dependent charge movement across ATP8A2. It was also noted that the sign of the ATP8A2-related current signal is positive, as observed for the SERCA-related transient current (inset of Figure 6B) that is attributed to the translocation and release of Ca2<sup>+</sup> ions into the SR vesicle interior [27] (see Section 3.1). We point out that the movement of positive charge in one direction is electrically equivalent to the displacement of negative charge in the opposite direction. Therefore, the ATP8A2 current signal was associated with ATP-dependent translocation of the negatively charged PS toward the outside of the proteoliposomes (the ATP8A2 cytoplasmic side facing the external aqueous solution) [32].

It is worth noting that PC, also present in the proteoliposomes, is not a substrate for ATP8A2 and has an electrically neutral head group. As mentioned above, PC does not stimulate ATP8A2 dephosphorylation; however, the enzyme can be phosphorylated by ATP in the absence of PS and PE and in the presence of PC alone [108]. Interestingly, an ATP concentration jump on ATP8A2 reconstituted in proteoliposomes containing 100% PC yielded no significant transient current, indicating that enzyme phosphorylation by ATP (E1→E1P reaction step) does not generate any electrical signal [32].

Useful information was also provided by ATP concentration jump experiments on proteoliposomes containing selected mutants of ATP8A2 [32]. In particular, the E198Q mutation was examined. Glu-198 is located in the DGET motif of the cytoplasmic A-domain of ATP8A2 that facilitates dephosphorylation of the phosphorylated intermediate. It was reported that phosphorylation by ATP is allowed in E198Q, while dephosphorylation is blocked with resulting E2P accumulation [108]. The absence of an electrical current upon an ATP concentration jump on proteoliposomes containing E198Q indicated that the electrogenicity of ATP8A2 is not related to the E1→E1P step (phosphorylation by ATP) or with the E1P→E2P conformational transition of the enzyme, which is in agreement with the experiments on 100% PC proteoliposomes. This finding is important to address the issue of whether ion countertransport occurs from the cytoplasmic to the exoplasmic side of the phospholipid flippase. In fact, it was shown that no charged substrate is countertransported during the E<sup>1</sup> → E1P → E2P reaction sequence [32], thereby distinguishing the P4-ATPase from the P2-type ATPases, which transport ions in opposite directions and are therefore referred to as antiporters.

It was concluded that the electrogenicity of ATP8A2 is related to a step in the ATPase transport cycle directly involved in translocation of the phospholipid [32]: dephosphorylation of the E2P intermediate, activated by lipid binding from the exoplasmic side, and/or the subsequent E<sup>2</sup> → E<sup>1</sup> conformational transition of the dephosphoenzyme, which is associated with release of the lipid to the cytoplasmic side [104]. It is noteworthy that the findings of the SSM study of the mammalian phospholipid flippase [32] support the view that the P4-ATPase is most likely an electrogenic uniporter, as also recently reported for a bacterial Cu+-transporting P1B-ATPase [82].

#### **4. Conclusions**

The charge transport mechanism of various P-type ATPases has been conveniently investigated by adsorbing the membrane-bound ATPase on the SSM and by activating the enzyme with a concentration jump of a suitable substrate. The transient current observed with the SSM method is a direct measurement of charge movements occurring during reactions involved in the transport cycle of the ATPase. This kind of measurements allows the identification of electrogenic partial reactions, which in turn can be used to localize charge translocation in the reaction cycle of the ion/lipid pump. It was shown that charge displacement in P-type ATPases is associated with transitions between different conformational states that facilitate the binding or release of the charged substrate.

The SSM method can provide useful information about the transport activity of P-type ATPases, as well as the molecular mechanisms of ATPase regulation, as discussed in the case of the SERCA enzyme. This is actually a very interesting research topic, which has yet to be examined in detail for some P-type ATPases, such as mammalian P4-ATPases [89]. However, also for the well-characterized P2-type ATPases Na+,K+-ATPase, and SERCA, a complexity of regulatory mechanisms has emerged [4,19], which requires further detailed investigation. We think that the SSM method can be usefully employed to characterize the interaction of P-type ATPases with specific regulatory partners, which can be small molecules, soluble proteins, transmembrane peptides, or the surrounding lipid bilayer [110].

Another interesting field of application of the SSM technique is related to the drug discovery process. It has been shown that drug/protein interactions can be conveniently monitored on SSMs. In particular, the SSM technique has demonstrated its usefulness to investigate the effects of various pharmaceutically relevant compounds on P-type ATPases [29,33,111], which are important targets for a variety of drugs [112–114]. Therefore, this technique represents a robust and reliable assay in drug development and evaluation studies on membrane transport proteins, as it can be used to quantify the effectiveness and potency of drugs directed toward specific protein targets, and to characterize novel drug candidates.

**Funding:** This research received no external funding.

**Acknowledgments:** Fondazione CR Firenze ID 2018.0886 for financial support.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**

1. Kühlbrandt, W. Biology, structure and mechanism of P-type ATPases. *Nat. Rev. Mol. Cell Biol.* **2004**, *5*, 282–295. [CrossRef] [PubMed]


© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Molecules* Editorial Office E-mail: molecules@mdpi.com www.mdpi.com/journal/molecules

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-0365-2690-4