Ligand uptake in Mycobacterium tuberculosis truncated hemoglobins is controlled by both internal tunnels and active site water molecules [version 1; peer review: 1 approved, 1 approved with reservations]

Mycobacterium tuberculosis, the causative agent of human tuberculosis, has two proteins belonging to the truncated hemoglobin (trHb) family. Competing interests: No competing interests were disclosed. retained water molecules at the active site. We found that a single mutation allows Mt-trHbN to acquire ligand migration rates comparable to those observed for Mt-trHbO, confirming that ligand migration is regulated by the internal tunnel architecture as well as by water molecules stabilized in the active site. This is a succinct and generally well written paper reporting experimental data on ligand binding to Mt-trHbN and to two mutants (VG8F anndVG8W)).These latter have been designed to test hypotheses regarding the possible routes by which access is gained to the heme by small neutral gaseous molecules. The experiments and the molecular modelling that supports them, and which provides mechanistic insights, have been carefully performed. The results are of interest to the field as they add to the body of accumulated evidence that proteins, including those with the function of binding small neutral molecules, provide specific, and often dynamic, channels to permit rapid access to the binding site. Furthermore the kinetics of binding are seen to be strongly influenced by single amino acid substitution in the access channels. the results the This manuscript describes CO association kinetic constant measurements, ● NO decomposition, and molecular dynamics simulations on the wild type truncated Hb from Mycobacterium tuberculosis (Mt-trHbN) and two mutants (VG8F and VG8W) which introduce modifications in the two-tunnel system of the protein. The data are then compared to those from Mt-trHbO, suggesting that ligand migration is regulated by the internal tunnel architecture as well as by water molecules stabilized in the active site. Abstract: line 11 Mt-trHbN should not be in Italics. Abstract: line 12


Introduction
Mycobacterium tuberculosis, the causative agent of human tuberculosis, affects approximately two billion people world-wide, causing over three millions deaths each year 1 . The genome of this pathogenic organism includes two genes, glbN and glbO, which encode for two proteins, termed here truncated hemoglobin N (Mt-trHbN) and truncated hemoglobin O (Mt-trHbO), belonging to the truncated hemoglobin (trHb) family of heme proteins, widely distributed in eubacteria, cyanobacteria, microbial eukaryotes and plants 2,3 .
The truncated hemoglobin family exhibits a three-dimensional structure similar to the common globin fold of myoglobin, but significantly smaller. The secondary structure of trHbs consists of four α-helices arranged in a two-over-two antiparallel sandwich instead of the common three-over-three helix globin fold. Phylogenetic analysis has distinguished three different groups of truncated hemoglobins, classified as groups I, II and III, also called N, O and P, respectively.
It has been shown that group I Mt-trHbN catalyzes the detoxification of • NO in the presence of O 2 4,5 . The first step of this mechanism involves O 2 migration and binding. Subsequently, • NO migrates to the active site and reacts with the heme-bound O 2 to yield an unstable peroxynitrite adduct, which isomerizes to generate the relatively innocuous nitrate anion.
Several studies have examined the role of internal tunnels in ligand migration in trHbs 2,[5][6][7][8] . Three internal tunnels were found in the truncated hemoglobin family: a long tunnel (LT) topologically positioned between helices B and E, and two short tunnels, known as the E7 Gate (E7 gate) and the short tunnel G8 (STG8), which are roughly normal to the LT, as depicted in Figure 1. The E7 tunnel corresponds to the highly conserved E7 pathway widely studied in both myoglobin and hemoglobin [9][10][11] . The STG8 tunnel is analogous to that found in Mt-trHbN, next to the key residues VG8 and IH11.
Previous results indicate that WG8, an absolutely conserved residue in groups II and III truncated hemoglobins, is involved in hindering ligand migration in Mt-trHbO by blocking both STG8 and LT ( Figure 1) [12][13][14] . In addition, the presence of a smaller residue at the G8 position in the Mt-trHbO mutant (WG8F) was observed to increase the small ligand association constant, although the molecular details of this process were not investigated [12][13][14] . It has also been noted that internal water molecules can block the heme accessibility, thus delaying ligand binding [15][16][17] .
By performing CO association kinetic constant measurements (k on CO), • NO decomposition, and molecular dynamics (MD) simulations, we addressed molecular mechanisms that control ligand association in M. tuberculosis truncated hemoglobins.

Protein purification
All chemicals and reagents were obtained from Sigma Aldrich, unless indicated otherwise. The trHbN protein was purified using standard techniques reported for other bacterial globins 19 . Briefly, mutated constructs were used to transform E. coli BL21 DE3 pLysS. Starter cultures grown overnight in LB supplemented with kanamycin (50 µg ml -1 ) and chloramphenicol (35 µg ml -1 ) were used to inoculate 6 batches of 1 L LB medium at 1% (v/v), supplemented with kanamycin and 3 µM FeCl 3 . Once cultures reached an OD 600 of around 0.4, expression of trHbN was initiated by the addition of 1 mM IPTG and grown for a further 4 h. Cells were harvested by centrifugation at 5500 rpm for 20 min at 4°C and stored overnight at -20°C. After thawing, cells were resuspended in 40 ml buffer (10 mM TRIS-HCl (pH 7.0) with 1 mM EDTA, 10 mM DTT, 45 µg ml -1 phenylmethylsulphonyl fluoride, 500 µg ml -1 RNase and 50 µl DNase), homogenized using a Douce homogeniser and Figure 1. Schematic representation of the potential tunnels described in truncated hemoglobins. The three pathways, Long Tunnel (LT), E7 Gate (E7 gate) and Short Tunnel G8 (STG8) for ligand migration through the tertiary structure of a typical trHb are shown. ultracentrifuged at 44,000 rpm for 1 h at 4°C. The supernatant, red in color, was loaded onto a 30 ml DEAE Sepharose Fast Flow column (Pharmacia Biotech) equilibrated with 10 mM TRIS-HCl (pH 7.0), washed with the same buffer until the UV trace returned to baseline, and eluted via a gradient from 0 to 1 M NaCl in 10 mM TRIS-HCl using an Akta purifier (GE Healthcare Bio-Sciences, Amersham Biosciences, U.K. Ltd.). Fractions that were most red in color were concentrated using a Vivaspin 20 concentrator (Sartorius Stedim Biotech) to around 5 ml and loaded onto a gel filtration Superdex 75 column, equilibrated with 0.15 M NaCl in 10 mM TRIS-HCl (pH 7.5); again, fractions with the most color were collected, combined and stored at -80°C. Purity was checked using gel electrophoresis and analysis of the heme-to-protein ratio (410 nm and 280 nm in the UV-visible absorption spectrum).

Kinetic stopped flow measurements of CO binding
Rapid mixing experiments were conducted with a thermostated stopped flow apparatus (BioLogic SFM-300). Kinetics of carbon monoxide (CO) binding to determine the k on CO were measured on the deoxy state of mutant and wild type globins at 20ºC. Solutions containing 5 µM protein in a 100 mM sodium phosphate at pH 7.0 were degassed in a nitrogen atmosphere and reduced with an equimolar concentration of sodium dithionite and mixed with increasing CO concentrations. The observed pseudo first-order rate constant (k obs ) was determined by fitting the absorbance decay resulting from association of the protein with CO, to a single exponential function. Kinetic rate constants (k on CO) were obtained from the slope of the plots of k obs as a function of CO concentration.

NO decomposition
To determine rates of nitric oxide ( • NO) decomposition by wild type and mutant Mt-trHbN proteins, • NO was added, as ProliNON-Oate, to a solution of 50 mM KPi buffer with 50 µM EDTA (pH 7.5), 100 µM NADPH and 100 nM E. coli ferredoxin reductase inside a thermoregulated, magnetically stirred reaction vessel. Mt-trHbN (2 µM) was added at the apex of the signal response to 2 µM ProliNONOate and • NO decay was followed until depleted using an • NO electrode (World Precision Instruments). Rates of • NO decay were calculated for each protein by determining the time taken for peak • NO concentrations to decay by 0.5 µM and were expressed per µM heme, determined spectrally by the peak in the Soret region at 410 nm.

Set up of the simulations
The starting structure corresponds to Mt-trHbN crystal structure (PDB entry 1IDR http://www.rcsb.org/pdb/explore/explore. do?structureId=1IDR), at 1.9 Å of resolution 20 . Amino acids protonation states were assumed based on environment of the residue in the crystal structure. All the solvent-exposed His were protonated at the N-δ delta atom, as well as HisF8, because of its coordination to the heme iron. An octahedral box of 10 Å of radius, which corresponds to 5234 explicit water molecules was added to the system. TIP3P water molecules were used by tLEaP module of the AMBER12 package 21 . The param99 Amber force field was used for all the aminoacid parameters 22 except heme parameters which were developed in our group 23 and strongly validated for being used in several studies of heme proteins 24-30 . Periodic boundary conditions were used for all the simulations performed with a 9 Å cutoff. Particle mesh Ewald (PME) summation method for treating the electrostatic interactions. The SHAKE algorithm was used to keep constant the non-polar hydrogen equilibrium distance. Temperature and pressure were kept constant with Langevin thermostat and barostat, respectively, as implemented in the AMBER12 program 21 . The equilibration simulation protocol was performed as follow: (i) slowly heating the system from 0 to 300K for 20 ps at constant volume, by using harmonic restraints of 80 kcal/mol A 2 for all C α atoms and (ii) pressure equilibration of the whole system during 1 ns at 300K with restrained atoms in (i). (iii) Unconstrained 100 ns molecular dynamics simulation at constant temperature (300K) was performed.
In silico mutant proteins were built by using tLEaP module of AMBER12 package 21 , and underwent the same protocol used for wild type protein. Root Mean Square Deviation (RMSD) was used as structure stability controls. All structures were observed to be stables during the time scale of the simulation ( Figure S1).

Ligand migration free energy profiles
The free energy profile for the CO migration process inside the protein tunnel/cavity system was computed by the Implicit Ligand Sampling (ILS) approach that post-processes, using a probe molecule, an MD simulation performed in the absence of the ligand. This method was thoroughly tested for heme proteins 32 . ILS calculations were performed on a rectangular grid (0.5 Å resolution) that includes the whole simulation box (i.e. protein and the solvent) and the probe used was a CO molecule. Calculations were performed on 5000 frames taken from the last 90 ns of simulation time. The values for grid size, resolution and frame numbers were tested in a previous study 32 . Analysis of the ILS data was performed using an ad-hoc fortran-90 program available upon request 32 . Moreover, ILS has been shown to yield quantitative results for ligand migration processes when compared with more costly free energy methods that treat the ligand explicitly.

CO association kinetic constant measurements
Although CO is not the natural ligand of the hemeproteins, it is widely used as a probe for ligand association studies due to its ease of use. In order to address the molecular determinants controlling ligand migration we performed CO ligand association constant measurements of wild type Mt-trHbN and two mutants: VG8F and VG8W. Kinetic traces for CO binding were measured through the absorption changes at the CO adduct peak position (λ=423 nm; Figure 2). Association of CO is well described by a single exponential decay, whose rate constant (k obs ) depends linearly on CO concentration and the slope can be interpreted as k on CO. A significant k on CO decrease for VG8F (715 ± 27 mM -1 s -1 ), and an even larger decrease for VG8W (48 ± 1 mM -1 s -1 ) was observed in relation to that observed for the wild type protein (4495 ± 357 mM -1 s -1 ) (Figure 3). Table 1 summarizes the measured k on CO values for wild type and mutant Mt-trHbs O and N, and is presented alongside literature data.

Molecular dynamics simulations
Small ligand association in the trHb family is presumably regulated by two main processes: i) ligand migration from solvent bulk to the protein distal site cavity, ii) displacement of water molecules from the distal site cavity [15][16][17]35 . With this in mind, we performed classical MD simulations as they allow us to investigate both processes involved in ligand association. Ligand migration was studied using ILS calculations for the wild type, as well as both VG8F and VG8W mutant proteins. Displacement of retained water molecules in the distal site was considered by performing classical MD simulations and analyzing the solvation structure in each active site.
The wild type Mt-trHbN presents two tunnels available for ligand migration, the LT and the STG8 ( Figure 4A). On the one hand, the LT connects three internal cavities: (trHb : CO) 1 , (trHb : CO) 2 and (trHb : CO) 3 . The STG8, on the other hand, has only the distal site cavity, (trHb : CO) 1 , which is directly connected to the solvent. The VG8F mutant conserves both tunnels, although they are constrained compared to those in the wild type. In the VG8W case, however, the energy profiles suggest a completely blocked STG8 and a LT for which the accessibility to the iron heme is partially reduced.
In order to quantify the contribution of the single G8 mutation we computed free energy profiles for CO migration through both LT and STG8 tunnels ( Figure 4B, 4C). The free energy was set to a value of 0 kcal/mol where CO ligand is fully solvated-at 13 Å and 24 Å from the Fe atom, for STG8 and LT respectively. Wild type Mt-trHbN presents small barriers (~2 kcal mol -1 ) for CO migration from the solvent to the active site cavity (trHb : CO) 1 through both tunnels.
The active site water molecules occupancy was computed for all three systems by performing 200 ns of MD simulations with explicit water molecules. In each case a water molecule was able to access the active site and was stabilized by the iron and the distal site residues ( Figure 5). Specifically, in both wild type and VG8F Mt-trHbN a water molecule was present for approximately 40% of the length of the simulation ( Figure 5A, 5B). The VG8W mutant active site, on the other hand, is occupied by water molecules in 80% of the simulation time, probably due to the hydrogen bonding capacity of W ( Figure 5C).      determined in a reaction mixture containing buffer, NADPH and E. coli FdR, to enable cyclic restoration of heme iron to the oxyferrous state. Figure 6A shows that in the absence of protein (red trace) decay of the • NO signal was monophasic until • NO was exhausted. The decay of NO in the presence of Mt-trHbN (black trace) was biphasic, with an almost linear initial rapid rate in decay, which was used to compare the various Mt-trHbN derivatives, followed by a slower rate in decay. This suggests that • NO is not being turned over in a cyclic manner, but is simply binding available heme. • NO consumption results show that the VG8F and VG8W mutans have a statistically significant reduced • NO binding capacity compared to HbN ( Figure 6B).

Dataset 1. Experimental and theoretical calculations raw data
http://dx.doi.org/10.5256/f1000research.5921.d42091 Detailed legends describing the raw data can be found in the text file provided.

Discussion
CO association kinetic constant measurements as well as MD simulations of Mt-trHbN wild type and site-specific mutants were performed to analyze the role of tunnels and water molecules in the ligand association process. ILS calculations showed that the main tunnels of wild type Mt-trHbN, STG8 and LT, were partially blocked in the VG8F mutant and STG8 was nearly completely blocked in the VG8W mutant. The analysis of water molecules showed that VG8W increases the probablity of the presence of a water molecule in the distal site, which may interfere with the association process. Consistently, the association kinetic constants of CO for both Mt-trHbN mutants showed a decrease of slightly less than one order of magnitude when VG8 is replaced with F and two orders of magnitude when VG8 is replaced with W. Moreover, our data also showed that both mutants have less capacity of • NO binding than wild type Mt-trHbN.
Interestingly, the Mt-trHbN VG8W mutant presents a similar k on CO to the wild type Mt-trHbO, showing that a single residue is responsible for the differential accessibility in these proteins. The results support the idea that STG8 and LT are the main channels for CO migration in the deoxygenated Mt-trHbN, as blocking these tunnels decreases the ability of CO to access the heme pocket. Although in both the mutated Mt-trHbN and wild type Mt-trHbO the STG8 is blocked by WG8, the LT remains open in Mt-trHbN, allowing CO access into the heme cavity, whereas the main tunnel for CO migration in Mt-trHbO is the E7, as was previously described 7 . This fact shows that although the k on CO from mutant trHbN and wild type trHbO members are very similar, the ligand enters through different pathways, evidencing the complexity of mechanisms that regulate the ligand association process in these proteins.

Competing interests
No competing interests were disclosed.

Michael Wilson
School of Biological Sciences, University of Essex, Colchester, UK This is a succinct and generally well written paper reporting experimental data on ligand binding to Mt-trHbN and to two mutants (VG8F anndVG8W)).These latter have been designed to test hypotheses regarding the possible routes by which access is gained to the heme by small neutral gaseous molecules. The experiments and the molecular modelling that supports them, and which provides mechanistic insights, have been carefully performed. The results are of interest to the field as they add to the body of accumulated evidence that proteins, including those with the function of binding small neutral molecules, provide specific, and often dynamic, channels to permit rapid access to the binding site. Furthermore the kinetics of binding are seen to be strongly influenced by single amino acid substitution in the access channels.
Although the results support the general conclusions drawn by the authors some clarification of a number of points would be helpful. These are given below. Figs 2A and B does time appear not to start at t=0? 1.

Why in
In Fig and discussion is it proposed that the water molecule is bound to the iron (common for ferric but not for ferrous iron) or stabilised in that location only by hydrogen binding. It is presumed that for the modelling the iron is in the ferrous state as the authors are discussing CO binding.

2.
Although the authors discuss the decomposition of NO catalysed by Mt-trHbN in the presence of oxygen the assays do not make it clear that this is the reaction under study. No mention of the oxygen concentration is made in the legend to Fig 6. It seems oxygen is present to account for the disappearance of NO. From Fig 6 it is stated that NO binds to the protein but is not degraded (e.g. to nitrate via peroxynitrite). Is this because regeneration of the reduced heme (necessary for oxygen binding) is so slow or is it because as ferric heme is reduced by the NADPH/ferredoxin system NO binds before oxygen and thus no turnover occurs as it is known that the NO-ferrous complex does not react with oxygen? In any case it is not made clear to what chemical step the measured kinetics refer. It would improve the 3. manuscript if the authors clarified these points.

Competing Interests: No competing interests were disclosed.
I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

Why in Figs 2A and B does time appear not to start at t=0?
We thank the reviewer for noticing that specific issue. We decided to discard the first 8ms because they were too noisy.

In Fig and discussion is it proposed that the water molecule is bound to the iron (common for ferric but not for ferrous iron) or stabilised in that location only by hydrogen binding. It is presumed that for the modelling the iron is in the ferrous state as the authors are discussing CO binding.
We agree with the reviewer that the water bounds in general very weak to the ferrous iron. However, there is evidence showing that in a polar distal site (with polar residues), as in the case of truncated hemoglobins, water molecules remain inside stabilized by the polar residues, and thus slow ligand binding to ferrous iron (Olson and Phillips Jr, 1997;Ouellet et al., 2008).

2.
Although the authors discuss the decomposition of NO catalysed by Mt-trHbN in the presence of oxygen the assays do not make it clear that this is the reaction under study. No mention of the oxygen concentration is made in the legend to Fig 6. It seems oxygen is present to account for the disappearance of NO. From Fig 6 it is stated that NO binds to the protein but is not degraded (e.g. to nitrate via peroxynitrite). Is this because regeneration of the reduced heme (necessary for oxygen binding) is so slow or is it because as ferric heme is reduced by the NADPH/ferredoxin system NO binds before oxygen and thus no turnover occurs as it is known that the NO-ferrous complex does not react with oxygen? In any case it is not made clear to what chemical step the measured kinetics refer. It would improve the manuscript if the authors clarified these points.
As suggested by the review we modified the caption of Figure 6, and also we add a sentence.
The legend of the caption can be changed to: repeats. (B) Mean rates of •NO decay in the presence of wild type Mt-trHbN or sitedirected mutants from 3 technical repeats ± S.E.M *P < 0.05, unpaired t-test.
The sentence added is: The chemical step being measured in this assay is the reaction between •NO and the oxyferrous heme; once this reaction has concluded, we assume that the heme is restored from ferric back to ferrous. We are unsure why the reaction is single turnover but it could be due to (a) rapid binding of •NO to the ferrous complex before oxygen can bind, rendering it unable to bind oxygen and initiate the reaction or (b) due to slow reduction of Mt-trHbN by the non-native E. coli FdR.
δ atom, as well as the proximal HisF8, because of its coordination to the heme iron". page 5, second column, line 12 It is not clear what the authors mean when they write that the STG8 "has only the distal site cavity, (trHb : CO) 1 , …", especially if this sentence is coupled with Figure 4A, where (trHb : CO) 1 seems to be connected to STG8 through (trHb : CO) 2 . ○ page 5, Title of Table 1 It is probably better to change "..for wild type and mutants Mt-trHbs O and N" to "..for wild type and mutants of Mt-trHbN and Mt-trHbO" ○ page 6, Figure 4 legend In the legend of Panel C there is no mention of the (trHb : CO) 3 site. molecules at the active site." By writing "These mutations affect both the tunnels accessibility as well as the affinity of distal site water molecules, thus modifying the ligand access to the iron" Introduction: page 3, first column, line 8 The authors might want to include a review on trHbs more recent than that indicated in reference (3). There are several of them published in the last few years.
As suggested by the review we modified the references by more recent ones: Davidge & Dikshit (2013). ○ Introduction: page 3, first column, line 23 The paragraph starting from line 23 is a bit misleading because the authors try to generalize the description of the protein matrix tunnels in trHbs by mixing what happens in trHbNs and trHbOs. This is confusing since it might give the impression that three tunnels co-exist in trHbs, which is not true. In this respect, Figure 1 contributes a lot to make confusion, since it is not clear which trHb protein represents and it seems that it has three co-existing tunnels. It is probably better to keep separate trHbNs and trHbOs, both in the text description and in Figure 1. The authors should describe the tunnel features in trHbN (short and long tunnel) and trHbO (cavities, small E7 residues an possible E7 gating), and show two panels in Figure 1 with depicted the tunnel/cavity systems in Mt-trHbN (panel A) and Mt-trHbO (panel B), possibly using a similar protein orientation and highlighting the role of the G8 residue in the two cases.
As suggested by the review, we modified the sentence to clarify.
"Three internal tunnels were found in the truncated hemoglobin family:" by "Three different internal tunnels have been characterize among the trHb members, in general one or two of these tunnels is found in each protein:" We also change Figure 1 and its caption as suggested.
Caption Figure 1. Schematic representation of the two pathways for ligand migration presented in M. tuberculosis trHbN. The Long Tunnel (LT) and Short Tunnel G8 (STG8) are shown in orange. ○ Introduction: page 3, second column, line 7 The sentence regarding the "internal water molecules" is too generic as it is written now, since it is not clear if the authors refer to globins, to trHbs or to Mt-trHbs. The authors should rephrase the sentence to clarify this issue.
As suggested by the review we clarify the sentence: "It has also been noted that in myoglobin, M. Tuberculosis trHbN as well as in T. fusca trHbO, internal water molecules were observed to block the heme accessibility, thus delaying ligand binding" ○ Introduction: page 3, second column, line 10 ○ The authors should say that the experimental measurements and the MD simulations have been performed only on Mt-trHbN and mutants, and not, for instance, on Mt-trHbO.
As suggested by the review we clarify the sentence by adding explicitly the name of the protein studied "By performing CO association kinetic constant measurements (…) of Mt-trHbN, we addressed molecular mechanisms that control ligand association in M. Tuberculosis truncated hemoglobins" Materials and Methods: page 3, second column, line 40 The purification paragraph seems to refers only to trHbN. What about its mutants? The authors should add a sentence to clarify this issue.
We clarify this by modifying "The trHbN protein was" by "The trHbN protein variants were" ○ Materials and Methods: page 4, first column, line 44 The following sentence is not written fully correctly: "Amino acids protonation states were assumed based on environment of the residue in the crystal structure. All solvent-exposed His were protonated at the N-δ delta atom, as well as HisF8, because of its coordination to the heme iron".
One possibility is to rephrase it as follows: "The protonation state of the amino acids was assumed based on the environment of the residues in the crystal structure. All solvent-exposed His residues were protonated at the Nδ atom, as well as the proximal HisF8, because of its coordination to the heme iron".
We thank the reviewer for the suggestion, the phrase was modified as suggested. ○ page 5, second column, line 12 It is not clear what the authors mean when they write that the STG8 "has only the distal site cavity, (trHb : CO) 1 , …", especially if this sentence is coupled with Figure 4A, where (trHb : CO) 1 seems to be connected to STG8 through (trHb : CO) 2 .
We modified the sentence by adding information: On the one hand, the LT connects three internal cavities: (trHb : CO) 1 , (trHb : CO) 2 and (trHb : CO) 3 . The STG8, on the other hand, connected to both the cavity (trHb : CO) 2 and the solvent.  Table 1 It is probably better to change "..for wild type and mutants Mt-trHbs O and N" to "..

for wild type and mutants of Mt-trHbN and Mt-trHbO"
We thank the reviewer for the suggestion; the phrase was modified as suggested.