Study of pH(Low) Insertion Peptides (pHLIPS) Interaction with Lipid Bilayer of Membrane

The pH-dependent interactions of pHLIPs  (pH (Low) Insertion Peptides) with lipid bilayer of membrane provides an opportunity to study and address fundamental questions of protein folding/insertion into membrane and unfolding/exit, as well as develop novel approach to target acidic diseased tissue such as cancer, ischemic myocardium, infection and others. The main goal of the work presented here is to answer the following questions: What is the molecular mechanism of spontaneous insertion and folding of a peptide in a lipid bilayer of membrane; What is the molecular mechanism of unfolding and exit of a peptide from a lipid bilayer of membrane; How polar cargo attached to a peptide’s inserting end might affect the process of insertion into a lipid bilayer of membrane; How sequence variation will affect a peptide’s interactions with a lipid bilayer of membrane (partitioning into bilayer at neutral and low pH; apparent pK of insertion) with the main goal to identify the best pHLIP variants for imaging and therapy of pathological states such as cancer and others. It has been demonstrated that pHLIP insertion into a membrane is associated with the protonation of Asp/Glu residues, which leads to an increase of hydrophobicity that triggers the folding and insertion of the peptide across a lipid bilayer. The insertion of the pHLIP is unidirectional and it is accompanied by the release of energy.. Therefore, the energy of membrane associated-folding can be used to favor the movement of cellimpermeable polar cargo molecules across the hydrophobic membrane bilayer when they are attached to the inserting end of pHLIP. Both pH-targeting behavior and molecular translocation have been demonstrated in cultured cells and in vivo. Thus, there is an opportunity to develop a novel concept in drug delivery, which is based on the use of a monomeric, pH-sensitive peptide molecular transporter, to deliver agents that are significantly more polar than conventional drugs. Understanding the molecular events that occur when a peptide inserts across a membrane, folds, or exits from it and unfolds provides crucial information for the development of new drug delivery agents, as well as improving our understanding of the first step of membrane-associate protein folding. The promise of exploiting tumor acidosis as a cancer biomarker has not been fully realized in clinical practice, even though the acidity has been a known property since the work of Otto Warburg nearly a century ago. The problem has been to find a practical way to target acidity. pHLIP reversibly folds and inserts across membranes in response to pH changes, and this discovery has led to a novel way to target acidic tissue. Steady state biophysical studies have revealed the molecular mechanism of pHLIP action, which is based on the increase of hydrophobicity of carboxyl groups when they become protonated under mildly acidic conditions, leading to peptide insertion into a membrane. It has been shown that pHLIP can target acidic tissue and selectively translocate polar, cell-impermeable molecules across membranes in response to low extracellular pH. As noted in the Molecular Imaging and Contrast Agent Database (MICAD) at NCBI, a pHLIP labeled with a fluorescent dye, or a PETand SPECTagents ( 64 Cu-DOTA, 18 F, 99 Tc) is a marker for in vivo acidity,. All prior studies in vivo were carried out with the WT-pHLIP sequence and showed that a good contrast and tumor to blood ratio can be achieved only more than 24 hours after pHLIP injection, when it has accumulated in the tumor and largely cleared from the blood. However, for the use of pHLIP-based radioactive imaging agents in the clinic, a more rapid background signal reduction is absolutely essential. We have conducted research in order to address this important need, to tune tumor targeting properties, and to broaden our understanding of the molecular mechanism of pHLIP action. A family of 16 pHLIP variants has been designed based on chemical and physical principles and comprehensive biophysical studies were performed with nonlabeled peptides. We have successfully established a set of design criteria and identified the pHLIP candidates for imaging and therapeutic applications, including lead compounds for PET/SPECT and fluorescence/MR imaging.

The main goal of the work presented here is to answer the following questions: -What is the molecular mechanism of spontaneous insertion and folding of a peptide in a lipid bilayer of membrane; -What is the molecular mechanism of unfolding and exit of a peptide from a lipid bilayer of membrane; -How polar cargo attached to a peptide's inserting end might affect the process of insertion into a lipid bilayer of membrane; How sequence variation will affect a peptide's interactions with a lipid bilayer of membrane (partitioning into bilayer at neutral and low pH; apparent pK of insertion) with the main goal to identify the best pHLIP variants for imaging and therapy of pathological states such as cancer and others.
It has been demonstrated that pHLIP insertion into a membrane is associated with the protonation of Asp/Glu residues, which leads to an increase of hydrophobicity that triggers the folding and insertion of the peptide across a lipid bilayer. The insertion of the pHLIP is unidirectional and it is accompanied by the release of energy.. Therefore, the energy of membrane associated-folding can be used to favor the movement of cell-impermeable polar cargo molecules across the hydrophobic membrane bilayer when they are attached to the inserting end of pHLIP. Both pH-targeting behavior and molecular translocation have been demonstrated in cultured cells and in vivo. Thus, there is an opportunity to develop a novel concept in drug delivery, which is based on the use of a monomeric, pH-sensitive peptide molecular transporter, to deliver agents that are significantly more polar than conventional drugs. Understanding the molecular events that occur when a peptide inserts across a membrane, folds, or exits from it and unfolds provides crucial information for the development of new drug delivery agents, as well as improving our understanding of the first step of membrane-associate protein folding.
The promise of exploiting tumor acidosis as a cancer biomarker has not been fully realized in clinical practice, even though the acidity has been a known property since the work of Otto Warburg nearly a century ago. The problem has been to find a practical way to target acidity. pHLIP reversibly folds and inserts across membranes in response to pH changes, and this discovery has led to a novel way to target acidic tissue. Steady state biophysical studies have revealed the molecular mechanism of pHLIP action, which is based on the increase of hydrophobicity of carboxyl groups when they become protonated under mildly acidic conditions, leading to peptide insertion into a membrane. It has been shown that pHLIP can target acidic tissue and selectively translocate polar, cell-impermeable molecules across membranes in response to low extracellular pH. As noted in the Molecular Imaging and Contrast Agent Database (MICAD) at NCBI, a pHLIP labeled with a fluorescent dye, or a PETand SPECT-agents ( 64 Cu-DOTA, 18 F, 99 Tc) is a marker for in vivo acidity,.
All prior studies in vivo were carried out with the WT-pHLIP sequence and showed that a good contrast and tumor to blood ratio can be achieved only more than 24 hours after pHLIP injection, when it has accumulated in the tumor and largely cleared from the blood. However, for the use of pHLIP-based radioactive imaging agents in the clinic, a more rapid background signal reduction is absolutely essential. We have conducted research in order to address this important need, to tune tumor targeting properties, and to broaden our understanding of the molecular mechanism of pHLIP action. A family of 16

Introduction
The stability and folding of membrane proteins are strongly constrained by the formation of secondary structures in the lipid bilayer environment, driven by the hydrophobic effect and hydrogen bonding. Consideration of these factors has led to versions of a thermodynamic framework model for the folding and unfolding of helical membrane proteins (1)(2)(3)(4)(5). One concept is that spontaneous insertion and folding includes the formation of helical intermediates at the bilayer surface, followed by insertion, and that unfolding includes the same steps, but in reverse order. Because folding to form a helix is coupled to insertion, a significant experimental challenge in testing the concepts is to separate the process of peptide partitioning into a membrane from the folding events leading to secondary structure. We have studied a water-soluble membrane peptide, pHLIP (pH Low Insertion Peptide), that binds to the surface of a lipid bilayer in an unstructured monomeric state at neutral pH (6)(7)(8)(9)(10)(11)(12).
Lowering the pH triggers the spontaneous insertion of the peptide across the bilayer and the formation of a transmembrane helix. Because the pH drop can be accomplished by rapid mixing, kinetic analysis can be used to examine steps in the process, and the predictions of the thermodynamic models can be tested.

Results
pHLIP is a remarkable 35-residue peptide found as a transmembrane helix in bacteriorhodopsin, yet soluble when isolated in aqueous solution. At neutral and high pHs pHLIP is monomeric at concentrations less than 8-10 M, and equilibrates between unstructured forms in aqueous solution (state I) and bound to the surface of a lipid bilayer if one is available (state II) (6,9) (Figure 1a supplementary information).
In an acidic environment the equilibrium is shifted toward a monomeric transmembrane helical form (state III) (6,9), and the process of insertion is accompanied by an energy release of about 1.8 kcal/mol in addition to the binding energy (6-7 kcal/mol) locating the peptide at the surface (11). The pKa of the transition from state II to state III is 6.0 (6,11). The transmembrane orientation of the peptide at low pH was previously established by FTIR spectroscopy (6); here, we additionally confirmed a transmembrane orientation by oriented circular dichroism (OCD) spectroscopy (Figure 1b supplementary information). A characteristic OCD spectrum for transmembrane helix orientation was obtained using direct insertion of pHLIP peptides into supported bilayers (in contrast to earlier experiments, where bilayers were assembled with membrane peptides (13,14). The protonation of two Asp residues in the transmembrane region leads to an increase of hydrophobicity that result in the folding and insertion of the peptide across a membrane (8). Increasing the pH promotes the unfolding and exit of the peptide from the core of the lipid bilayer.
The insertion of pHLIP across a membrane is unidirectional: the C-terminus goes inside a cell or vesicle, and the N-terminus stays outside (7,9). Neither partitioning of an unstructured peptide onto the bilayer surface at neutral pH nor insertion as a transmembrane helix at low pH promotes membrane fusion or leakage of vesicles, red blood cells, or cancer cells (7,8,10). Fluorescence and CD spectroscopy have been used to monitor transitions between the states (6,8,9). The pHLIP peptide gives an opportunity to observe membrane-associated transitions between surface coil and transmembrane helix and vice versa. An enabling advantage of our system is that an initial state is defined, with the peptide bound to the surface of a membrane as an unstructured monomer. Transitions were induced by rapid changes of pH by fast mixing of the aqueous solution of pHLIP pre-incubated with POPC at pH8.0 or pH4.0 with equal volumes of appropriately diluted solutions of HCl or NaOH, respectively. Membrane-associated peptide folding/insertion and unfolding/exit were monitored by the changes of CD and fluorescence signals we previously used to study equilibrium states (Figure 1a-b, d-e). Such parameters (mostly fluorescence) are widely used for the observation of conformational changes that occur in polypeptides interacting with a membrane or detergent (15)(16)(17)(18). For fast fluorescence measurements, a filter was used to capture the emission above 320 nm, but changes of the entire fluorescence spectrum were also recorded in a global mode with use of emission monochromator (Figure 1c, f). The obtained in stopped-flow mode fluorescence spectra clearly show that the increase of fluorescence is accompanied with shift of position of maximum, which indicative to the peptide insertion into hydrophobic core of membrane.
Since the pHLIP peptide contains a Pro residue in the transmembrane part, we speculated that proline might act as a helix breaker and/or exhibit isomerization during folding. To examine these possibilities, a pHLIP variant was synthesized with Pro replaced by Ala. We observed that the variant shows some helical structure in solution at neutral pH ( Figure 2 supplemental information), and that vesicle binding of the variant further promotes the coil-helix transition. Since replacement of Pro led to changes in solution and in the peptide-membrane interactions, we asked whether the presence of prolyl isomerase (cyclophilin A), which promotes Pro isomerization (19), would affect folding and insertion of the non-mutated peptide. Experiments were carried out at various temperatures. No differences were observed in membraneassociated folding in absence and presence of prolyl isomerase (data not shown), so we conclude that the proline is simply acting as a helix breaker in solution and on the bilayer surface. A detailed description of the experimental protocol is presented in the supplementary section.
To avoid crowding of the peptide on the surface of vesicles (parking problem) we set the peptide:POPC molar ratio at ~ 1:140 (11). Given the surface binding energy of 6-7 kcal/mol (from our earlier measurements), the bound to free ratio is ~ 10 4 to 10 5 , so the initial state in the insertion experiments is predominantly an unstructured surface bound state of the peptide (state II, Figure 1 supplementary information). In the unfolding experiments, the initial state, at low pH, is predominantly a transmembrane helical configuration of the peptide (>95% state III).
As seen by fluorescence, the entire process of insertion (transition form state II to state III) upon a rapid drop of pH is found to be well described by a pseudo-first order kinetics model with 4 consecutive steps involving State II, three intermediates, and State III: The rate equations for the time dependence can be written as differential equations: where f i are the changes of fluorescence or CD signals associated with the i-th transition from one intermediate to another, and n is the number of intermediates used in the model. For the kinetic model with three intermediates the function used to fit the experimental data is 4  3  4  2  3  2  4  3  2  4  3  4  1  3  1  4  3  1   4  2  4  1  2  1  4  2  1   4  3  4  2  4  1  3  2  3  1  2  1   3  2  3  1  2  1  3  2  1   1  4  4  3  4  2  1  3  3  2  1  2   5   4  3  4  2  3  2  4  3  4  1  3  1   3  2  3  1  2  1  4  2  4  1  2  1   1  4  4  3  4  2  1  3  3  2  1  2   3  2  1  4   3  2  3  1  2  1   3  2  1  3  2  1  2 The fluorescence kinetic curves measured at various temperatures ( Figure 2a) (6) where k i are the rates presented in Table 2 The activation Gibbs free energy, ΔG i # is then: where h is Planck's constant and k b is Boltzmann's constant. Thermodynamic activation parameters are summarized in Table 2 and their changes during the transition from one state to another are presented in Figure 2c.

Discussion
To interpret the results, we have constructed a model that is based on and consistent with the data, but which is incompletely constrained by it. Thus, some features are viewed with confidence while others are less certain. The model assumes single pathways of insertion and exit, and is shown in figure 3. Starting with the surface bound peptide of State II, and following the drop in pH, we see a rapid formation of helix, in two steps (Folding intermediates 1 and 2, forming at 8 ms and 112 ms), each producing about half of the total helix. The helix might be a single straight helix, or it might consist of several short helices with breaks in between (these possibilities are experimentally indistinguishable). This part of the interpretation is secure: helix formation clearly is faster than most of the fluorescence changes. It is possible that helix rotational orientation or a slight sinking of the helix at the interface could account for the fast (28ms) first fluorescence change of ~23%. Following the rapid helix formation, the rest of the insertion process is reported by fluorescence, and is about 1500 times slower, occurring in three kinetically distinct steps over the next ~45s.
While we cannot be certain of the exact nature of the kinetic intermediates during insertion, we suggest a few thoughts as working ideas. From the surface helix formation to the fully inserted transmembrane helix, two intermediate states are apparent, folding intermediates FI 3 and FI 4 . Some features of the process to consider are that significant rearrangement of lipid must occur during the peptide insertion, that the polar C terminus of pHLIP must traverse the bilayer, and that the proline may provide a point of flexibility. Our previous studies showed that there is substantial lipid perturbation when the peptide is in its surface bound configuration, and that the perturbation is much reduced when pHLIP adopts its transmembrane configuration (8,10,11). It is possible that a large collective lipid reorganization is involved, but we have no data with which to assess such a possibility. We therefore invoke a bending and partial insertion of the helix as an idea, suggested by the proline, and sketch the intermediate shown as FI 3  propagation of the helix-coil transition is accompanied by a rapid exit of part of the polypeptide within the next 5 ms (UI 2 ). The remaining 30% of the membraneembedded helical structure unfolds and exits within 65 ms (including the C terminus), and equilibrium is established between pHLIP in its unfolded membrane-bound (U m ) and soluble forms. Unfolding experiments carried out at 7°C showed that the peptide exit also occurs in two steps with characteristic times of 17.6 and 152 msec (data not shown). An interesting possibility is that the two deprotonation steps happen separately, and account for steps in the exit pathway.
The existence of interfacial folded intermediates reported by several authors has been based on indirect measurements (22)(23)(24), and the results of molecular dynamics simulations of membrane-associated folding and insertion of various peptides are not definitive (25)(26)(27)(28). Here, we present a direct observation of the formation of an interfacial helical intermediate as a step in the process of folding and insertion of a peptide into a membrane.
We conclude that an interfacial helical intermediate is a required step during the process of pHLIP folding/insertion. Helix formation reduces the free-energy penalty associated with the partition of the peptide backbone into the low dielectric environment of the bilayer, despite the fact that the coil-helix transition is associated with a loss of entropy. Helix insertion, most probably, is accompanied by a significant perturbation of lipids, although the lipid perturbation is reduced in the overall process.
The insertion of pHLIP is slow, since the peptide is initially located on the outer leaflet of the bilayer, and it takes time to cross the membrane and reorganize lipids around the transmembrane helix. In contrast to the insertion, unfolding and exit occur much faster, perhaps since the peptide can be conceptualized as occupying a small channel across the lipid, so that the peptide can quickly exit without significant lipid reorganization. Such a channel would close immediately after exit of the peptide. We have demonstrated in vivo that pHLIP can target diseased tissues with elevated levels of extracellular acidity, such as tumors (8,12,29) and that the energy of the insertion events can be used for the selective translocation of polar cell-impermeable cargo molecules across the membranes of liposomes and cells (7,12    The Arrhenius plot was constructed according to the Arrhenius equation (6 Table 1.
See discussion in the text.

pHLIP peptide
The pHLIP sequences: AEQNPIYWARYADWLFTTPLLLLDLALLVDADEGT and its Pro to Ala variant AEQNPIYWARYADWLFTTALLLLDLALLVDADEGT were prepared by solid-phase peptide synthesis using Fmoc

Steady-state fluorescence and CD measurements
Steady-state fluorescence measurements were carried out on a PC1 spectrofluorometer

OCD measurements
For oriented circular dichroism measurements we prepared the supported bilayer on The blank spectrum was subtracted from the sample spectrum to get final line presented in Figure 1b supplementary information.

Stopped-flow fluorescence and CD measurements
The stopped-flow fluorescence and CD measurements were carried out on a SFM-300 Changes of the pHLIP fluorescence signal were recorded through a 320 nm cutoff filter using an excitation wavelength of 280 nm. The fluorescence signal, recorded over 80 sec, was corrected for photobleaching. Changes of the scattered light signal from the liposomes were recorded through the 320 nm cutoff filter using an incident wavelength of 320 nm. Each kinetic curve was recorded several times and then averaged, excluding the first 3-4 shots. The shift of the entire fluorescence spectrum during folding/unfolding was also recorded in a global mode using an emission monochromator, with an excitation wavelength of 275 nm to minimize scattered light at short wavelengths (in a separate experiment the spectra were recorded with 280 nm of the excitation wavelengths). Each spectrum was recorded several times and averaged. All spectra were corrected for the spectral sensitivity of the instrument by comparing the spectrum of a tryptophan solution obtained with the same instrument with a standard tryptophan spectrum. In a control experiment, the signal was measured from the liposomes in the absence of peptide. At an excitation of 275 nm the scattering signal was insignificant, even at short wavelengths of the spectra.
The most challenging measurement was to monitor the changes of the CD signal (at 225 nm), which occurred ~100 times faster than the fluorescence changes. Since replacement of the Pro residue led to alteration of the peptide-membrane interactions, we carried out experiments (folding/insertion) at various temperatures in the presence of prolyl isomerase (cyclophilin A), which catalyses the isomerization of Pro residues in peptides. pHLIP was preincubated with the cyclophilin A (Sigma) at a ratio of 1:100 for two hours before mixing with liposomes, and fluorescence (at various temperatures) and CD kinetics measurements were performed in presence of cyclophilin A. All data obtained were the same as in the absence of enzyme.

Data analysis
The kinetic equations were solved by integration of (1) in Mathematica 5 (Wolfram Research). Nonlinear least squares curve fitting procedures were carried out in Origin 7 and MatLab. The goodness-of-fit was assessed by the aadjusted R-square statistics (adjR 2 ) and root mean squared error (RMSE) according to the standard formula. The transmembrane orientation of the helix has been confirmed by OCD. Transitions between states can be monitored by changes of fluorescence (c) and circular dichroism (CD) (d) spectral signals. The fluorescence and CD spectra of pHLIP at pH8 (black lines) indicate an unstructured configuration with tryptophan residues fully exposed to solvent. Incubation of pHLIP with liposomes at pH8 (blue lines) induces the partial burial of tryptophan residues inside the lipid bilayer without helix formation.
Decreasing the pH to 4.0 by the addition of HCl (red lines) induces the insertion of pHLIP and helix formation

Abstract
We have used pHLIP® [pH (low) insertion peptide] to study the roles of carboxyl groups in transmembrane (TM) peptide insertion. pHLIP binds to the surface of a lipid bilayer as a disordered peptide at neutral pH; when the pH is lowered, it inserts across the membrane to form a TM helix. Peptide insertion is reversed when the pH is raised above the characteristic pK a (6.0). A key event that facilitates membrane insertion is the protonation of aspartic acid (Asp) and/or glutamic acid (Glu) residues, since their negatively charged side chains hinder membrane insertion at neutral pH. In order to gain mechanistic understanding, we studied the membrane insertion and exit of a series of pHLIP variants where the four Asp residues were sequentially mutated to nonacidic residues, including histidine (His). Our results show that the presence of His residues does not prevent the pH-dependent peptide membrane insertion at ∼ pH 4 driven by the protonation of carboxyl groups at the inserting end of the peptide. A further pH drop leads to the protonation of His residues in the TM part of the peptide, which induces peptide exit from the bilayer. We also find that the number of ionizable residues that undergo a change in protonation during membrane insertion correlates with the pH-dependent insertion into the lipid bilayer and exit from the lipid bilayer, and that cooperativity increases with their number. We expect that our understanding will be used to improve the targeting of acidic diseased tissue by pHLIP.

Introduction
Extracellular acidification is a hallmark of different pathologies, including cancer, inflammation, ischemic stroke, and atherosclerotic plaques. Acidosis might be a useful biomarker for diagnosis or treatment if means can be found to target tissue acidity. We have found that a peptide derived from helix C of bacteriorhodopsin, 1  pHLIP is monomeric at low concentrations, with a mostly unstructured conformation in neutral and basic solutions (state I). If lipid vesicles or membranes are present at neutral pH, pHLIP binds to their external surface with an energy of 6-7 kcal/mol (state II). 4 In the membrane-attached state, pHLIP remains largely unstructured. 1 However, if the solution pH is lowered, pHLIP inserts to form a transmembrane (TM) α-helix (state III). The insertion is fully reversible and unidirectional, with the C-terminus being translocated across the membrane. 3 The pK a of peptide insertion into lipid bilayers is 6.0, and the energy difference between the attached state and the inserted state is 1.8 kcal/mol at 37 °C. 4 The pHLIP sequence is relatively rich in acidic residues (Table 1). At neutral pH, the combined negative charges of these residues, together with the carboxy terminus, constitute a large energetic barrier to pHLIP insertion across the membrane. The estimated energetic cost of the transfer of a single aspartic acid residue from water to the hydrophobic core of the membrane is unfavorable by 3.6 kcal/mol for the unprotonated (negatively charged) state, but only by 0.4 kcal/mol for the protonated (noncharged) state. 5 Simultaneously moving four charged Asp residues, one Glu residue, and the carboxy terminus into the membrane would cost 21.6 kcal/mol, assuming 3.6 kcal/mol for each carboxyl group, and peptide partitioning into the membrane at equilibrium would be about 1:10 16 . Thus, for pHLIP to be able to insert into membranes, protonation of a large fraction of the acidic residues can be expected, and knowledge of the protonation pattern of the acidic residues of pHLIP is an essential part of understanding the molecular mechanism of the membrane insertion process for any peptide containing carboxyl groups. Two classes of carboxyl groups are of interest: those that remain buried in the membrane after pHLIP is inserted into the membrane and those that traverse the hydrophobic core of the membrane during insertion. 6 Accordingly, we have studied both the pH-driven membrane insertion and the exit process for a series of peptides where the key aspartic acid residues are sequentially mutated.

Results
Previous studies in our laboratories revealed that sequence variations in the TM region of pHLIP can disrupt the delicate balance that preserves its water solubility. For example, a simultaneous change in the two aspartic acid residues at positions 14 and 25 to the homologous glutamic acid (Asp14/25Glu) resulted in a loss of pH-dependent membrane insertion due to aggregation of the peptide in aqueous solution 7 (we have recently developed new pHLIP variants with several Glu residues, which preserve pHdependent properties; unpublished data). In order to reduce the likelihood that the introduced variations in the peptides used in this work could cause aggregation, we decided to follow a dual strategy to increase their water solubility: (i) we added an Asp tag to the N-terminus (noninserting end) to increase the number of charges in the molecule, which typically improves the solubility of hydrophobic peptides 8 and 9 ; this resulted in the replacement of the N-terminal sequence AAEQ with DDDED (Table 1); and (ii) we used the TANGO algorithm 10 to define the region of the pHLIP sequence with the highest aggregation tendency and found this to be the stretch from residue 21 to residue 30 (coinciding with the most hydrophobic region of the peptide).
We then mutated Leu26 to Gly, which greatly reduced the predicted aggregation tendency.
We incorporated these modifications into a series of pHLIP variants, where four aspartic acid residues were sequentially mutated to nonacidic polar residues. The aspartic acid residues at the C-terminus of the peptide that transitorily traverse the core of the membrane upon insertion (Asp31 and Asp33) were replaced with polar but uncharged asparagine residues. On the other hand, for the Asp residues that are located in the core of the membrane after insertion (in positions 14 and 25), histidine was chosen as the replacement residue, as it is expected to be partially charged at neutral pH (thus improving water solubility) while being only slightly polar in its uncharged state (the transfer energies from water to the bilayer interior are 0.43 and 0.11 kcal/mol for the neutral forms of Asp and His, 5 respectively) so that the insertion properties of pHLIP may not be altered. The peptides were named D0-D3 according to the number of aspartic acid residues present in the regions of interest (TM and Cterminus; the positively charged N-terminus is not expected to interact with the membrane). For the variants with three aspartic acids, two alternatives were studied: one that kept Asp14 (D3a peptide) and the other that kept Asp25 (D3b peptide).
We conducted experiments to test the state of the variants in solution, where pHLIP is largely found as an unstructured monomer. 11 Sedimentation velocity experiments were conducted to determine the oligomerization state of the different peptide variants in aqueous buffer. Previous analysis of wild-type (wt) pHLIP (at 7 μM in 10 mM phosphate buffer and 100 mM NaCl, pH 8) 11 showed that pHLIP is mostly monomeric, but a small oligomer population is observed (∼ 6%). We performed our sedimentation velocity experiments under the same conditions, but without NaCl in the solution. For each peptide, we observed a peak with a sedimentation coefficient of 0.72 ± 0.12 S (Table 2 and Fig. 1), which corresponds to a molecular mass of 3.4 ± 0.8 kDa. This is in agreement with the expected monomer masses of the different peptides (4126 Da for wt and ∼ 4300 Da for the different variants), with the differences being ascribed to shape effects from the extended peptide. In the case of D1 and D0, a minor peak with a sedimentation coefficient of 3.3 ± 0.3 S was also observed. This component represents 5 ± 2% of the total population, and its sedimentation coefficient corresponds to a molecular mass of 43 kDa (roughly consistent with the presence of an octameric or decameric particle). Comparison of our results with the previous report for wt suggests that the presence of oligomers is reduced at lower ionic strength. For the particular case of the D1 and D0 peptides, they seem to have a slightly higher oligomerization tendency in solution, but they are still 95% monomeric. Thus, our results suggest that all the peptide variants remain soluble and are essentially monomeric. For the rest of the experiments, we employed peptide concentrations (1.5-5 μM) lower than that used for sedimentation analysis (7 μM); thus, the level of oligomers present for D1 and D0 is expected to be lower.
Fluorescence spectra of the peptides in aqueous solution at neutral pH showed that, in all cases, the emission maximum is centered around 347-349 nm (Fig. 2, black lines, and Table 2), indicating that the two tryptophan residues of the peptides are largely exposed to aqueous solution, as in fully unfolded proteins, and consistent with the slightly low sedimentation coefficient. This finding represents an improvement over the previously studied Asp14/25Glu mutant peptide, where peptide aggregation shifts the emission maximum to 342 nm in buffer at pH 8. 7 A similar fluorescence maximum was also observed for the Asp14/25Asn mutant under the same conditions. 2 The presence of mostly unstructured species in aqueous solution for each of the studied peptides was confirmed by circular dichroism (CD) experiments, since the observed CD spectra were characterized by a minimum at 203 nm ( Fig. 3, black lines), as observed for pHLIP in state I.
The two lipid-interacting states of the pHLIP variants were then examined: state II, where wt pHLIP is mostly unstructured and attached at the bilayer surface, and state III, where wt pHLIP forms a TM helix at low pH. 1  The D2 variant, where both Asp residues are replaced by His residues, also demonstrates a pH-dependent membrane interaction. However, the spectral pattern is slightly different from those for wt and D3 variants: the fluorescence intensity of D2 in the presence of POPC decreases in the pH range 8-6, with no significant changes in the spectral maximum at pH 8-7 and with a small shift to lower wavelengths at pH 6 ( Fig. S1). The amount of the helical structure of D2 at neutral pH is slightly higher than those of wt and D3 ( Fig. 2 and Table 2), while no change is seen in the pH range 8-6. As an explanation, we suggest that D2 partitions somewhat more deeply into the membrane lipid bilayer than wt and D3 at neutral pH values, since His residues are expected to be only partially charged at neutral pH values, enhancing the hydrophobicity of the peptide TM and its affinity for the lipid bilayer. The decrease in fluorescence signal in the pH range 8-6 might be attributed to the partial quenching of emission of at least one of the Trp residues by one of the partially protonated His residues. At the same time, at neutral pH values, the peptide C-terminus containing four negative charges (two Asp, one Glu, and the C-terminus) does not partition into the membrane, keeping the peptide at the membrane surface. A further drop of the pH to pH 3-4 is associated with a fluorescence spectral maximum blueshift, an increase in fluorescence intensity (Fig. 2), and the appearance of a more pronounced negative band at 222 nm on CD spectra (Fig. 3), which is usually an indication of peptide insertion into the bilayer. 1 Reduction of pH leads to the protonation of negatively charged groups at the C-terminus and peptide insertion into the membrane. At the same time, we expect that protonation of His residues at low pH should occur; this might lead to the peptide's exit from the lipid bilayer or, alternatively, the formation of a pore channel in the lipid bilayer, where positively charged His residues would be pointed toward the channel. Calcein encapsulation control experiments that rule out the formation of pores in the membrane by the D2 and D3 peptides were performed ( Fig. S2). Thus, most probably, the pK a for the protonation of His is shifted to very low pH values when it is embedded in a lipid bilayer. We carried out fluorescence pH titrations to compare the behaviors of D2 and wt peptides at pH values lower than 3.5 ( Fig. S3). While no fluorescence change was detected for wt at acidic pH values, we observed that an additional process was present for D2 (with an apparent pK a of 2.5), characterized by a fluorescence decrease and a redshift of the spectral maximum, which might be associated with peptide exit from the lipid bilayer.
To establish the orientation of each helix in the membrane, we performed oriented circular dichroism (OCD) measurements in which the light beam is oriented perpendicular to the planes of a stack of oriented lipid bilayers containing the peptides of interest. Theoretical calculations and experimental data indicate that helices oriented with axes parallel with the membrane surface (perpendicular to the incident light) give CD signals distinctly different from those of helices oriented across the bilayer (parallel with the incident light). 12, 13 and 14 In the range of 190-240 nm, the peptide CD spectrum is dominated by π-π* and n-π* transitions. 15 The π-π* transition in a helix splits into three components, one of which gives rise to a negative Gaussian band near 205 nm, with its electric transition dipole parallel with the helical axis.
When the incident light propagates parallel with the helical axis, the electric field vector is orthogonal to the 205-nm π-π* dipole transition, and there is no interaction between the electromagnetic wave and the dipole, leading to the disappearance of the negative band at 205 nm in a CD spectrum. Thus, when the supported bilayers are oriented perpendicular to the light propagation, a helix with a TM orientation will have a CD spectrum that contains a positive 190-nm band and a negative 225-nm band. If the helix adopts a membrane surface orientation on the supported bilayer, then all transitions are seen, and the OCD spectrum is the same as for a peptide CD spectrum in solution, with randomly oriented helices. Our data clearly indicate that D2 adopts a TM orientation at pH 3.5-4.5, while increasing the pH leads to peptide exit and the appearance of a membrane surface orientation of the helix (Fig. 4). The OCD spectrum at pH 1.9 does not correspond to a TM helix. Thus, we conclude that the pK a of both or at least one of the His residues is significantly shifted from 6.3-6.9 16  How does the number of carboxyl groups affect the pK and cooperativity of insertion?
We monitored the pH-induced changes in the position of the fluorescence emission maximum of the peptides, which provide details about peptide insertion into the lipid bilayer, in the presence of POPC (Fig. 6). A plot of the positions of the spectral maxima follows a sigmoid behavior as a function of pH, corresponding to the transition between the interfacial state and the inserted state for all variants (except for D0). Fitting the experimental data provides the two main parameters that describe the insertion process: pK a and cooperativity (m). The pK a of membrane insertion obtained for wt pHLIP is 5.94 ± 0.09, which is in agreement with previous reports. 1 and 7 For the different variants, shifts of the pK a to lower values (∼ 5.2) were detected (Fig. 7a).
The reason for this decrease is unclear, but it might be related to the lower number of aspartic residues or to the presence of histidines in the TM region of the pHLIP variants. We do not think that the N-terminal DDDED sequence will influence the pK a values of the peptides in our study, since its polarity should preclude hydrophobic interaction with the lipid bilayer; thus, it is not expected to be involved in the insertion process. However, we cannot rule out that it might reduce the overall membrane affinity of the peptide. While the pK a values for the variants changed very little, we observed a gradual decrease in the cooperativity of the insertion process (m parameter) for peptides with fewer Asp residues, as the titration occurred progressively over a wider pH range (∼ 1 pH unit for wt and ∼ 2 pH units for D1) (Figs. 6 and 7b). Our data indicate that the cooperativity of insertion is linked to the number of protonatable residues. Cooperativity and pK a might also respond to the position of protonatable groups in the peptide sequences and their proximity to each other. When pHLIP is at the surface of the vesicle and the pH is lowered, the protonation of one Asp residue might facilitate the protonation of other protonatable residues, shifting their pK a values. The protonation of the first Asp residue might induce partial insertion of the peptide into the membrane. In this scenario, the protonation of the neighboring Asp residues would be energetically favored to shield the negative charge (i.e., the pK a value of the neighboring Asp is shifted to higher values in a more hydrophobic environment) and then a positive feedback would be established, triggering membrane insertion.
How do the number and the location of Asp residues affect peptide exit from the membrane? The CD and fluorescence changes associated with wt pHLIP lipid insertion at acidic pH are completely reversible. 11 Here we also followed changes in the CD and fluorescence signals and in the reversibility of biotin translocation across the membrane. The ellipticity increase associated with each peptide insertion into the membrane was found to be essentially reversible for wt and D3b (Fig. 3, broken blue lines overlap with continuous blue lines), while for D3a, D2, and D1, the reversibility was only partial. Since changes in the CD signal upon acidification for D2-D0 are less pronounced than those for wt and D3, the reversibility of the D2-D0 membrane insertion was also assessed by changes in the fluorescence signal (Fig. S4). It is interesting to note the different levels of reversibility of the two D3 peptides: the insertion process is significantly more reversible in D3b (90%) than in D3a (70%) (Fig. 5b), suggesting nonequivalence of the two buried positions. We observed an overall linear relationship between the number of aspartic acid residues interacting with the membrane and the degree of α-helix formation reversibility (Fig. 5b). The results obtained for the reversibility of the biotin translocation (exit process) were also in agreement (Fig. 5b).
An important consideration in the interpretation of the exit data is the time course of equilibration of the pH inside the liposomes, so we encapsulated the membrane-

Discussion
We have previously observed that even conservative changes in the pHLIP sequence can lead to peptide aggregation in solution at neutral pH. 7 Our results show that all the peptides in this study are soluble in solution, being essentially monomeric (the addition of a D-tag at the N-terminus and the L26G mutation appear to favor peptide solubility). Spectral data obtained with D3-D0 peptides indicate that the lower is the number of negatively charged groups in the peptide sequence, the deeper are the peptide partitions into a lipid bilayer and the greater is the helicity. At the same time, TM orientation (at least for the D3-D2 peptides) requires protonation of the Asp/Glu residues and the terminal carboxyl group at the C-terminus, which can readily go across a membrane in its noncharged form. We confirmed our previous finding 2 suggesting that TM Asp residues are not essential for peptide insertion. Interestingly, we have observed here that membrane insertion upon acidification occurs in our peptides in the presence of two His residues in the predicted TM region. Histidines have been used in the past to drive the insertion of peptides into membranes at neutral pH values. 17  We conclude that protonation of negatively charged residues located in the TM or in the C-terminal inserting end must occur in order to preserve the pH-dependent ability of pHLIP to interact with the membrane. These residues act as switches for pHLIP membrane insertion, as the negative charges of their side chains block membrane insertion. Acidification causes the protonation of these side chains, resulting in an increase in the overall hydrophobicity of the peptide, which leads to TM helix formation, shielding the hydrophobic residues of pHLIP from water molecules. When the pH is raised to near neutrality, the negatively charged state of the carboxyl groups is again favored, decreasing the peptide hydrophobicity and resulting in exit from the TM position. Peptide exit from the lipid bilayer is completed when deprotonation of Asp/Glu residues located in the hydrophobic core of the membrane occurs and the TM helix is destabilized. Thus, pHLIP variants where Asp14/Asp25 were replaced by Glu, with a higher pK a (pK a = 6.5), 7 might be more effective for targeting tumors with higher pH e values. Our present results suggest that the number of Asp residues in the TM region can also modulate the pK a value. Thus, a peptide containing an extra Asp in the TM region might have a higher pK a and might be directed to tumors more effectively. Another important factor to be considered is the broadness of the pH transition of the peptide, which is dictated by the cooperativity of the transition. On one hand, for the case where the peptide pK a is lower than the tumor pH e but the transition is broad (m value is low), a significant part of the pH transition could intercept the pH e value, resulting in a significant pHLIP tumor insertion. However, such a scenario will also lead to more accumulation in healthy tissue. Since it is usually desirable to have a high tumor/organ ratio, an insertion transition of high cooperativity might be best. This would ensure greater differentiation between the amount of inserted peptides and the amount of noninserted peptides over a narrow range of pH values, favoring selective tumor targeting, since the difference in pH between normal tissue and cancerous tissue may be only 0.5-0.7 units. However, we must bear in mind that the measured pH e provides an indication of the average acidity outside the cell for a given tumor and can vary between different tumor regions. Furthermore, pH e may not reflect the precise pH on the exterior surface of the cells, since the cells pump protons to the extracellular medium and ΔpH will lead to proton accumulation at the membrane surface. 25 Another feature that is expected to shift the equilibrium toward the membrane-inserted form is the presence of Asp/Glu residues at the C-terminus of the peptide. After being translocated across the plasma membrane into the cytoplasm, where the pH is neutral, these groups would be deprotonated. Since the translocation of charges across membranes is unfavorable, the inserted form would be stabilized.
pHLIP shows promise as a means of targeting cells in acidic tissues and delivering agents for therapy and imaging. At the same time, we are learning more about the binding and insertion of peptides at the membrane surface. Here we have shown that variation in the positions and numbers of carboxyl group titrations modulates the pK and cooperativity of insertion.

Peptide synthesis and assessment of monomeric state
Peptides were made by solid-phase synthesis, using standard 9- The peptides described in Table 1 Equation (1) where F a = (f A + S A pH) and F b = (f B + S B pH); f A and f B are the spectral maxima for the acidic and basic forms, respectively; S A and S B are the slopes of the acidic and basic baselines, respectively; and m is the cooperativity parameter. Fitting by nonlinear least squares analysis was carried out with Origin software.

Circular dichroism
Samples were prepared as in the fluorescence experiments, but the final molar lipid/peptide ratio was 300:1, with the final peptide concentration varying from 2 to 5 μM. CD spectra were recorded in Jasco J-810 and MOS450 Biologic spectropolarimeters interfaced with a Peltier system. Spectra were recorded at 25 °C using 2-or 5-mm cuvettes, the scan rate was 50 nm/min, and 10-30 averaging steps were performed. Raw data were converted into mean residue ellipticity according to 30 : where Θ is the measured ellipticity, l is the path length of the cell, c is the protein concentration, and N is the number of amino acids.
For the study of membrane attachment, insertion, and its reversibility, the typical procedure was as follows: The samples were incubated with POPC vesicles at pH 8 for 90 min, the spectra were recorded, the pH was lowered to 4.0, and the measurements were performed after 30 min. Finally, the pH of the sample was biotin at its C-terminus was taken as 100% reversibility, and that of pHLIP labeled at its N-terminus was taken as 0%.
Peptides were labeled at the C-terminal Cys residues using the membrane-  Tables   Table 1. Sequence of the peptides.           kcal/mol (12). We previously showed that pHLIP insertion is associated with the protonation of Asp/Glu residues, which leads to an increase of hydrophobicity that triggers the folding and insertion of the peptide across a lipid bilayer (13,14). The insertion of the pHLIP is unidirectional: the C-terminus crosses the lipid bilayer, and the N-terminus stays outside (11,15). The energy of membrane associated-folding can be used to favor the movement of cell-impermeable polar cargo molecules across the hydrophobic membrane bilayer when they are attached to the inserting end of pHLIP (15)(16)(17). Both pH-targeting behavior and molecular translocation have been demonstrated in cultured cells and in vivo (16)(17)(18)(19)(20)(21). Thus, there is an opportunity to develop a novel concept in drug delivery, which is based on the use of a monomeric, pH-sensitive peptide molecular transporter to deliver agents that are significantly more polar than conventional drugs.
In our initial kinetic study we found that pHLIP inserts into a POPC phospholipid bilayer in several steps: first, an interfacial helix is rapidly formed (~100 ms), which is then followed by a slow transmembrane insertion pathway (~1 minute) that contains several intermediates. The exit of the peptide from the bilayer core proceeds ~800 times faster and through different intermediates (22). It remained unclear why it takes 1000 times longer for the helix to insert across a bilayer after it is formed on the surface, and what the intermediates are on the insertion/exit pathways. Another question we wanted to address is how polar cargo might affect the process of insertion, and thus, translocation of cargo across the bilayer.

Materials and methods
Due to the limit of space, a detailed description of experiments could be found "in the Supporting Material".

Design of pHLIP variants
Our previous data indicated that the pHLIP peptide forms a helix, after a pH drop, 1000 times faster than it inserts into a lipid bilayer, and insertion occurs through several steps. The insertion time and nature of the intermediates might result from the presence of four protonatable groups at the C-terminus of the peptide, which have to cross the membrane to complete the process of insertion. In order to cross the energy barrier presented by the hydrophobic membrane core at any appreciable rate, it is reasonable to hypothesize that these charged groups should be at least partially protonated. To test the idea, we asked if the number of protonatable groups at the Cterminus would correlate with the rates of insertion and exit, as well as examining the number of intermediate states along the insertion/exit pathways. Two truncated pHLIP variants were designed: pHLIP-2 and pHLIP-1, where the number of protonatable groups (shown below in red) was reduced to two and one, respectively. Additional Asp residues were placed at the N-terminus to preserve peptide solubility.  (16). Information about Log P measurements can be found in the supporting material. We have shown previously that the pHLIP-4 is capable of translocating biotin-Peg (23). All constructs were purified to remove unreacted peptide and cargo.

Steady-state study of pHLIP variants
We employed fluorescence and CD/OCD spectroscopic techniques to demonstrate that pHLIP variants and their cargo conjugates preserve pH dependent membrane-inserting properties ( Fig. S1-4). The data clearly indicate that all peptides preserve ability to interact with lipid bilayer of membrane in a pH-dependent manner. The data also suggest that the peptides may be partly pulled away from the membrane core by the polar cargo molecules attached to their C-termini (especially pHLIP-2 and pHLIP-2E, which are more hydrophobic and partition more deeply into the membrane and have higher helix content at pH8 compared to pHLIP-4) ( Fig. S3 and Table S1). Because we had moved protonatable residues from the C to the N-terminus and attached polar (non-charged) cargo, we checked for effects on the pH-dependencies of the insertion (Fig. 1c, f and 4S). The perturbation of the insertion pKa by truncation of the Cterminus and attachment of the polar cargoes is small.

Mathematical models for fitting of kinetics data
In our earlier kinetic studies, we used a sequential mathematical model to fit the kinetic data and to find the rates and contributions of individual components (22). To simplify the mathematical model, only forward reactions were taken into consideration. In the present work we have made an attempt to describe the processes by taking into account both forward and backward reactions. We have considered several linear models: two-state (no intermediates): three-state (single intermediate): and four-state (two intermediates) models: The transitions between states are described by a set of differential equations (Appendix S1-3), which can be solved, but the functions obtained are very complex and will contain a number of variable parameters increasing with the complexity of the applied model. It is not practical to perform fitting of the experimental data using such complex functions: a slight variation in input data dramatically affects the solution, thus making it unreliable. However, the solution for fluorescence varation with time can be presented in a general form as a sum of the exponential functions: for the (IV) model (3) where τ i are the characteristic times for each transition or ν i = 1/τ i are the characteristic rates of the transitions, and f i are the characteristic contributions. Thus fitting of the measured curves can be performed using exponential functions and the characteristic rate constants ν i can be found directly from the experiment. However, we wish to emphasize that the characteristic rates (or times) and contributions need to be related to the rate constants (k i ) and contributions for the transition from one intermediate to another, which fully reflect the transitions. Due to the complexity of the problem, we proposed to establish relations only between the characteristic rates and the real rate constants without considering the contributions. By making a few reasonable assumptions, simple approximate relations between k and ν can be established. Thus, for the two-state model (see Appendix S1): for the three-state model (see Appendix S2): , and for the four-state model (see Appendix S3): , , The experimental kinetic data were fitted by the single, or double, or triple exponential functions (eqs. 1-3), which are general solutions for the two-, three-or four-state models, respectively.

Kinetics of insertion of pHLIP variants with truncated C-terminus at various temperatures
The insertion of each of the pHLIP-4, -2, -1 peptides into a lipid bilayer was triggered by a drop of pH from 8 to 3.6 and was monitored at various temperatures (7, 11, 18, 25ºC) by the increase of tryptophan fluorescence (Fig. 1 a-c). The pHLIP-4 peptide inserts across the bilayer as a helix within 30-50 sec (at various temperatures), the pHLIP-2 -within 3-8 sec, and for the pHLIP-1 the process is completed in 80-400 ms, which is about the same as the time of helix formation (90-100 ms) ( Table 1). Thus the processes of helix formation and insertion occur practically simultaneously in the absence of protonatable side chains at the C terminal tail, and the number of protonatable residues at the inserting end does not affect the formation of helical structure, but correlates with the time of peptide insertion into the lipid bilayer.
To ensure that the addition of Asp residues to the N-terminus of truncated pHLIP variants does not alter the kinetics, we tested the following sequence:

pHLIP-6: AEDQNDPIYWARYADWLFTTPLLLLDLALLVDADEGTCOOH
where an additional two Asp residues (total 6 Asp) were placed at the N-terminus of pHLIP-4. The steady-state and kinetics data obtained for pHLIP-6 were very similar to the data obtained with pHLIP-4 ( Fig. S5), which confirms our suggestion that the modification of the N-terminus does not interfere with the process of insertion.

Activation energies of pHLIP variants insertion into bilayer
Arrhenius plots were constructed to establish activation energies (E a ) and frequency factors (A) for the transitions between the states for each pHLIP variant (Fig. 1d). The points were fitted by the Arrhenius equation: A global fit was used in the analysis of the second transition for pHLIP-2 and -1, and for the second and third transitions for pHLIP-4, since slopes of the corresponding curves are very similar to each other (established by separate fitting of each data set).
In the global fitting we used activation energy as a shared parameter to establish differences in the frequency factors for various transitions. The activation energy barrier for the pHLIP-1 and -2 is 13.2 kcal/mol. The frequency factor for the pHLIP-1 transition to the final state is 4.2x10 10 , which is an order of magnitude higher than the frequency factor for the pHLIP-2 (1.

Kinetics of insertion of pHLIP variants with cargoes
The  (Table S1). The cargo "pulls" the peptides to the membrane surface, affecting their state II positions.
Thus pHLIP-2 and pHLIP-2E with cargoes start their journey into the membrane to adopt TM configuration from a more superficial membrane surface configuration than peptides with no cargo, which are more membrane-embedded at high and neutral pHs.  S6). When the size of the pH jump is reduced, both peptide folding and bilayer insertion slow down. The first (fast) rate of the insertion is very similar for all pHLIP variants (Table 3) We observed an interesting behavior of pHLIP-4 when the pH was dropped from 8 to 6, and a "kink" in the fluorescence and CD kinetic curves appeared (Fig. 3). The kinetic curves of the insertion and folding at the pH8→6 jump were fitted by threeexponential functions with negative amplitudes for the second component (shown in red in Table 3). The physical meaning of a negative amplitude is that the spectral signal changes in the opposite direction. These changes indicate that after the pH is dropped, pHLIP-4 first partitions into the lipid bilayer with helical structure formation, but later comes out of the membrane with a reduction of helical content, and then finally "dives" into the membrane slowly with an increase of helical content.

Exit/unfolding of pHLIP variants
We also investigated the reverse processes of exit/unfolding of the pHLIP variants when the pH is changed from 3.6 to 5, 6 and 8 (Fig. S7, Table 4

Insertion/folding and exit/unfolding of single-Trp pHLIP variants
Wishing to better understand the intermediates, we used tryptophan residues positioned along the sequence to follow insertion and exit of different parts of pHLIP-4 into and out of a lipid bilayer (Fig. 4, Table 5). The characteristic times of the transitions for the single-Trp variants are similar to those of pHLIP-4, while double the time is required for the pHLIP-W2 and -W3 to insert and adopt its final TM configuration when the pH is dropped from 8 to 3.6. For the pH8→6 transition, a "kink" is observed similar to that of pHLIP-4 within the same time scale of 4-7 sec.
The most pronounced kink is observed for the pHLIP-W3, and a less pronounced kink is seen for pHLIP-W1 and -W2. As mentioned above, the kink is interpreted as a partial exit and unfolding of the pHLIP-4 peptide in the path to the inserted and folded state when the pH is dropped from 8 to 6. Based on this view, we infer that the Cterminal end of the peptide, which has four protonatable groups, tends to exit the bilayer to a greater extent than other parts of the peptide.
Exit and unfolding for the pH3.6→8 transition happens fast for all single-Trp variants (within 350 ms), but much more slowly for the intermediate transition driven by the pH3.6→6 jump. (Fig. 4 c-d). Interesting changes were observed for pHLIP-W3 with a pH increase from 3.6 to 6: while the fluorescence decays progressively for pHLIP-W1 and -W2, pHLIP-W3 exhibits an initial increase of fluorescence, which then decays slowly. Our interpretation is that the changes are related to the movement of Trp residues across the bilayer as the pHLIP-W3 peptide exits from the membrane.

Rates of pH equilibration in POPC liposomes
It is known that the pH inside a liposome equilibrates progressively with the external pH after a pH jump (24). However, the rate of equilibration depends on the magnitude of the pH changes, the concentrations of other ions present, the charges on the lipid headgroups, the buffering capacities inside and outside of the vesicles, and other factors. One of the widely used methods to follow changes is to monitor fluorescence changes of the pH-sensitive dye fluorescein (FITC) encapsulated into the liposomes.
FITC carries 2 negative charges at pH9 that are protonated as the pH is lowered. Since some of the charged residues of pHLIP peptides are located near the bilayer surfaces (on inner or outer leaflets), we chose to use lipid bound FITC to probe pH changes near the inner leaflet rather than bulk pH changes. We followed pH equilibration after the addition of acid or base using liposomes containing 1% FITC labeled phospholipids. Biphasic kinetics were seen for a pH jump from pH8→3.6 with characteristic rates of about 0.04 and 0.003 sec -1 (Fig. S8a), the data are in a good agreement with the rates measured previously (25). Thus the fastest component of the pH changes inside a liposome is of the same order of magnitude as the slowest component of pHLIP-4 insertion, while pHLIP-2, -2E and -1 peptides fold and insert into the lipid bilayer much faster than the pH equilibrates on the bilayer inner leaflet.
In the case of a pH jump from pH8→6, the first component slows down to 110 sec (0.009 sec -1 ) and the second component is not detectable within 20 min.
We also measured the FITC fluorescence changes when the solution pH is raised from 3.6 to pH 8 and to pH 6 by addition of NaOH (the solution already contained H + and Clions to mimic our unfolding experiments). In both cases, the characteristic time of the first increase of FITC fluorescence associated with the pH changes on the inner leaflet of the bilayer is about 20 sec (Fig. S8 c-d), after which it takes tens of minutes for the pH to be fully equilibrated.

Discussion
In this study we designed several pHLIP variants and examined how elements of the pHLIP peptide and polar cargoes attached to the inserting end determine the pathways and kinetics of peptide insertion across and exit from a lipid bilayer. Based on our results, we have developed a model that describes our current view of the polypeptide membrane entry and exit pathways, as well as cargo translocation across the bilayer (Fig. 5). The model assumes a sequential pathway for the processes of insertion and exit, and takes the state II as a starting point, where the peptide is bound to the surface of the lipid bilayer in predominantly unstructured configuration. A drop of pH leads initially to the protonation (or partial protonation) of the carboxyl groups located in the TM part of the peptide, which are positioned closer to the hydrophobic core of the bilayer and, most probably, have the highest values in the sequences for the pK a of protonation, since the other titratable groups are not as constrained by nearby side chain hydrophobicity to lie near the interface. It is known that the pK a of the protonation/deprotonation of residues depends on the dielectric properties of their environment (26,27), and it was shown previously that buried Asp residues in the Chelix of bacteriorhodopsin, from which pHLIP is derived, have higher pK a 's of protonation than those exposed to a more polar aqueous environment (28). Protonation

Synthesis of peptides
All variants were prepared by solid-phase peptide synthesis using Fmoc

Conjugation of biotin and biotin-Peg to the pHLIPs
For conjugation with biotin and biotinPeg, pHLIP peptides were mixed with biotinmaleimide or biotin-dPeg 3 -maleimide (Quanta Biodesign Ltd) in DMSO at a ratio of 1:10 and incubated at room temperature for about 8 hrs and then at 4ºC until the conjugation reaction was completed. The reaction progress was monitored by HPLC.
The product was purified using reverse phase C18 HPLC, lyophilized and characterized by SELDI-TOF mass spectrometry.

Liposome preparation
Large unilamellar vesicles were prepared by extrusion. 25  Bilayers with the peptide solution were allowed an additional 6 hour equilibration.
Measurements were taken at 3 steps during the process: when the monolayers were incubated with the excess of liposomes, soon after spaces between slides were filled with the peptide solution and 6 hours after the second measurement. 14 slides (28 bilayers) were assembled and OCD spectra were recorded on a MOS-450 spectrometer with 2 s sampling time. Control measurements were carried out of the peptide between slides with and without supported bilayers and in the presence of an excess of POPC liposomes.

Octanol-water partition coefficient measurements
The polarity of biotin-maleimide and Peg-biotin-maleimide was determined by the assessment of the relative partitioning between aqueous and octanol liquid phases. The biotin and biotin-Peg was dissolved in a 10 mM phosphate buffer pH8 (0.5 ml) followed by the addition of n-octanol (0.5 ml). The solutions were mixed by vortexing for 24 hrs at room temperature and then left for another 48 hrs in order to reach equilibrium. After the phase separation, the absorption maximum of biotin at 300 nm was recorded for each phase. The molar extinction coefficients in n-octanol and phosphate buffer are assumed to be the same, and the ratio of the OD readings was used directly to calculate the partition coefficients, P = ODn-octanol/ODwater, and Log 10 P values, which reflect the relative polarity of the cargoes.

Probing pH changes on the lipid bilayer inner leaflet by FITC fluorescence changes
To probe the kinetics of pH changes on the inner leaflet of vesicle lipid bilayers when the external pH is suddenly changed, we used POPC liposomes containing 1 mol% of FITC-DHPE. FITC-DHPE is a pH-sensitive fluorescent dye conjugated with lipid headgroups, and the dye absorbance and fluorescence decreases with a decrease of the pH from 9 to 4. The pH83. solution. An F-test was used to compare fits of different models to select the best one.  Table S1.

Two-State model
The two-state model is used to describe fast processes of folding of the pHLIP-2E variant, kinetic curves of which are fitted well by the single-exponential function. This model doesn't assume existence of intermediate states. (1.1) The transition from the state A to B is described by the differential equation: The variables A and B designate relative populations of the corresponding states. k 1 and kˉ1 are the rates constant for the forward and backward reactions, respectively.
We assume that initially all pHLIP molecules are in the state A and hence the initial conditions are: Exact solution of the differential equation 1.2 is the single-exponential function: Some of the CD kinetic data were fitted by the single-exponential function: where the characteristc rate ν 1 expressed in a form of the rate constants is: If we assume that equilibrium between the states A and B is strongly shifted to the right, meaning that k 1 >> kˉ1 and the difference between the rate constants at least an order of magnitude: (1.8) then we can estimate the rate of the forward reaction from the characteristic rate obtained in result of fitting of experimental data by single-exponential function:

Three-State model
In the majority of cases it was not possible to get an adequate fitting of the experimental data by the single-exponential function. Therefore we introduced a threestate model, which assumes the existence of a single intermediate: 1) The transitions from one state to another are described by the differential equations: The variables A, B and C designate relative populations of the corresponding states.
We assume that initially all the pHLIP molecules are in the state A and hence the initial conditions are: Finally the equilibrium will be reached and the equilibrium populations can be easily To obtain the time evolution of all states one can exclude B and C from the system (2.2-2.4) and obtain the differential equation for A: In general form the solution of the equation 2.7 is given by the two-exponential function: where the characteristic rates v 1 and v 2 are expressed in the form of rate constants: The population of the state B is found from the equation 2.2: and finally the population of the state C is given: and C by S A , S B , S C . Then the spectral signal of the whole system is: Substituting here the expressions for the populations of the different states using equations 2.8; 2.11 and 2.12 one can obtain: (2.21) Experimentally it was found that most of the pHLIP-1 and -2 kinetic curves could be adequately fitted by the two-exponential function: Therefore the experimental measurements S exp provide five parameters: two characteristic rate constants v 1 and v 2 and three characteristic fluorescence amplitudes g 0 , g 1 and g 2 . Comparing S theor and S exp we can find the relationships between the theoretical and experimental parameters: And the rates are given by the equations 2.9 and 2.10. Unfortunately, the theoretical description involves seven parameters: four rate constants k 1 , kˉ1, k 2 , and kˉ2, and three fluorescence/CD amplitudes S A , S B , S C against five experimental parameters, which make it impossible to find the parameters unless we would make assumptions. First, we concentrate our attention only on the rate constants. Second, we noticed that v 1 >> v 2 , thus the equations 2.9 and 2.10 can be expanded into a series. The major terms in this expansion are: If we assume that the equilibrium between the states A, B and C is strongly shifted to the right, meaning that k 1 >> kˉ1 and k 2 >> kˉ2 , and the difference between the rate constants is at least an order of magnitude: , (2.28) then the rate of the forward reaction could be estimated from the characteristic rate obtained as the result of the fitting of the experimental data by the single-exponential function: (2.29)

Four-state model
The adequate fitting of the pHLIP-2E-bt and -btPeg kinetic data was achieved only by the three-exponential function. Therefore we introduced four-state model, which assumes existence of two intermediates: The transitions in this system are described by the set of equations: The variables A, B, C and D designate relative populations of the corresponding states.
We assume that initially all pHLIP molecules are in the state A and hence the initial conditions are: Finally the equilibrium will be reached and the equilibrium populations can be easily found by the graph technique: Solution of these equations is given by the three-exponential functions with the characteristic rates v 1 , v 2 , v 3 and it is rather cumbersome. We can assume that the first transition is very fast and the equilibrium is strongly shifted toward the state B, which means kˉ1  0. Then To solve this set one can exclude D: and then exclude C: Solution of this differential equation is given by with similar expressions for C and D. The first characteristic rate v 1 is given by the equation 3.8, and v 2 and v 3 are determined by: If we assume that the rates of consequent stages significantly decrease, i.e. k 2, kˉ2 >> k 3, kˉ3, then one can expand expression 3.16 into series and find solution in a simple form: We can reasonably assume that the equilibrium (3.1) between the states B, C and D is strongly shifted to the right, meaning that k 2, >> kˉ2 and k 3 >> kˉ3. The difference should be at least an order of magnitude: (3.19) and the rate constants are: (3.20)

Introduction
The promise of exploiting tumor acidosis as a cancer biomarker has not been fully realized in clinical practice, even though the acidity has been a known property since the work of Otto Warburg nearly a century ago. The problem has been to find a practical way to target acidity. While studying membrane protein folding, we discovered a peptide (pH (Low) Insertion Peptide called pHLIP  ) that reversibly folds and inserts across membranes in response to pH changes, and this discovery has led to a novel way to target acidic tissue. Our biophysical studies have revealed the molecular mechanism of pHLIP action, which is based on the increase of hydrophobicity of carboxyl groups when they become protonated under mildly acidic conditions, leading to peptide insertion into a membrane (1)(2)(3)(4). We have shown that pHLIP can target acidic tissue and selectively translocate polar, cell-impermeable molecules across cell membranes in response to low extracellular pH (1,(5)(6)(7)(8)(9)(10). As Where R is the gas constant and T is the temperature in Kelvin.

pH-dependence
The

Design of pHLIP variants
We have reported the basic molecular mechanism of the interaction of WT-pHLIP with lipid bilayers (1,3,11). In these studies we found three states of the peptide: State I, in solution as an unstructured monomer at neutral pH when no lipid membrane is present; State II, bound at the surface of a lipid bilayer as a largely unstructured monomer at neutral pH, and State III, inserted across the bilayer as a monomeric helix at acidic pH. To broaden our understanding of the main principles of pHLIP peptide interactions with membranes and to select the best sequences for clinical use, we employed our knowledge and designed 16 variants of the WT-pHLIP (Var0) sequence:

Biophysical studies
First, we studied the interaction of the pHLIP variants with lipid bilayers, employing the fluorescence and CD spectroscopic techniques previously used ( Table S1 and Figure S1).
The partitioning of pHLIP variants into bilayers at pHs 8 and 4-5 was assessed in titration experiments by measuring changes of intrinsic peptide fluorescence ( Figure   1). On average, the partitioning of the variants into the membrane at low pH is about 50 fold higher than at high pH. As expected, the truncated variants Var5-13 have lower affinities for the membrane at neutral pH compared to WT and other variants, since a number of hydrophobic residues were removed from the sequences. In general, variants with Glu residues have slightly higher affinities compared to the same variants with Asp (since Glu is more hydrophobic due to the additional methylene group The above numbers reflect both the differences in binding affinities, and to some degree, the strength of helical structure formation in the membrane. The largest differences are observed for Var3, Var5 and Var7. The smallest differences are found for Var2 and Var15 due to the strong interactions of these peptides with the bilayer in state II at pH8. The pKa of peptide insertion is determined by the pKa of protonation of Asp/Glu residues in the TM and at the inserting end of the peptide. As charged carboxyl groups located at the transition zone between the aqueous and hydrophobic environments sink more deeply into the bilayer, the pKa is progressively raised by the reduced dielectric constant, and the pKa of Asp/Glu residues increases. Thus, all truncated variants (Var5-13), which demonstrated weaker interactions with lipid bilayer surfaces at neutral and high pHs, would have lower pKa's of insertion. The pKa of pHLIP variant insertion was measured by following the shift of the position of the intrinsic peptide emission maximum as the pH is changed from 8 to 2 ( Figure 2). As expected, the shift of pKa to lower pHs correlates with the truncation of the pHLIP sequence. Thus, the pKa of Var10, with a single Asp residue in its TM, is the lowest (pKa=4.5). All variants containing Glu residues (Var7-9, 11, 13) have higher pKa values compared to the related sequences with Asp residues (Var5-6, 10, 12), again as expected if the increased hydrophobicity of the Glu causes a deeper association with the bilayer, lowering the dielectric and raising the pK.
To this point, we have considered the equilibrium energies of progressive binding and insertion events, but it is also important to note that sequence variation will alter the kinetics of insertion. The barrier for insertion includes the resistance to the passage of the inserting terminal end of the peptide from the aqueous compartment outside to the aqueous compartment inside a cell or liposome. We have found in recent work that the kinetics of insertion correlates with the number of protonatable groups at the inserting end or the presence of polar cargo (12). The characteristic membrane insertion time for WT-pHLIP is about 30 sec, and slightly faster kinetics were observed for Var14, which is the reverse sequence of WT. Var14 has the same number of protonatable residues at its N-terminus as WT has at its C terminus, except for the free C-terminus itself. The data are consistent with the view that the acetylated N-terminus of Var14 inserts across the bilayer and the amidated C-terminus stays outside, reversing the direction of insertion. The characteristic time of Var1-2 insertion, which has fewer protonatable residues at its C-terminus than WT, was about 1-2 sec, which correlates well with our recent findings (12). Further truncation of the peptide inserting end, which reduces the number of protonatable residues, resulted in faster peptide insertion, completed within first 30-100 msec for Var3-13. Var15-16 have no protonatable residues in the TM, while they have the same number of Asp/Glu as WT at the inserting end, and as a result the time of insertion was reduced to 10-11 sec relative to WT. The kinetics data are shown in Figure 3.

Tumor targeting and distribution of Alexa750-pHLIPs in organs
We

pH-insensitive K-pHLIPs
As a control, we have previously used K-pHLIP-WT, where the two key Asp residues in the TM part of the peptide were replaced by Lys residues (1). K-pHLIP cannot interact with a membrane in a pH-dependent manner over the range of neutral and low pHs we are studying, and therefore should not target acidic tumors. Because of the potential clinical applications of Var3 and Var7 we synthesized and tested K-versions of each: Each of these peptides failed to interact with lipid bilayers in a pH-dependent fashion treated with peptides at pH5.9. The cellular uptake of Var3 was 1.12 times higher at low pH compare to neutral pH, and uptake of Var7 was 1.34 times higher at low pH.
Our biophysical data show that the affinity of Var3 for a membrane at pH8 is higher than the affinity of Var7 at the same pH, therefore we see slightly higher cellular uptake of Var3 at pH7.4 compare to Var7 and a less pronounced difference in cellular uptake between neutral and low pHs for Var3. For the cases of K-Var3 and K-Var7, the fluorescence signal was very similar when cells were treated at either high or low pHs, moreover the fluorescence signal at pH7.4 was slightly higher than at pH5.9.
Also, the distributions of fluorescent signals for D-and E-variants and K-variants were different. K-Var3 and K-Var7 were distributed as cytoplasmic dots. In contrast to pHLIP variants, K-pHLIPs are positively charged and partially aggregated, and, most probably, might be taken up by endocytosis and trapped in endosomes in a pHindependent manner, while cellular uptake of Var3 and Var7 is pH-dependent.
Biodistribution and tumor targeting of Alexa750-K-pHLIPs were also investigated using xenograft models in mice. In contrast to Var3 and Var7, K-pHLIPs did not  Table S5).
The tumor to muscle ratio at 4 hours for Var3 and Var7 labeled with BODIPY and Alexa750 was 6.00.9, while for K-Var3 and K-Var7 it was 2.60.3 (Figure 5g).
Thus, the replacement of the Asp/Glu residues by Lys in the TMs of Var3 and Var7 leads to the loss of both pH-dependent interactions with membranes and targeting of acidic tumors, supporting a central role for the carboxyl groups in targeting.

Discussion
Rapid cell growth and an inadequate blood supply produce hypoxic conditions that cause a partial use of glycolysis in tumor cells, resulting in acidification of the cytosol, to which the cell adjusts by pumping protons into the external environment. But, hypoxia and low blood supply are not the only mechanisms responsible for the development of an acidic environment within solid tumors (14). Malignant cancers have an elevated glucose uptake even under normal oxygen conditions, overwhelming the mitochondrial capacity and using glycolysis for the overflow. This condition is known as "aerobic glycolysis" or the Warburg effect (15). Cells exhibiting a Warburg effect catabolize glucose at a high rate (16,17), and the use of glycolysis results in a much higher level of production of H + and lactic acid, which are pumped across cell plasma membranes into the extracellular space, where they accumulate in poorly perfused regions (18)(19)(20). In addition to the lactic acid-output, intracellular titration of acid with bicarbonate and the engagement of the pentose phosphate shunt releases CO 2 from tumor cells. Expression of carbonic anhydrase 9 and 12 on the tumor cell surface catalyzes the extracellular trapping of acid by hydrating cell-generated CO 2 into HCO 3 − and H + (21,22). These mechanisms combine to create an acidic extracellular milieu favoring tumor growth, invasion and development.
The pHLIP peptides can exploit tumor acidity as a useful biomarker. Based on the results of our previous and current investigations, design principles can be formulated to set directions for different clinical uses: i) all pHLIP peptides have a membrane-binding sequence, which contains a number of hydrophobic residues essential for membrane targeting; ii) a pHLIP is stable across a lipid bilayer at low pH; iii) to achieve pH-dependent targeting, at least one protonatable group (Asp or Glu, or any other) needs to be in the sequence; iv) in addition to the membrane-binding sequence, a pHLIP peptide may contain a membrane inserting sequence (the C-terminus for most investigated pHLIPs or the acetylated N-terminus of the reverse pHLIP sequence), which crosses the lipid bilayer and emerges in the cytoplasm.
The protonatable residues could be only in the inserting sequence or in       The titration of the pHLIP variants with increasing concentration of POPC liposomes at pH8 (a) and pH4-5 (b) are followed using intrinsic peptide fluorescence changes.

Conjugation of pHLIP variants with fluorescent dyes
The lyophilized powder of a peptide was dissolved in a solution containing 3M urea, and the peptide solution was transferred to buffer using a G-10 size-exclusion spin column, or was dissolved directly in phosphate buffer at pH8 for biophysical studies.
The concentrations of the peptides were determined by absorbance (for Var0-2 and For studies on cultured cells and animals each pHLIP variant was conjugated with AlexaFluor750 C 5 -maleimide (Invitrogen, catalog #A30459), and Var3, Var7, K-Var3 and K-Var7 were conjugated with Tetramethylrhodamine-5-maleimide, single isomer (Invitrogen, catalog #T6027) or BODIPY-TMR C 5 -maleimide (Invitrogen, catalog #B30466) in DMF or DMSO at a ratio of 1:1.1 of dye:peptide and incubated at room temperature for about 6 hours and then at 4ºC until the conjugation reaction was completed. The reaction progress was monitored by reverse phase HPLC. The purity of products was assessed by analytical HPLC and peak identity was confirmed by SELDI-TOF mass spectrometry.

Liposome preparation
Large unilamellar vesicles (LUVs) were prepared by extrusion. POPC (1-palmitoyl-2oleoyl-sn-glycero-3-phosphocholine, Avanti Polar Lipids, Inc.) dissolved in chloroform at a concentration of 1 mg/ml were desolvated on a rotary evaporator and dried under high vacuum for several hours. The phospholipid film was then rehydrated in 100 mM phosphate buffer, pH 8.0, vortexed for 2 hours, and repeatedly extruded through membranes with 100 or 50 nm pore sizes to obtain LUVs.

Oriented circular dichroism measurements
Oriented circular dichroism was measured from supported bilayers deposited on a stack of quartz slides with special polish for far UV measurements, with spacers of 0. hour incubation at 100% humidity. Then, excess vesicles were carefully removed and the slides were stacked to make a pile filled with the peptide solution (5 μM) at pH 4.
Bilayers with the peptide solution were allowed an additional 6 hour equilibration.
Measurements were taken at 3 steps during the process: when the monolayers were incubated with the excess of liposomes, soon after spaces between slides were filled with the peptide solution and 6 hours after the second measurement. 14 slides (28 bilayers) were assembled and OCD spectra were recorded on a MOS-450 spectrometer with 2 s sampling time. Control measurements were carried out of the peptide between slides with and without supported bilayers and in the presence of an excess of POPC liposomes.

Assumptions made for the analysis of pH-dependence data
Peptide intrinsic fluorescence changes are used to follow the insertion as a function of pH (transition from the state II to state III as the pH is lowered).  with POPC when the pH is dropped from 8 to 3.6 by the addition of an aliquot of HCl). However, due to the fact that the quantum yields in states II and III are slightly different, there was a slight non-linearity, so we completed the analysis by estimating the contributions of the states for WT-pHLIP and found that the apparent pK shifts no more than on 0.05 pH units toward lower pHs. Since this shift is less than the experimental error, we present the pKa values for transitions from the state II to III based on the analysis of the positions of spectral maxima. However, long circulation times of pHLIP variants in the blood could be very well suited for fluorescence or MR imaging, as well as for the delivery of therapeutic agents. We find that all pHLIP variants that have higher affinities for membranes at neutral pH show high tumor/organ ratios at 24 hours post-injection.

Quantification of cellular uptake
-An important parameter is the difference between the Gibbs free energy of peptide binding to membranes at low pH vs. high pH. This parameter reflects the difference in the affinity of the peptide to membrane at low and neutral pH and the strength of formation of transmembrane helix at low pH. A larger ΔΔG will ensure a greater differentiation between the inserted and non-inserted peptides as a function of pH.
-For applications in vivo, the kinetics of peptide insertion across the lipid bilayer are important for rapid equilibration with tissues and clearance from the blood. Based on insertion kinetics we can group all pHLIP peptides into: i) the peptides with protonatable residues both in the TM and in the inserting end show the slowest kinetics of insertion (minutes); ii) peptides that are truncated at the inserting end and have few or no protonatable residues at the inserting end, partition in bilayer much faster (seconds), iii) peptides that have protonatable residues at the inserting end but not in the TM show intermediate times of insertion (~20 s), and finally, peptides with only one protonatable residue in the TM have the fastest kinetics of insertion, which coincides with time of formation of helical structure (~100 msec).
-Exit kinetics should be different in cells and liposomes, since peptides that have protonatable residues that are translocated across a cell membrane move them to the neutral pH of the cytoplasm. To exit the cell membrane, these groups must be protonated, which is much less likely in the cytoplasm, so these peptides become anchored in a cell. Such an "anchor" can significantly (by orders of magnitude) reduce the rate of peptide exit, and the peptide could stay in the plasma membrane for weeks. Such an effect would explain our observation that mouse tumors are stably labeled with fluorescent WT-pHLIP. Tables   Table S1. The three states of each pHLIP variant. The spectral parameters of the pHLIP variants in states I, II and III are presented. The parameters were obtained from analysis of the fluorescence and CD spectra shown in Figure 1: the maximum position of the fluorescence spectrum  max , in nm; S -the normalized area under the spectra (normalization was done on the area under the spectrum in State I);  225 x 10 3 , deg cm 2 dmol -1 -the molar ellipticity at 225 nm.  Table S2. Mean NIR fluorescence with standard deviation calculated for each organ collected at 4 hours after Alexa750-pHLIP variant administration, n is the number of animals. The mean fluorescence values of organs were normalized to the signal in kidney for each mouse, then averaged for each variant and presented in Figure S3.  Table S3. Mean NIR fluorescence with standard deviation calculated for each organ collected at 24 hours after administration of Alexa750-pHLIP variants, n is the number of animals. The mean fluorescence values of each organ was normalized to the signal in kidney at 4 hours (Table S2) for each mouse, then averaged for each variant and presented in Figure S3.