Predicting small molecule fluorescent probe

Predicting small molecule fluorescent probe localization in living cells using QSAR modeling. 2. Specifying probe, protocol and cell factors; selecting QSAR models; predicting entry and localization RW Horobin1, F Rashid-Doubell2

Biotech Histochem Downloaded from informahealthcare.com by University of Glasgow on 11/03/13 For personal use only.

1School

of Life Sciences, College of Medical, Veterinary and Life Sciences, The University of Glasgow, University Avenue, Glasgow G12 8QQ, Scotland UK, and 2School of Medicine, Royal College of Surgeons in Ireland–Medical University Bahrain, PO Box 15503, Busaiteen, Kingdom of Bahrain

Abstract We describe the practical issues and the methodological procedures that must be carried out to construct and use QSAR models for predicting localization of probes in single cells. We address first the determination of probe factors starting with a consideration of the chemical nature of probe molecules present. What is their identity? Do new compounds arise in incubation media or intracellularly? For each probe, how many distinct chemical species are present? For each probe species, the derivation of the following numerical structure parameters, or descriptors, is set out with worked examples of electric charge and acid/base strength (Z and pKa); hydrophilicity/lipophilicity (log P); amphiphilicity (AI and HGH); conjugated bond number and largest conjugated fragment (CBN and LCF); width and length (W and L); and molecular and ionic weights, head group size and substituent bulk (MW, IW, HGS and SB). Next, protocol factors are specified by focusing separately on the mode of introduction of the probe to the cells, other application phenomena, and factors that influence directly observations of outcomes. Cell factors then are specified by considering separately structural and functional aspects. The next step is to select appropriate QSAR models and to integrate probe, protocol and cell factors to predict the interactions of the probe with the cell. Finally, we use an extended case example to explore the intracellular localization of certain photodynamic therapy dyes to illustrate these procedures. Key words: intracellular localization, PDT, QSAR models, single cell systems

In Part 1 of this paper (Horobin et al. 2013a), we described small molecule fluorescent probes and their applications, and provided a general account of using QSAR rules to predict patterns of probe Correspondence: Richard W. Horobin, School of Life Sciences, College of Medical, Veterinary and Life Sciences, The University of Glasgow, University Avenue, Glasgow G12 8QQ, Scotland UK. Tel: 44 (0)1796 474 480, E-mail: Richard.Horobin@glasgow. ac.uk © 2013 The Biological Stain Commission Biotechnic & Histochemistry 2013, 88(8): 461–476.

DOI:10.3109/10520295.2013.780635

uptake and localization in the context of single cell systems. Here in Part 2, we consider the practical issues and methodological procedures required to construct and use QSAR models for single cells. We defer until Part 3 discussion of integration of models to allow analysis of more complex events and of multicellular organisms. We discuss first details concerning probe factors, especially estimation of structure parameters, protocol factors and cell factors. Next, we consider the choice of appropriate QSAR core models and the integration of these with 461

quantities of contaminating dyes. In either case, dyes with significantly different chemical characteristics may be present simultaneously in the mixture.

the various factors listed above. Finally, we present a case example of the intracellular localizations using QSAR modeling of a group of dyes used in photodynamic therapy research.

Determining probe factors How many and which probe species are present?


Answers to this question often are surprisingly complex. Some complexities and suggestions about how best to deal with them are discussed here using a checklist format. What is the identity of the probe? The simplest approach is to seek answers in hardcopy or online resources. Sources of particular value include: • Conn’s Biological Stains: editions 9 and 10 of this handbook complement each other and the earlier volume includes reagents of historical interest. For bibliographic details for this resource, see Lillie (1977) and Horobin and Kiernan (2002) for the 9th and 10th editions, respectively. Many fluorescent probes in current use, however, post date both editions. • There are several other relevant handbooks and monographs including those by Green (1990) and Rost (1995). Hard copy versions of the earlier editions of the Molecular Probes catalogue (Haugland 1996) also provide useful resources. All suffer from the limitation noted above. • Online sources are more current, but in some cases less reliable. Useful examples include the NIH databases, ChemIDplus and PubChem. Online catalogues such as those from Life Technologies and Sigma-Aldrich often are informative. • Although Google Images may offer a useful shortcut to molecular structures for some probes, one may need to be a chemist to use this major resource at all, and a sceptical chemist to use it without error (see Appendix 1). Several problems can preclude, and do confuse, QSAR analyses. For example: • Some proprietary probes have undisclosed structures. • Certain structures of probes given in the literature or on websites are incorrect. • Other structures, although not incorrect, are presented in a way likely to mislead non-chemists. • Commercial lots of probes may comprise mixtures of dyes or at least contain substantial 462

Examples of such problems are given in Appendix 1. From this point forward, the correct identity of the probe of interest is assumed to be known, or if impure or a mixture, the identities of all the compounds present are assumed to be known. The issues of stability and reactivity can therefore be addressed. Is a compound likely to be chemically or biochemically transformed, either inside or outside a live cell? More specific questions follow from this.

Are new chemical compounds generated within the incubation medium or live cell under physiological conditions? Because new compounds can be formed, such an assessment must be made for each probe present. It is convenient to categorize these phenomena as: chemical reactivity, i.e., instability of the probe; accidental or artifactual biochemical reactivity of the probe; and biochemical reactivity of the probe that is intended or at least acknowledged and allowed for in the protocol. Failure to account for such phenomena can preclude, and often will confuse, QSAR analyses. Examples of these problems are given in Appendix 2. While it is useful also to consider any concomitant modification or damage of the cell resulting from dye–cell reactions, such effects are considered under the heading of cell structural factors and are discussed later (Table 5). The next question is: for each probe compound, how many distinct chemical species are present under physiological conditions? Owing to a number of physicochemical phenomena, a single chemical compound may be present in different forms with different physicochemical characters. Each of these species must be treated separately in a QSAR analysis as reported by Franco and Trapp (2008) concerning their physicochemical modeling. This assessment must be carried out for both compounds present in the incubation medium (or its equivalent) and those present in the varied compartments of a live cell. The more common phenomena considered here are: acid/ base equilibria, which are influenced by the varied pH of different cellular compartments (Table 1); pseudobase/carbinol formation; and oxidation/ reduction phenomena.

Biotechnic & Histochemistry 2013, 88(8): 461–476

Such problems also may preclude, and do complicate, QSAR analyses. For examples, see Appendix 3.

Obtaining numerical structure parameters (descriptors)


We consider parameters cited most frequently in published QSAR reports that address intracellular probe localization. These parameters were listed in Table 1 of Part 1 of this paper (Horobin et al. 2013a) under the rubric, “Summary account of numerical structure parameters used to describe and model fluorescent probes.” Seeking parameter values: options and examples Some parameter values may be obtained from the literature in the usual manner, e.g., by searching: • Research papers and review articles using freely accessible databases such as PubMed and PubChem. Proprietary databases such as ChemSpider, CrossFire Beilstein, NIST Chemistry Webbook or SciFinder also can be used for those with access. • The monograph literature in addition to the above by using Google Scholar. Table 1. Typical pH values for various cellular compartments; taken from various reviews: Loiselle and Casey (2010), Llopis et al (1998), Mindell (2010), Rivinoja et al. (2012), and Takahashi et al. (2001) Cellular structure Cytosol Endoplasmic reticular lumina Endosomes, early & late

Typical pH values When external pH ~ 7.0, cytosolic pH ~ 7.0 7.2 5.5–6.5

Golgi apparatus: Trans Golgi network Cis Golgi network Lysosomes

6.0 6.7 4.5–5.0

Mitochondrial matrix: Non-respiring Respiring Nucleus

Secretory vesicles

As cytosol Up to 1.0 pH unit above cytosolic pH Can be 0.3–0.5 pH units higher than cytosolic pH, see Seksek and Bolard (1996) 5.5

Google also can be used to find on-line sources of additional kinds. Resources generated by individual academics include course handouts and substantial data compilations on personal websites. Other materials come from institutional educational sites that are posted by both universities and the news media; other substantial information sources are vendor based. Although this approach sometimes may provide values of parameters such as log P and pKa, little or no information about more specialized parameters can be expected. In the latter case, values must be estimated or measured, as for log P or pKa, Possibilities and problems vary with each parameter as discussed below. Because multiple species of a single compound can be present simultaneously, a caution is offered. The pKa and log P values required are those related to individual species, termed log P and pKa micro values. For discussion of this issue, see Avdeef (2003) and Perrin et al. (1981). When seeking information, especially online, note that some sources present the macro rather than the required micro values for these parameters. Also, online sources may quote log P values for neutral species even when queried for the ionic species that exist under physiological conditions.

Z/pKa: procedures Why these parameters are significant Intracellular distribution of probes may be influenced directly and strongly by electric charge as outlined in Part 1 of this review (Horobin et al. 2013a), e.g., owing to the presence of membrane potentials across organelle membranes. Indirect effects of electric charge also arise by the effect of charge on hydrophilicity/lipophilicity, because ionic substituents increase probe hydrophilicity. Evaluating the electric charge of a probe under physiological conditions often requires knowing, at least approximately, the acid and base strengths of the probe’s substituents, which usually are assessed by their pKa values. Consequently, pKa also is a significant factor that influences both uptake and localization. Weak bases accumulate in acidic organelles, such as lysosomes, whereas strong bases localize in the nucleus if hydrophilic, or favor the mitochondria if lipophilic. Indirect effects of pKa on uptake also may occur. The cationic form of a weak base may be hydrophilic and membrane impermeable, while the electrically neutral free-base species may be lipophilic and therefore permeable. Unfortunately, published chemical structures can mislead the non-chemist, because the species

Symposium: QSAR for localizing fluorescent probes 463


illustrated may or may not be those present under physiological conditions. As an example, consider rhodamine B. Under neutral conditions this dye exists predominantly as a zwitterion, while the free acid and spiro species, as shown in Fig. 1, are present only under acidic conditions. For sources of pKa values in a large (six volume) collection of critically selected values, see Smith and Martell (1974). There also is an electronic version of these data (Smith et al. 2012). When, as often is the case, no experimental value is available, an estimate may be made in various ways. These are summarized in the classic monograph by Perrin et al. (1981), which contains worked examples for the various methods of estimation. A number of commercial software packages also provide pKa predictions for drug-like molecules into which category dyes and small molecule probes fall. Nine such programs have been critically compared and evaluated (Chenzhong and Hicklaus 2009).

Log P procedures and case examples Why this parameter is significant Both probe uptake into cells and subsequent intracellular distribution are influenced by probe lipophilicity, for which the log P parameter is a model, as outlined in Part 1 (Horobin et al. 2013a). Sources of log P values Information concerning approximately 15,000 compounds is available in publications from the Pomona College MedChem Database (Leo et al. 1971, Hansch and Leo 1979). This material also is available electronically as the MedChem database maintained by BioByte Corp. As a widely used alternative to direct measurement of partition coefficients for both actual and proposed compounds, log P values often are obtained using software-implemented computational systems based on knowledge of the molecular structure of the compound. Available systems vary in terms of their physicochemical and computational basis, ease of use, range of application, and accuracy of outcome as critically reviewed by Mannhold et al. (2009). In particular, many commercially available software-implemented systems, including the CLOGP method of Hansch and Leo (1979), do not provide log P values

Possible alternatives to the pKa parameter Although other acidity and basicity functions are available, e.g., the Hammett acidity function, the pH scale is the most commonly used. This is not surprising, because it is well suited for dilute aqueous solutions and therefore for biological systems.

H3CH2C

CH2CH3 N

CH2CH3 N CH2CH3

O

H3CH2C

CH2CH3 N

COO

a)

CH2CH3 N CH2CH3

O

COOH

c) H3CH2C

N

b)

CHCH3 N CH2CH3

O

CH3 COOH

H3CH2C

CH2CH3 N

O

CH2CH3 N CH2CH3 COOH

d)

H3CH2C

e)

CH2CH3 N

CH2CH3 N CH2CH3

O O C O

Fig. 1. Variant structural formulae of rhodamine B found in hardcopy and online sources (Summer 2012). Structure (a) is the zwitterionic species present under physiological conditions. Structures (b) and (c) are chemically in error. Structures (d) and (e), while not chemically impossible, illustrate species (free acid and spiro forms, respectively) that are unlikely under physiological conditions.

464



for ionic species. The choice of procedure thus is a compromise, because the QSAR models used and discussed here require log P values for ionic species. Consequently, we have used the manual method developed by Hansch and Leo (1979) despite its distinct lack of user friendliness compared to software-mediated systems. The manual Hansch and Leo (1979) approach is described in some detail in the cited book chapter, which also includes many worked examples of log P estimation of both uncharged and ionic species. The procedure is based on splitting a molecule into fragments composed of an atom or group of atoms whose contributions to overall log P (fragment or ƒ values) have been derived from experimental measurements made on a set of test compounds. This approach is described as “group contribution,” “fragmental” or “substructure based.” Certain structural effects, e.g., chain length and chain branching, or interactions between polar fragments, e.g., electronic and steric effects, however, are not accounted for by merely aggregating ƒ values. To deal with such complications, bond and interaction factors (F) are used, which again are derived from experimental measurements on a set of test compounds. The cited chapter of Hansch and Leo (1979) provides an extensive tabulation of both ƒ and F values. When the appropriate values are available, log P values for a known structure are obtained as follows: inspect the molecular structure to identify fragments with tabulated ƒ values, sum the ƒ values, inspect the molecular structure to identify structural features or interactions with tabulated F values, and sum the F values. The totalled ƒ and F values provide an estimated log P value.

Example estimations These are carried out on three fluorescent probes with very different physicochemical characteristics. Note that the symbolic formalism used in the following examples is that of Hansch and Leo (1979). Example 1 is pyranine whose structure is given in Fig. 2a. Only fragment values are required for this estimation. Fragments: 6ƒφCH 6ƒφC 4ƒφC 3ƒφSO3 1ƒφOH 6(0.36) 6(0.23) 4(0.13) 3(4.53) 1(0.44) Calculated log P 10.0

Example 2 is Nile red whose structure is given in Fig. 2b. Both fragment and factor values are required for this estimation. Fragments: 8ƒφCH 5ƒφC∗ 1ƒφC 1ƒφC 1ƒφ N 1ƒφ O 1ƒφ CO 1ƒφN 2ƒCH3 2ƒCH2

a)

O S

OH

3

Log P = –10

O S

SO

3

b) H C 3

H C 3

H 2 C

3

Log P = 5.7

N N

CH

2

O O

c) HGH = –4.4

AI = 5.8 H C 3

N

CH3

CH3

Log P = 1.4

Fig. 2. Graphical and numerical representations of the lipophilicity and amphiphilicity of three probes. a) Pyranine, a hydrophilic probe with no distinct hydrophilic (head group) or lipophilic (tail) domains, hence not significantly amphiphilic. b) Nile red, a lipophilic probe with no distinct hydrophilic domain or substantial lipophilic domain, hence not significantly amphiphilic. c) TMA-DPH, a lipophilic probe with distinct hydrophilic and lipophilic domains, hence strongly amphiphilic. Red color coding indicates lipophilicity, blue indicates hydrophilicity; strong colors indicate domain properties, pale colors indicate overall properties. 8(0.36) 5(0.44) 1(0.23) 1(0.13) 1(1.12) 1(0.08) 1(0.59) 1(0.93) 2(0.89) 2(0.66) Factors: 2FbYN 2FφP2(N/O) 1FφP2(N/CO) 2(0.2) 2[0.8(1.12 0.08)] 1[0.08(1.12 0.59) Calculated log P 5.8 – 0.1 5.7

Example 3 is the probe, TMA-DPH (trimethylammonium diphenylhexatriene), whose structure is given in Fig. 2c. Both fragment and factor values are required for this estimation. Fragments: 15ƒφCH 3ƒφC 3ƒCH3 1ƒφN 15(0.36) 3(0.13) 3(0.89) 1(4.41) Factors: 3Fbx 1 3(0.9) Calculated log P 4.1 2.7 1.4

Possible alternatives to this parameter A major disadvantage of the procedure described above is its complexity. Consequently, investigation of an alternative software-based procedure has been explored by Dapson (2013) and is reported in this symposium issue. To date, all of our papers concerning QSAR have used the manual Hansch and Leo (1979) approach and the limiting parameter values in the published models reflect this. Alternative computational methods would need either to provide an alternative set of limiting values or the


alternative parameter values must be amenable to being transformed into equivalent log P values. For a preliminary investigation into these possibilities, see Dapson and Horobin (2013).


AI and HGH: procedures with case examples Why these parameters are significant Both probe uptake and subsequent intracellular localization are influenced by amphiphilicity as outlined in Part 1 (Horobin et al. 2013a). This property requires a molecule to possess distinct lipophilic and hydrophilic domains, sometimes termed “tail” and “head group,” respectively. Consequently, the parameters modeling these features (AI or amphiphilicity index and HGH or head group hydrophilicity) are significant. For a graphical illustration of the meaning of these parameters, see Fig. 2. Sources of AI and HGH values No tabulation of these parameters has been published, so the values must be estimated. Both are the nominal partition coefficients of limited molecular domains and are estimated using the Hansch and Leo (1979) manual procedure as below. Inspect the molecular structure to identify the lipophilic and hydrophilic domains (tail and head group). Then for each domain, inspect the domain structure to identify fragments with tabulated ƒ values. Sum the ƒ values, inspect the domain structure to identify structural features or interactions with tabulated F values, then sum the F values. The total of the ƒ and F values provides estimated AI and HGH values.

Example estimations These will be carried out on the same three fluorescent probes as used for the sample log P estimates. Example 1 is pyranine which has several separate hydrophilic groups surrounding a lipophilic core as shown in Fig. 2a. Because no distinct tail or head group can be defined, no meaningful AI or HGH values can be assigned. Example 2 is Nile red, shown in Fig. 2b, which is less symmetric and possesses a fused and unsubstituted phenyl ring that constitutes a small lipophilic domain. Because the hydrophilic groups are scattered across the molecule, however, there is no distinct head group and the dye is not amphiphilic. Example 3 is TMA-DPH, shown in Fig. 2c, which possesses both lipophilic and hydrophilic domains and is amphiphilic. Estimation of the AI and HGH

466

parameters can be carried out as follows. For AI, only fragment values are required. Fragments: 15ƒφCH 31ƒφC 15(0.36) 3(0.13) Calculated AI 5.8

For HGH, both fragment and factor values are required. Fragments: 3ƒCH3 1ƒφN 3(0.89) 1(4.4) Factors: 3Fbx 1 3(0.9) Calculated HGH 1.7 – 2.7 4.4

For certain other amphiphilic probes however, the boundary between the lipophilic and hydrophilic domains cannot be distinguished by mere visual inspection of the structural formulae. An iterative approach is required that involves repeated calculations of variant substructures to define the head group, because, as noted in Part 1 of this review, HGH – 1 is a necessary condition for significant amphiphilicity. Possible alternatives to these parameters Because both AI and HGH are nominal log P measures, see comments under log P above. CBN and LCF: procedures and case examples Why these parameters are significant As outlined in Part 1 (Horobin et al. 2013a), probe uptake is influenced by the overall size of a probe’s conjugated system (CBN value), while significant intracellular nucleic acid binding requires a planar conjugated domain that exceeds a minimum size and LCF value. Sources of CBN and LCF values No tabulations of these parameters have been published, so they must be estimated. To establish the CBN value of a dye, inspect the molecular structure to identify all conjugated components (Appendix 4 gives examples of such elements) and sum the number of bonds in all conjugated regions. Example 1 is the mitochondrial probe, JC-1, which contains several unconjugated alkyl chains that protrude from a planar conjugated domain (Fig. 3a). In the figure, one can count 28 bonds in the conjugated domain, so CBN 28. Example 2 is fura-2 salt, a calcium probe that has 29 conjugated bonds split into two domains, so CBN 29 (Fig. 3b). Various unconjugated aliphatic moieties are appended to the conjugated domains. Example 3 is sulforhodamine B, a fluorescent probe with a variety of applications, which contains 25 conjugated bonds (Fig. 3c). Various unconjugated


[28] Cl

C2H5 N

Cl

Cl

C2H5 N

N

C2H5 Cl

C2H5

N

[18]

C2H5 O

C2H5

SO3

a)

[7]

O2CH2C N O

H3C

[8]

O


O C O

SO3

CH2CO2

CH2 CH2 O O

C2H5

N

C2H5

N

N

c)

N

O2CH2C

[21]

CH2 CO2

b) Fig. 3. Dye structures that illustrate the difference between the CBN and LCF structure parameters. The overall size of the conjugated (aromatic) domain or domains in a probe is modeled by the conjugated bond number (CBN). The size of the largest conjugated domain is modeled by the largest conjugated fragment (LCF) parameter, which may be identical to or less than CBN. LCF domains are color coded blue. The dyes illustrated are: a, JC-1; b, fura-2 salt; and c, sulforhodamine B. The contribution of each conjugated domain to CBN and LCF is indicated by the bracketed numbers.

ionic and alkyl groups protrude from the conjugated region whose CBN 25. If the LCF is required, inspect the molecular structure to identify all conjugated components as above, inspect the structure to identify electronic separations between conjugated regions that produce separate conjugated fragments (separations are of several types, see Appendix 5) and if a separation is found, count the numbers of bonds in each fragment. The LCF is the count for the largest fragment. Example 1 is JC-1, which contains only a single conjugated domain (Fig. 3a). The LCF value, therefore, is the same as the CBN value, i.e., 28. Example 2 is fura-2 salt, which has two conjugated domains separated by a short aliphatic chain (Fig. 3b). The largest such domain, color coded blue in the figure, contains 21 bonds, so the LCF 21. Example 3 is sulforhodamine B (Fig. 3c), which also has two separate conjugated domains, the largest of which, color coded blue in the figure, contains 18 bonds, so LCF 18. To non-chemists this may seem surprising, but the pendent phenyl group with its bulky ortho sulfonate substituent cannot lie in the same plane as the tricyclic xanthine moiety; the resulting sterically forced rotation prevents conjugation of the two aromatic domains.

Possible alternatives to these parameters Some modeling software computes the number of pi atoms (atoms connected by double bonds) in each conjugated system (Dapson and Horobin 2013). This is useful for correctly identifying certain atoms and conjugated systems that may not be evident to non-chemists. Width (W) and length (L): procedures and case examples Why these parameters are significant Binding of cationic probes to nucleic acids is influenced strongly by the shape of the planar fluorophore as outlined in Part 1 (Horobin et al. 2013a). More rectangular probes tend to intercalate with DNA in solution, and to bind selectively to ribosomal RNA in nucleoli and cytoplasm in living eukaryotic cells. Probes that are more rod shaped tend to be minor groove binders; in living cells, these bind selectively to chromatin. Such differences in aspect ratio of the fluorophore may be defined in terms of W and L of molecular structures (Del Castillo et al. 2010). Sources of W and L values A tabulation of such data is available for some nucleic acid targeting probes (Del Castillo et al. 2010). As described in detail by Del Castillo et al.



(2010), W and L can be determined using appropriate molecular modeling software. These investigators also showed, however, that an alternative simplistic approach often is satisfactory. One can count the numbers of bonds along the length of a fluorophore to obtain L and count the number of bonds orthogonal to the long axis of the fluorophore to obtain W. Example estimations using the simplistic approach Figure 4 shows, respectively, the structural formulae of acridine orange and bisbenzimide, which are intercalative and groove binding nucleic acid probes. The bonds in each fluorophore that are perpendicular to the long axes, where their sum is the simplistic W value, are color coded red in the figure. The bonds paralleling the long axis of each conjugated fluorophore, where their sum is the simplistic L value, are color coded green in the figure. The W/L ratio is 0.38 for acridine orange and 0.18 for bisbenzimide, which may therefore be termed “rectangular” and “rod shaped,” respectively. Alternative parameters A variety of shape and steric parameters have been developed by medicinal chemists for pharmaceutical studies, and certain of these are being explored for application as fluorescent probes. For example, some modeling software allows the distance between atomic centers to be estimated (Dapson and Horobin 2013).

a) H3C

N

N H

CH3

b)

CH2 N NH H3C H2C CH2 H2C

N N H

N

CH3

CH3

OH

H N N

Fig. 4. Structural formulae of acridine orange (a) and bisbenzimide (b), which are intercalative and groove binding nucleic acid probes, respectively. Simplistic W (width) values are obtained by counting the number of bonds in the fluorophore orthogonal to its long axis; these are coded red. Simplistic L values are obtained by counting the number of bonds in the fluorophore parallel to its long axis; these are coded green. The W/L ratio is 0.38 for acridine orange and 0.18 for bisbenzimide.

468

MW, IW, HGS and SB: procedures and case examples Why these parameters are significant The overall size of a probe influences its ability to pass through gap junctions and plasmodesmata and overall size can be modeled by molecular or ionic weight (MW, IW) as is outlined in Part 1 (Horobin et al. 2013a). The size of the hydrophilic domain (head group) of an amphiphilic probe influences the rate of flip-flop across a membrane bilayer, hence membrane permeability, or conversely, retention of a probe within the plasma membrane. Size may be modeled by the sum of the atomic weights of atoms within the head group (HGS). The size, or “bulk,” of the substituents attached to the conjugated domain that constitutes the fluorophore of a probe can influence possible binding mechanisms with double-stranded DNA (Müller et al. 1973). Examples of overall size Inspecting the probes whose structures already have been illustrated, it is apparent visually that fura-2 and TMA-DPH (Figs. 3b and 2c, respectively) are the largest and smallest respectively. Also, the IW of fura-2 is 636 and that of TMA-DPH is 291; these values are the extreme numerical values of this parameter for this data set. Because of the visually unassuming character of –SO3– in conventional twodimensional chemical structures, the relative sizes of sulforhodamine B and HPTS are much harder to estimate from mere visual inspection. Use of numerical parameters, however, gives IWs of 558 and 455, respectively and renders the size differences obvious. Examples of head group size Inspection of the structures of the two amphiphilic probes, TMA-DPH and sulforhodamine B (Figs. 2c and 3c, respectively), suggests that the former has the smaller head group. This difference, however, is dramatically obvious when the HGSs, of 59 and 235, respectively, are compared. Examples of substituent bulk Preliminary modeling of substituent bulk (SB) as the sum of the atomic weights of the constituent atoms suggests that rectangular fluorophores able to bind to DNA fail to intercalate and instead bind in the minor groove when SB is greater than 100 (Müller et al. 1973, Horobin et al. 2013b). Thus, although the fluorophore of the well known nuclear probe, DRAQ5, is rectangular, with W/L 0.33, the value of SB is greater than 140 and the dye binds in the minor groove of double-stranded DNA.


Possible alternative molecular volume parameters A wide variety of alternative measures of molecular volume are available. For example, some modeling software sums atomic weights for specified constituent atoms, which provides estimates of HGS and SB without requiring the manual procedures described above (Dapson and Horobin 2013).


What additional structure parameters might be useful? Drug designers have devised and explored many structure descriptors. For example, the E-Dragon software (developed by Prof. Todeschini’s Milano Chemometrics and QSAR Research Group), widely used by pharmaceutical chemists, enables the estimation of more than 1600 descriptors. While it is likely that some may prove of use for predicting probe localization the only additional parameter currently of known significance is solubility. This is an underestimated limiting factor. If a probe is poorly soluble, then transfer from solvent to cell, and sometimes the maximum possible amount of probe in a cell target, will be restricted. For a general discussion, including procedures for estimating solubility from structural information, see Avdeef (2003). Tabulations of solubility data for fluorochromes and dyes exist (e.g., Baughman et al. 1996, Green 1990, Horobin and Kiernan 2002, Lillie 1977), but the accuracy of these data

is uncertain, because determining valid solubilities of dyes is difficult, as explained by Baughman et al. (1996). A final phenomenon can be mentioned at this point, i.e., probe reactivity with cell components. This can occur, for example, with the covalently binding isothiocyanates, FITC-fluorescein and SITS, and also with the reactive ester Cy dyes. If these species are applied from an external solution, and if they react rapidly, they bind to proteins in the plasma membrane and the resulting fluorescent labeling internalizes only if the membrane internalizes. Slower reacting probes, such as the various “Tracker” compounds, are intended to react with the proteins of the organelles to which the other physicochemical features of the probe directs them, hence the terms LysoTracker, MitoTracker etc. Of course, hopes will not always be satisfied.

Specifying protocol factors A detailed picture of factors that can influence uptake, interaction and observation of probes with cells should be assembled. For this purpose, it is convenient to focus separately on the mode of introduction of the probe to the cells, other application phenomena, and factors directly influencing observations of outcomes. For check lists relevant to these steps, see Tables 2, 3 and 4, respectively.

Table 2. Protocol factors concerning entry of probes into cells, with examples of consequences, especially regarding probe localization How entry is achieved Immersion of cells in a solution of the probe (incubation)

Incubation following a membrane permeabilization procedure

By direct insertion through the plasma membrane

Particular methods

Probe localization & other consequences

Probe is in free solution

Following passive diffusion, probes enter the cytosol; following endocytosis they enter endosomes/lysosomes

Probe is solubilized by a “carrier” such as serum albumin (SA) or cyclodextrin

SA has esterase activity, and can modify AM ester probes extracellularly; in some cell lines SA is endocytosed and enters endo-/lysosomes

Probe is solubilized within liposomes

Probes in liposome membranes enter the plasma membrane; if in liposomal lumena, probes enter the cytosol

Following a direct insult, e.g. electroporation, mechanical scraping, or sonication

Probes enter the cytosol

Following membrane modification, e.g. with saponins or surfactants

Re-sealing of plasma membrane may be slow, leading to loss of cell constituents

For instance ballistically or by microinjection

Probes enter the cytosol


Table 3. More protocol factors, and some of their consequences Type of protocol factor

Consequences

Varies from suspension cultures of single cells to organ cultures

Probe penetration easiest in cell monolayers, slowest in organ cultures

Most comprise monolayers of cultured cells attached to glass or resin surfaces

Monolayers have least overlapping of organelles

Timing of probe entry into the cell

As a single bolus or pulse or continuously (incubation)

Pulse labeling required if a controled post-entry wash-through is used

Timing of post-entry incubation

Incubation times vary from minutes to days

Longer incubation times increase risk of cell damage

Temperature of insertion or incubation media

From body temperature to that of a refrigerator (~ 4o C)

Low temperatures inhibit endocytosis

Probe concentration/form

From nanomolar solutions to colloidal particles

Molecules in solution may be pinocytosed; colloidal particles may be phagocytosed

Nature of the specimen


Range of variation of factors

Higher concentrations may be toxic, or give nonspecific uptake patterns Presense of co-solutes, e.g., serum albumin

May facilitate uptake of some probes, inhibit that of others

Ionic strength

If high, can induce probe aggregation; or osmotic damage to endosomes/lysosomes

pH

Influences membrane permeability of probes; low pH may facilitate entry of weak acids, high pH that of weak bases

Special comment is required when the interactions of probes with cells are investigated using microscopy. Observations of this kind often are made by graduate students in departments of chemistry or physics. Most of these investigators, therefore, have little biological knowledge and they may also lack an appreciation of the fundamentals of microscopy. The latter problem is exacerbated when the microscope is an institutional resource and is used with minimal expert assistance. The semi-automated character of much modern instrumentation exacerbates the problem further by inducing false optimism. For such bench workers, the overarching advice is to make friends and allies of biologists and microscopists.

Specifying cell factors The following issues must be considered for each cell type present. Initially one must ask whether a single cell line is being investigated, whether the probe in question been applied to such cells before, and if so, what happened. Both published 470

work and possible in-house information should be sought. It then is convenient to divide the analysis of cell factors into “structural” and “functional” topics as outlined in Tables 5 and 6.

Marrying probe/protocol/cell factors to appropriate QSAR models to predict interactions of probe with cell: concept and case example Currently available QSAR models were summarized in Part 1 of this review (Horobin et al. 2013a). These models fall into several categories which, taken together, allow prediction of an overall outcome of the interactions of a probe species with a cell. For each chemical species, each of the following groups of QSAR models (the terms reflect those used as section headings in Part 1) must be considered: modeling probe features that control cell uptake, access of membrane impermeable probes to the cytosol, modeling probe features that control intracellular localization, relocalization/redistribution


Table 4. Protocol factors related to the mode of observation of the interaction of probes with cells Type of protocol factor

Specifics

Consequences

Timings of observations At a set time following probe insertion

Typically specified in the protocol

May miss rapid or slow localization changes; ignores cell-line variations

Repeatedly or continuously Increases likelihood of problems due to probe fading, or photodamage to the cell


Relating to microscopy Specimen thickness

Even using a single cell line this can vary with protocol; cell monolayers usually being thinest

Thinner specimens take up probes faster, and have less overlap of organelles

Optical issues

Glare must be minimized, and resolution maximized

Glare and low resolution both limit observable detail

High resolution images should be obtained routinely

Needed for interpretation of localization, even if publication images are of low resolution

Probe stability

Varies between probes and problems increases with lengthening incubation time

High rates of fading limit observations, and make quantitation difficult

Specimen stability

Probes may be toxic, and many increase photodamage of the cell

Damaging cell structure and function often reduces the value of observations

Geometrical patterns are observed, eg, “blobs” or “bubbles” or “rods” or “haze”

Patterns must be interpreted, e.g., “blobs” may be lysosomes or fat droplets; “bubbles” may be endosome membranes or photodamaged mitochondria

Relating to image interpretation Interpreting observations is an active process requiring skill and knowledge

of probes in damaged cells, and loss of dyes from live cells. Checking the numerical structure parameter values for each probe species against each model in each category together with the relevant cell and protocol factors allows the nature of the overall interaction to be predicted. A simplified graphical summary of this approach is given in Fig. 8 of Part 1 of this review. The procedure is illustrated below by a case example that corresponds to “Assessing the behavior of novel or mechanistically puzzling probes” in Part 1. Exploring the intracellular localization of certain PDT dyes: a case example Several investigations by Kuimova and associates (Kuimova et al. 2009, da Silva et al. 2011, Kuimova

2012) have explored and clarified the lifetimes and distribution of intracellular singlet oxygen produced by dyes, the phenomenon that underlies photodynamic therapy (PDT). Four porphyrin dyes with contrasting physicochemical character were investigated: chlorin, TDFPPM, TDFPPS and TMPyP. Despite careful work that included fluorescence microscopy, the distributions of the dyes within the cells and the mechanisms of the differential distributions were not always clear. In the case of chlorin, which was observed to accumulate in cytoplasmic granules, Kuimova et al. (2009) stated, “we have not been able to determine the exact organelles and/or structures in the cytoplasm in which this hydrophobic dye associates.” In the case of TMPyP, it was observed that initially it was “principally localized in lysosomes” (da Silva et al. 2011). After further incubation and


Table 5. Structural cell factors, including examples and consequences


Type of structural cell factor

Examples

Consequences

Overall form of the cellular specimen

Isolated cells or cells in organ culture or in whole organisms (either unicellular or multicellular)

Influence uptake rate and visibility of cell components

Number of cell types present

One if a single cell line; multiple if organ culture

Each type must be considered separately

Identity of organelles occurring in unusually large numbers

Mitochondria in cardiomyocytes and lysosomes in phagocytes

Large numbers of a single organelle can influence localization patterns

Presence of gap junctions (or of plasmodesma)

In epithelial sheets and in filamentous algae

Hydrophilic probes can move between cells

Presence of membrane-membrane contact between cells

In epithelial sheets and in myelin sheaths of neurons

Lipophilic probes can move between cells

Presence of extracellular walls outside the plasma membrane

Plant cellulose walls and bacterial peptidoglycan cell membranes

May trap probes and inhibit entry into cell

Presence of an extracellular matrix or superficial adherent layer surrounding cells

Eukaryotic bone & cartilage matrices, eukaryotic glycocalyces and bacterial biofilms

May trap probes and inhibit entry into cell

Change in cell structure due to probe or co-solute reacting/ interacting with cell components

Changes in membrane fluidity

Uptake of certain probes may be faciliated selectively in fluid regions

Permeabilization of membranes, e.g., due to photodamage

Probe may leak from organelles, e.g., lysosomes into cytosol Probe may leak from cell into incubation medium Organelle or cell contents may leak

illumination, “TMPyP ultimately tends to localize in the nucleus … Nevertheless, appreciable amounts … appear in the cytoplasm” (Kuimova et al. 2009). Again, it could not be determined in which cytoplasmic structures the dye was present. Moreover, after extended illumination it was observed that some dye “leaks out into the extracellular medium” (da Silva et al. 2011). Imaging data for the dye, TDFPPM, was given by da Silva et al. (2011, Fig. 7), and there it can be seen that the dye was present in both cytoplasmic granules and in non-punctate form within a region adjacent to the nucleus; no comments were offered about this. No observations about localization were reported for TDFPPS. Unavoidably, given the limited localization information available, the mechanisms suggested for sensitizer accumulation are sketchy. For example, the marked differences between chlorin and TMPyP were described merely as “Due to their vastly different lipophilicity” (Kuimova 2012); for TMPyP it was said only that “in the nucleus… it most likely 472

binds to DNA.” Concerning dye in the cytoplasm, it was stated that “it is unclear whether extra-nuclear TMPyP freely diffuses in the cytosol or if it binds, for example, to proteins” (Kuimova et al. 2009). Because intracellular localization is so important in PDT, the question becomes whether information regarding distribution of sensitizer dyes can be predicted by QSAR modeling in the absence of additional experimental data. To address this issue, we follow here the procedures used earlier in the present paper for the four dyes in our data set.

Specifying probe factors How many compounds and species were present? Three of the dyes had been synthesized in-house and their structures investigated, the fourth was purchased from a fine chemicals supplier. We assume, therefore, that each dye lot consisted of a single compound. None of the dyes is obviously chemically or biochemically unstable, but because all are photosensitizers, fading due to


Table 6. Functional cell factors, including examples and consequences Type of functional cell factor

Examples

Consequences Processes differentially favor uptake of:


Occurrence of endocytosis

Fluid-phase (pinocytosis)

Soluble or nanoparticulate probes

Adsorptive

Probes bound to plasma membrane

Phagocytosis

Microparticulate probes

Occurrence of exocytosis

Secretion and synaptic activity

Probes lost from cell if incubation prolonged

Presence of known transporters, pumps or channels

MDR pump in tumor cells

Some probes (e.g., rhodamine 123) are removed from the cell

Glutamate transporter in glia

Some probes (e.g., SR101) are selectively taken into the cell

Cell division

Stage of division during meiosis or mitosis

Probe-DNA binding may vary with stage

Unusually fluid plasma membrane

In sperm, following capacitation

Biomodification of a probe

In differentiating erythrocytes and in some tumor cell lines

Uptake of certain probes into the membranes is favored

De-esterification of AM ester probes in various organelles

See Appendix 2 for discussion of this topic

self-generated singlet oxygen is possible. Because this would probably yield colorless compounds, fading would not confuse discussion of probe localization. Because each of the four porphyrin dyes possesses potentially protonatable ring N atoms (Falk and Phillips 1964, Phillips 1960, 1963), several chemical species may have been present. Therefore, the two compounds with electron-rich conjugated systems, TDFPPM and TDFPPS, are expected by similarity with analogous compounds to have pKa values for ring Ns in the range 4–5. Consequently, dyes in the incubation media and in the cytosol would be mostly deprotonated, whereas species within low pH organelles, such as lysosomes, would be mostly protonated. The resulting overall electric charges are controlled by the protonation equilibria, the non-ionic character of the other components of TDFPPM, and the four sulfonic acid substituents of TDFPPS; the latter always are present as anions under physiological conditions. TMPyP, however, is an electron-deficient porphyrin and for this reason the pKa values for protonation of its ring N atoms are lower (approx imately 1.4 and 0.6; tabulation in Kano 2004). Consequently, this dye exists predominantly as a single species under physiological conditions,

i.e., as a tetracation owing to the presence of four quaternary nitrogen groups. The pKa values for ring N atoms of the chlorin dye should, based on comparison with related compounds (Cavaleiro et al. 1984), be near 4–5 and 3. Therefore, the dye in incubation media and cytosol would be mostly deprotonated, whereas the dye in acidic organelles would be mostly protonated. The resulting overall charges on the various species of these four compounds are listed in Table 7. Also included in Table 7 are other relevant structure parameters for the dyes that were estimated as described earlier in the present paper.

Specifying protocol factors Prior to microscopic inspection, all four dyes were applied to cells by adding them to the incubation media under normal physiological conditions. Initial exposure to dyes was in the dark for extended periods (24 h) using no permeabilization procedure. Effects of potentially toxic solvents (D2O, ethanol) and of fetal calf serum were assessed, and considered not to influence probe localization. Subsequent illumination, which permitted microscopic observation of fluorescence, was for varying periods from short to extended.


Table 7. Estimated structure parameters, plus predicted and observed localizations, for the various species of the sensitizer dyes expected to be present Dye

Z

Log P

AI

CBN

Chlorin

0

3.4

2.4

25

1.5

2.4

25

6.0

3.5

2

4.0

4


TDFPPS

TMPyP

Cytoplasmic granules; not in nucleus

64

Biomembranes, especially Golgi complex

Cytoplasmic granules, perinuclear cytoplasm; not in nucleus

3.5

64

Lysosomes

8.5

3.5

64

Lysosomal; if membrane damaged, nucleus

No observations reported

2

8.5

3.5

64

4

13.5

NA

56

Lysosomal; if membrane damaged & incubation prolonged: nucleus, ribosomes, & possible leakage from the cell

Initially lysosomal then cytoplasmic “haze” plus nucleus & leakage from the cell

0

Specifying cell factors Monolayers of HeLa cells were used with all dyes. Photodamage to membranes with significant exposure to dyes (plasma membranes, and in some cases endosome/lysosome membranes) is more likely with longer illumination times. Given the cell line and temperature used, uptake by pinocytosis was possible. What are the predicted accumulation sites and mechanisms? The values of the numerical structure parameters for each chemical species listed in Table 7 were inserted into the QSAR models summarized in Part 1 to predict localizations. If necessary, the process was modulated by considering the protocol and cell thumbnails given above. • For chlorin, the models predict that the neutral species would enter cells by passive diffusion and would have access to all cell structures. The hydrophilic protonated species then would be ion-trapped in low pH structures such as lysosomes. This prediction is consistent with the observed localization within cytoplasmic granules. • For the strongly lipophilic neutral species of DFPPM, the models predict that it would enter by passive diffusion and have access to all organelles within the cell. The dye would accumulate in, though not be trapped by, biomembranes, particularly those of the Golgi complex. The hydrophilic, fully protonated species, however, would be ion-trapped in low pH 474

Observed localization

Lysosomes

1 TDFPPM

Predicted localization

structures. This prediction is consistent with the two types of fluorescent structures observed: a large perinuclear region and cytoplasmic granules. • For the sulfonated dye, TDFPPS, the hydrophilic deprotonated species is predicted by the QSAR models to bind to plasma membrane proteins and could be internalized by adsorptive endocytosis. If localization in the endo-/ lysosome membranes resulted, extended illumination and subsequent incubation might result in membrane damage and release of the dye into the cytosol where it would bind to proteins. Extended incubation, however, could permit slow diffusion to the site of the most basic cell proteins, the nuclear histones. No information on localization was given in the papers cited. • For the hydrophilic cationic quaternary salt, TMPyP, the QSAR models predict that uptake by pinocytosis is possible, which could result in endo-/lysosomal localization. Owing to its extreme hydrophilicity, the dye would not bind to proteins and continuing illumination could result in damage and permeabilization of the vesicle membrane, which would release dye into the cytosol. The dye then could diffuse into all non-membrane delimited cell regions. If the cell is illuminated, it could result in photodamage to membranes, which would release dye from the cell. In addition, the literature suggests (GlavašObrovac 2009, Marzilli et al. 1992) that TMPyP existing free in the cytosol could interact with both DNA and RNA by varying binding modes.



These outcomes are consistent with the observed loss of TMPyP from the cell into the incubation medium, and staining of both nucleus and cytoplasm. Predicted intracellular localizations of probes agree with those of the experimentalists in one case and in two other cases suggest precise accumulation sites consistent with reported observations. The prediction for TDFPPS could not be evaluated, because no observations for this compound have been reported. In terms of mechanism, the QSAR predictions confirm the experimentalists’ suggestions in one case, and in other cases make new and/or more detailed suggestions. Overall, therefore, these four dyes demonstrate that QSAR modeling of dye uptake and localization can provide a value added approach to experimental studies that involve intracellular localization of dyes, probes, or indeed, of other xenobiotics.

Acknowledgment RWH thanks Dr. R. Aitken, School of Life Sciences, College of Medical, Veterinary and Life Sciences, University of Glasgow, for providing facilities. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.

References Avdeef A (2003) Adsorption and Drug Development. Solubility, Permeability and Charge State. Wiley-Interscience, Hoboken NJ. Chapter 6, sections 4.0, 4.9, 3.7. Baughman GL, Banerjee S, Perenich TA (1996) Dye solubility. In: Advances in Color Chemistry. vol. 4. Physicochemical Principles of Color Chemistry. Blackie, London. pp. 145–195. Cavaleiro JAS, Jackson AH, Ali SA (1984) The protonation of chlorins (dihdyroporphyrins). Tetrahed. Lett. 25: 229–232. Chenzhong L, Hicklaus MC (2009) Comparisons of nine programs predicting pKa values of pharmaceutical substances. J. Chem. Inf. Model. 49: 2801–2812. Dapson RW (2013) Alternative methods for estimating common descriptors for QSAR studies of dyes and fluorescent probes using molecular modeling software. 1. Concepts and procedures. Biotech. & Histochem. 88: 477–488. Dapson RW, Horobin RW (2013) Alternative methods for estimating common descriptors for QSAR studies of dyes and fluorescent probes using molecular modeling software. 2. Correlations between log P and the

hydrophilic/lipophilic index, and new methods for estimating degrees of amphiphilicity. Biotech. & Histochem. 88: 489–497. da Silva EFF, Pedersen BW, Breitenbach T, Toftegaard R, Kuimova MK, Arnaut LG, Ogilby PR (2011) Irradiation- and sensitizer-dependent changes in the lifetime of intracellular singlet oxygen produced in a photosensitized process. J. Phys. Chem. B 116: 445–461. Del Castillo P, Horobin RW, Blázquez-Castro A, Stockert JC (2010) Binding of cationic dyes to DNA: distinguishing intercalation and groove binding mechanisms by use of simple experimental and numerical models. Biotech. & Histochem. 85: 247–256. Falk JE, Phillips JN (1963) Physical and coordination chemistry of the tetrapyrrole pigments. In: Chelating Agents and Metal Chelates. Dwyer FP, Mellor DP, Eds. Academic Press, New York. pp. 441–490. Franco A, Trapp S (2008) Estimation of the soil-water partition coefficient normalized to organic carbon for ionizable organic chemicals. Environ. Toxicol. Chem. 27: 1995–2004. Glavaš-Obrovac L, Piantanida I, Marczi S, Mašic L, Timcheva II, Deligeorgiev TG (2009) Minor structural differences of monomethine cyanine derivatives yield strong variation in their interactions with DNA, RNA as well as on their in vitro antiproliferative activity. Bioorg. Med. Chem. 17: 4747–4755. Green FJ (1990) The Sigma-Aldrich Handbook of Stains, Dyes and Indicators. Aldrich Chemical Company, Milwaukee WI. Hansch C, Leo A (1979) Substituent Constants for Correlation Analysis in Chemistry and Biology. Wiley-Interscience, New York. Chapter 4, pp. 18–43. Haugland RP (1996) Handbook of Fluorescent Probes and Research Chemicals, 6th ed. Molecular Probes, Eugene OR. Horobin RW, Kiernan JA (2002) Conn’s Biological Stains: a Handbook of Dyes, Stains and Fluorochromes for Use in Biology and Medicine, 10th ed. BIOS Scientific Publishers, Oxford. Horobin RW, Rashid-Doubell F, Pediani JD, Milligan G (2013a) Predicting small-molecule fluorescent probe localization in living cells using QSAR modeling. 1. Overview and models for probes of structure, properties and function in single cells. Biotech. & Histochem. 88: 440–460. Horobin RW, Stockert JC, Rashid-Doubell F (2013b) Uptake and localisation of small-molecule fluorescent probes in living cells: a critical appraisal of QSAR models and a case study concerning probes for DNA and RNA. Histochem. Cell Biol. 139: 623–637. DOI: 10.1007/s00418013-1090-0. Kano K (2004) Molecular complexes of water-soluble porphyrins. J. Porphyr. Phthalocya. 8: 148–155. Kuimova MK (2012) Mapping viscosity in cells using molecular rotors. Phys. Chem. Chem. Phys. 14: 12671–12686. Kuimova MK, Yahioglu G, Ogilby PR (2009) Singlet oxygen in a cell: spatially dependent lifetimes and quenching rate constants. J. Am. Chem. Soc. 131: 332–340. Leo A, Hansch C, Elkins D (1971) Partition coefficients and their uses. Chem. Rev. 71: 525–616.



Lillie RD (1977) Conn’s Biological Stains: a Handbook on the Nature and Uses of the Dyes Employed in the Biological Laboratory, 9th ed., Williams and Wilkins, Baltimore. Llopis J, McCaffery JM, Miyawaki A, Farquhar MG, Tsien RY (1998) Measurement of cytosolic, mitochondrial, and Golgi pHi in single living cells with green fluorescent proteins. Proc. Natl. Acad. Sci. USA 95: 6803–6808. Loiselle FB, Casey JR (2010) Measurement of intracellular pH. Methods Mol. Biol. 637: 311–331. Mannhold R, Poda GI, Osermann C, Tetko IV (2009) Calculation of molecular lipophilicity: state-of-the-art and comparison of log P methods on more than 96,000 compounds. J. Pharm. Sci. 98: 861–893. Marzilli LG, Peth G, Lin M, Kim MS, Dixon DW (1992) Tentacle porphyrins: DNA interactions. J. Am. Chem. Soc. 114: 7575–7577. Mindell JA (2012) Lysosomal acidification mechanisms. Ann. Rev. Physiol. 74: 69–86. Müller W, Crothers DM, Waring MJ (1973) A nonintercalating proflavine derivative. Eur. J. Biochem. 39: 223–234. Perrin DD, Dempsey B, Serjeant EP (1981) pKa Prediction for Organic Acids and Bases. Chapman and Hall, London. p. 17.

Phillips JN (1960) The ionization and coordination behaviour of porphyrins. Rev. Pure Appl. Chem. (Australia) 10: 35–60. Phillips JN (1963) Physicochemical properties of porphyrins. In: Comprehensive Biochemistry, vol. 9, Florkin M, Stotz EH, Eds., Elsevier, Amsterdam. pp. 34–72. Rivinoja A, Pujol FM, Hassinen A, Kellokumpu S (2012) Golgi pH, its regulation and roles in human disease. Ann. Med. 44: 542–554. Rost F (1995) Fluorescence Microscopy, vol. 2. Cambridge University Press, Cambridge. Seksek O, Bolard J (1996) Nuclear pH gradient in mammalian cells revealed by laser microspectrofluorimetry. J. Cell Sci. 109 (Part 1): 257–262. Smith RM, Martell AE (1974) Critical Stability Constants, vols. 1–6. Plenum Press, New York. Smith RM, Martell AE, Motekaitis RJ (2012) NIST Critically Selected Stability Constants of Metal Complexes Database, Version 8.0, NIST standard reference database 46, US Department of Commerce, Gaithersburg, MD. Takahashi A, Zhang Y, Centonze E, Herman B (2001) Measurement of mitochondrial pH in situ. Biotechniques 30: 804–810, 812.

Supplementary material available online Appendices 1–5 to be found online at http:// i n f o r m a h e a l t h c a re . c o m / d o i / a b s / 1 0 . 3 1 0 9 / 10520295.2013.780635

476
