Comparison of Human and Canine P‐Glycoprotein Substrates at R, P, and H Binding Sites

TL;DR

This study compares human and canine P-glycoprotein drug interactions, finding that human data often does not accurately predict canine behavior.

Contribution

The study reveals that human P-glycoprotein substrate data is frequently inaccurate for predicting canine P-glycoprotein substrate status.

Findings

MRP1 has minimal impact on calcein AM efflux in cells with canine or human P-gp.

Identifying a drug's P-gp binding site can provide clinically relevant information.

Human P-gp substrate data is often inaccurate as a proxy for canine P-gp substrate data.

Abstract

P‐glycoprotein (P‐gp) greatly impacts substrate drug disposition, so much so that regulatory agencies recommend ascertaining the P‐gp status of active pharmaceutical ingredients (APIs) intended for human use. Arguably, the P‐gp status of drugs intended for canine patients is equally, if not more, important. Our research objectives were to assess whether human P‐gp substrate data can predict canine P‐gp substrate status and to explore the three previously reported binding sites within the P‐gp binding pocket, the H‐, R‐, and P‐sites. Competitive efflux assays employing cell lines expressing canine or human P‐gp were used to compare the degree of overlap or independence of the three binding sites in canine versus human P‐gp using site‐specific fluorescent P‐gp substrates rhodamine 123, calcein AM and Hoechst 33342. Because calcein AM can also be transported by multidrug resistance protein 1 (MRP1), experiments were performed to assess its potential influence on calcein AM efflux studies. Results indicate that: (i) MRP1 is either a non‐factor or negligible factor for cells expressing canine or human P‐gp respectively; (ii) determining an API's P‐gp binding site may provide clinically relevant information; and (iii) use of human P‐gp substrate data as a proxy for canine P‐gp substrate data will often prove inaccurate.

Figures

Western blots assessing MRP1 (~190 kDa; top left panel) and P‐gp (~170 kDa; top right panel) expression in various cell lines. Beta actin (~42 kDa; middle panels) was used as a loading control. Molecular weight markers (MW: 250, 150, 50, 37 kDa) are indicated along the left side of the respective images. Normalized MRP1 and P‐gp expression are indicated in the bottom panels. HeLa cells are the positive control cell line for MRP1 expression (see Note 2); K9KO are MDCK cells with canine P‐gp knocked out (Karlgren et al. 2017), MDCK cells, Madin Darby canine kidney cell line; hMDR1 over‐expresses human P‐gp (Karlgren et al. 2017); K9PGP cells overexpress canine P‐gp (Mealey et al. 2017).

Representative histograms representing flow cytometric analysis of a single well of either hMDR1 cells (A–C) or MDCKPGP cells (D–F) treated with the following fluorescent (“probe”) and non‐fluorescent P‐gp substrate combinations to confirm binding to their respective P‐gp binding sites: Rhodamine 123 and cyclosporine A for the R‐site; Hoechst 33342 and vinblastine for the H‐site; calcein AM and progesterone for the P‐site. Additional wells containing the negative control cisplatin and positive control PSC 833 were also evaluated. The x‐axis represents fluorescence intensity, and the y‐axis represents cell number. Experiments were run in triplicate and repeated on a separate day.

Bar chart of (MFI [experimental drug + fluorescent P‐gp substrate]:MFI [fluorescent P‐gp substrate alone]) ratios for human (x‐axis) and canine (y‐axis) P‐gp at the R‐site (rhodamine 123, left panel) and P‐site (calcein AM, right panel) and standard deviations. Ratios of 1 indicate the compound is not a P‐gp substrate, ratios > 2 indicate the drug is a P‐gp substrate, and ratios > ~5 for calcein AM and > ~10 for rhodamine 123 indicate strong P‐gp substrate status. CIS, cisplatin (10 μM); CYC, cyclosporine A (1 μM); DIG, digoxin (1000 μM); IVM, ivermectin (800 μM); LOP, loperamide (10 μM); PRO, progesterone (10 μM); PSC, PSC 833 (1 μM); VIN, vincristine (50 μM). CIS and PSC are the negative and positive controls, respectively. Experiments were run in triplicate and repeated on a separate day. *Indicates statistically significant differences in MFI ratios between canine and human P‐gp.

Suggested functional binding relationship of R‐, H‐, and P‐sites within the canine P‐gp binding pocket based on data generated in the competitive P‐gp efflux assay.

Tables

TABLE 1. Comparison of amino acids comprising the transmembrane domain of human and canine P‐glycoprotein.

Transmembrane domain	Human sequence (amino acid residue) canine sequence	Number (%) amino acid differences
1	(52) VVGTLAAIIHGAGLPLMMLVFGEMTDI (78) LVGTMAAIIHGAALPLMMLVFGNMTDS	4 (14.8%)
2	(112) TRYAYYYSGIGAGVLVAAYIQVSFWCLA (139) TTYAYYYSGIGAGVLVAAYIQVSFWCLA	1 (3.6%)
3	(186) IGDKIGMFFQSMATFFTGFIV (206) IGDKVGMFFQSIATFFTGFIV	2 (9.5%)
4	(213) KLTLVILAISPVLGLSAAVWAKILSSFTDK (242) KLTLVILAISPVLGLSAAIWAKILSSFTDK	1 (5.0%)
5	(291) KAITANISIGAAFLLIYASYALAFWYGTTLV (321) KAITANISIGAAFLLIYASYALAFWYGTSLV	1 (3.2%)
6	(329) GQVLTVFFSVLIGAFSVGQASPSIEAFA (356) GQVLTVFFSVLIGAFSIGQASPDIEAFA	1 (3.6%)
7	(712) VVGVFCAIINGGLQPAFAIIFSKIIGV (738) VVGIFCAIINGGLQPAFSIIFSRIIGI	4 (14.8%)
8	(754) LFSLLFLALGIISFITFFLQGFTFGKA (780) MFSVLFLVLGIISFITFFLQGFTFGKA	3 (11.1%)
9	(829) IGSRLAVITQNIANLGTIII (849) IGSRLAVITQNIANLGTIII	0 (0%)
10	(856) QLTLLLLAIVPIIAIAGVVEMKMLSGQALK (885) QLTLLLLAIVPIIAIAGVVEMKMLSGQALK	0 (0%)
11	(934) KAHIFGITFSFTQAMMYFSYAGCFRFGAYLV (964) KAHIFGVSFSITQAMMYFSYAGCFRFGAYLV	3 (9.7%)
12	(972) EDVLLVFSAVVFGAMAVGQVSSFAPDYA (999) QDVLLVFSAIVFGAMAVGQSSFAPDYA	2 (7.1%)

TABLE 2. Fluorescent and non‐fluorescent P‐gp substrates examined, the concentrations at which they were studied, and the rationale for their inclusion.

P‐gp substrate	Concentrations (μM)	Rationale for inclusion
Calcein AM	0.01	Fluorescent P‐site P‐gp substrate
Hoechst 33342	1	Fluorescent H‐site P‐gp substrate
Rhodamine 123	5	Fluorescent R‐site P‐gp substrate
Cisplatin	10, 20, 50	Non P‐gp substrate (− control)
Cyclosporine A	0.01, 0.1, 1	Non‐fluorescent R‐site substrate (human P‐gp) a
Vinblastine	1, 5, 10, 20	Non‐fluorescent H‐site substrate (human P‐gp)
Progesterone	1, 5, 10	Non‐fluorescent P‐site substrate (human P‐gp)
Digoxin	10, 100, 1000	Human P‐gp substrate (unknown site; unknown canine P‐gp status)
Ivermectin	0.008, 0.08, 0.8	Canine and human P‐gp substrate‐unknown site
Loperamide	10	Canine and human P‐gp substrate‐unknown site
PSC‐833	1	“Classic” P‐gp inhibitor (+ control)‐unknown site
Vincristine	12.5, 25, 50	Canine P‐gp substrate‐unknown site

TABLE 3. Mean fluorescence intensity (MFI) ratios (MFI [experimental drug + fluorescent P‐gp substrate]:MFI [fluorescent P‐gp substrate alone]) and standard deviations of human (hMDR1) and canine (MDCKPGP) P‐gp after incubation with site‐specific probes.

P‐gp binding site	Fluorescent–non‐fluorescent drug combination	hMDR1 cells	MDCKPGP cells
R‐site	Rhodamine 123 + cyclosporine A	11 (2.1)	43 (3.3)
R‐site (− control)	Rhodamine 123 + cisplatin	1 (0.1)	1 (0.1)
P‐gp inhibitor (+ control)	Rhodamine 123 + PSC 833	44 (4.1)	52 (2.3)
H‐site	Hoechst 33342 + vinblastine	2 (0.2)	2 (0.3)
H‐site (− control)	Hoechst 33342 + cisplatin	1 (0.02)	1 (0.1)
P‐gp inhibitor (+ control)	Hoechst 33342 + PSC 833	3 (0.2)	3 (0.1)
P‐site	Calcein AM + progesterone	5 (0.7)	6 (0.5)
P‐site (− control)	Calcein AM + cisplatin	1 (0.03)	1 (0.04)
P‐gp inhibitor (+ control)	Calcein AM + PSC 833	12 (1.0)	8 (0.9)

TABLE 4. Mean fluorescence intensity (MFI) ratios (MFI [experimental drug + CFDA):MFI [CFDA]) of human (hMDR1) and canine (MDCK‐PGP) cells.

CFDA (10 μM)+	hMDR1 cells	MDCK‐PGP cells
MK 571 (100 μM)	9.3	37
Cisplatin (10 μM)	0.9	1.0
Cyclosporine A (1 μM)	1.2	1.1
Vinblastine (20 μM)	1.5	1.6
Progesterone (10 μM)	1.0	1.0
Digoxin (1000 μM)	1.1	1.5
Ivermectin (0.8 μM)	1.1	1.2
Loperamide (10 μM)	1.5	1.4
PSC‐833 (1 μM)	1.1	1.0
Vincristine (50 μM)	1.6	1.3

Full text

Introduction

1

Regulatory agencies such as the U.S. Food and Drug Administration and the European Medicines Agency strongly recommend assessing active pharmaceutical ingredients (APIs) intended for human use as P‐glycoprotein (P‐gp) substrates during the drug development process (Wang and Sauerwald 2014). The primary rationale is to prevent drug–drug interactions (i.e., acquired P‐gp deficiency) associated with co‐administration of P‐gp substrate drugs. Assessment of drugs as canine P‐gp substrates is clinically important for preventing adverse drug reactions in dogs with either intrinsic or acquired P‐gp deficiency. Intrinsic P‐gp deficiency occurs in dogs with ABCB1‐1Δ, a 4‐bp deletion mutation in the ABCB1 or MDR1 gene (Mealey et al. 2001). Acquired P‐gp deficiency can affect any dog, even those with normal MDR1 genetic make‐up. Co‐administration of two P‐gp substrate drugs can result in competition for P‐gp‐mediated efflux at the blood–brain barrier, biliary canalicular cells, T‐lymphocyte cytoplasmic membrane, or other sites where P‐gp is expressed. The fact that serious adverse drug reactions can occur in dogs with P‐gp deficiency has been well documented (Campbell et al. 2017; Deshpande et al. 2016; Mealey and Fidel 2015; Mealey et al. 2008, 2022).

P‐gp has been described as promiscuous with respect to its substrate binding capacity. More than 200 structurally diverse compounds are substrates for human P‐gp (Subramanian et al. 2016). Although not fully understood, one of the reasons human P‐gp is able to bind and transport such diverse molecules is that its tertiary structure includes a binding pocket or cavity rather than a single, distinct binding site (Chufan et al. 2015; Loo et al. 2006). Within this cavity, three overlapping substrate interaction sites have been proposed, dubbed the R‐, P‐, and H‐site, due to their ability to bind Rhodamine 123, Progesterone, and Hoechst 33342, respectively (Shapiro and Ling 1998; Shapiro et al. 1999; Chufan et al. 2015). These binding sites should be considered more functional than structural since there is not a consensus on the specific amino acid residues involved.

One method of assessing drugs as P‐gp substrates involves administering the test compound to animals such as dogs homozygous for the MDR1 mutation or knockout mouse models (Swain et al. 2013; Paul et al. 2004). An obvious disadvantage of this approach is that it stands in contrast to the principles of humane experimental technique espoused by the USDA,1 namely the 3R's (replacement, reduction, refinement) with respect to using animals in research. Another disadvantage of live animal studies is their expense relative to in vitro studies. The study reported here represents a preliminary step toward fully replacing the use of dogs or mice for determining if a drug is a canine P‐gp substrate. In addition to the ethical advantage, the cost of cellular‐based efflux studies is substantially less than that of animal studies.

There are two general types of cell‐based methods to assess P‐gp substrate status. The first is what could be considered a direct method whereby bidirectional transport of the test compound is assessed using a P‐gp expressing cell line grown in transwells (Yabut et al. 2022). The test compound is applied to either side of the transwell and its concentration is measured over time by using either radiolabeled test compound or by developing an assay for quantitating the test compound. The second method could be considered an indirect method, consisting of the test compound competing with a fluorescent P‐gp substrate for efflux. The magnitude of increase in intracellular fluorescence when cells are incubated with the test compound and the fluorescent probe compared to that when cells are incubated with the fluorescent probe alone is used to predict its P‐gp substrate status. A potential advantage of the indirect method is that drug–drug interactions involving P‐gp efflux may be more accurately predicted because the fluorescent P‐gp substrates bind to specific P‐gp binding sites (H‐, R‐, or P‐site). For example, an exclusive R‐site substrate may not cause drug–drug interactions with an exclusive P‐site substrate if concurrently administered even though both are P‐gp substrates.

While one might postulate that P‐gp substrate data generated from human studies could be applied to canine P‐gp, several studies have indicated species differences in P‐gp substrate status (Kido et al. 2022; Katoh et al. 2006; Yamazaki et al. 2001; Zolnerciks et al. 2011). The amino acid sequence of the canine and human P‐gp binding pockets, located within the transmembrane domains, is not identical (Table 1); therefore it seems likely that species differences could exist in P‐gp substrate status, and are worthy of further exploration.

Our primary objective was to perform an initial comparison of the human and canine P‐gp binding pocket (H‐, R‐, and P‐binding sites) using intrinsically fluorescent P‐gp substrates that interact specifically with those respective sites. Our goal was to determine: (i) if there is evidence of H, R and/or P binding sites for canine P‐gp; (ii) if there is evidence of overlap between these sites; and (iii) whether substrate binding is quantitatively and qualitatively similar between canine and human P‐gp.

Materials and Methods

2

Drugs, Fluorescent P‐Glycoprotein Substrates, Cell Lines and Culture Conditions

2.1

Previously characterized cell lines expressing canine (MDCK‐PGP; Mealey et al. 2017) or human (hMDR1‐MDCK; Karlgren et al. 2017) P‐gp were used in these studies, the former being established in the authors' laboratory and the latter a generous gift from the Karlgren laboratory. HeLa cells, used as a positive control for multidrug resistance protein 1 (MRP1)2 were obtained from American Type Culture Collection (Manassas, VA). Cells were discarded after about 4 weeks in culture, which reflects approximately 12 passages. Fluorescent P‐gp substrates rhodamine 123 and calcein AM were purchased from Invitrogen (Carlsbad, CA), and Hoechst 33342 was purchased from Biosciences (Minneapolis, MN). The non‐P‐gp substrate cisplatin and the P‐gp substrates digoxin, cyclosporine A, ivermectin, loperamide, progesterone, PSC‐833, and vinblastine were purchased from Tocris Biosciences (Minneapolis, MN). The rationale for their inclusion is indicated in Table 2. Drug concentrations used in the assays described below are based on concentration ranges used in transport assays using human or rodent cell‐based efflux studies that have demonstrated their status as P‐gp substrates. The fluorescent multidrug resistance protein 1 (MRP1) substrate 5(6)‐carboxyfluorescein diacetate (CFDA) was purchased from Thermo Fisher (Eugene, OR) and the MRP1 specific inhibitor MK 571 was obtained from Tocris Bioscience (Minneapolis, MN). Cells were maintained in Dulbecco's modified Eagle medium (DMEM) containing 10% fetal bovine serum within an incubator tightly regulated at 37°C and 5% CO_2_ concentration.

Western Blot Assessment of P‐gp, MRP1, and β‐Actin Expression

2.2

Expression of P‐gp, MRP1, and the loading control β‐actin in all experimental cell lines was assessed via western blot. Total protein concentrations for whole cell lysates were determined using a Pierce BCA protein assay kit (Thermo Scientific, Rockford, IL). Twenty micrograms of total protein from each cell lysate were separated on 4%–15% Tris–HCl gradient polyacrylamide gels (Invitrogen, Carlsbad, CA) and electrotransferred to polyvinylidene difluoride membrane. The membrane was incubated in blocking buffer (5% fish gelatin in Tris buffered saline [TBS BioWorld, Dublin, OH]) for 1 h at room temperature, then incubated with primary antibody in 5% fish gelatin in TBS with 0.05% Tween 20 overnight at 4°C. For detection of MRP1, rat mAB MRPr‐1 antibody (Abcam, Cambridge, MA) was used at a dilution of 1:100 and for detection of P‐gp SN06‐42 (Invitrogen, Carlsbad, CA; Cat # MA5‐32282) was used at a dilution of 1:2000. Membranes were washed in TBS with 0.05% Tween 20 and incubated with 1:2500 horseradish peroxidase‐conjugated secondary antibody (Thermo Pierce, Rockford, IL) for 1 h at 4°C. After a final wash, secondary antibody binding was detected using Immobilon Western HRP substrate (Millipore, Burlington, MA). For β‐actin, the P‐gp and the MRP1 probed membranes were stripped for 20 min at room temperature with Restore Western Blot Stripping Buffer (Thermo, Rockford, IL). Membranes were rinsed in deionized water and equilibrated in TBS before reblocking. Membrane blocking was accomplished as described earlier, then incubated with rabbit anti b‐actin polyclonal antibody (ab16039; Abcam, Cambridge, MA) at 1:1000 for 1 h at room temperature in 5% fish gelatin in TBS with 0.05% Tween 20. Membranes were then washed and incubated with anti‐rabbit horseradish peroxidase‐conjugated secondary antibody (Thermo Pierce, Rockford, IL) in 5% fish gelatin in TBS with 0.05% Tween 20 for 1 h at room temperature. Bots were imaged on a BioRad ChemiDoc MP (BioRad, Hercules, CA). Band density was determined using the BioRad Image Labv4.1 quantification module.

Assessment of Rhodamine 123, Hoechst 33342 and Calcein AM as Fluorescent Substrates for Putative Canine P‐gp R‐, H‐, and P‐Site, Respectively

2.3

To determine if the canine P‐gp binding pocket consists of three overlapping binding sites (R‐, H‐, and P‐sites) as described for human P‐gp, we exploited the intrinsic fluorescence of three P‐gp substrates (rhodamine 123, Hoechst 33342, and calcein AM) that have previously been shown to interact with the R‐, H‐ and P‐binding sites, respectively. All experiments included the positive control, PSC‐833 and the negative control, cisplatin. Intracellular concentrations of each of these P‐gp substrates can be quantified using flow cytometry. We conducted previously described (Mealey and Burke 2023) competitive efflux assays whereby non‐fluorescent P‐gp substrates that have been demonstrated to bind to the R‐ (cyclosporine A; Shapiro and Ling 1998; Shapiro et al. 1999; Chufan et al. 2015), H‐ (vinblastine; Shapiro and Ling 1998) or P‐site (progesterone; Shapiro et al. 1999; Eneroth et al. 2001; Fröhlich et al. 2004) were co‐incubated with the corresponding fluorescent P‐gp substrate. Briefly, 8 × 10^4^ hMDR1‐MDCK or MDCK‐PGP cells were incubated overnight in 12‐well plates. Prior to seeding, cells underwent counting and trypan blue viability assessment using a Corning Cell Counter with Axion Biosystems Axis Vue analysis software (Axion Biosystems, Atlanta, GA). Media containing the pre‐diluted fluorescent probe (rhodamine 123, Hoechst 33342, or calcein AM) alone or the fluorescent probe plus the corresponding, site‐specific P‐gp substrate were added to wells as a media change at the concentrations indicated in Table 2. Plates were incubated for 1.5 h under standard cell culture conditions (described above). After incubation, the media was aspirated off, cells were trypsinized and washed with ice cold Hank's balanced salt solution (HBSS). Cells were centrifuged in a pre‐chilled rotor (5 min at 150 g), the pellet was resuspended in 300 μL ice cold HBSS and subjected to flow cytometric analysis. For experiments using rhodamine 123 and calcein AM, cells were analyzed on a Guava easyCyte flow cytometer (Cytek Biosciences, Fremont, CA) with a 50 mW blue laser set at 488 nm. A minimum of 10,000 events were collected for each well. The fluorescence emission of rhodamine 123 and calcein AM was collected in the green channel with a 525/30 bandpass filter. For experiments using Hoechst 33342, a separate instrument (Guava easyCyte HT; EMD Millipore Corporation, Hayward, CA) was used with a 100 mW violet laser (405 nM) and the fluorescence emission of Hoechst 33342 collected in the blue channel with a 450/45 bandpass filter. Triplicate wells of each drug combination and of the fluorescent probe alone were analyzed, and all experiments were repeated on a separate day to ensure repeatability. The mean of the fluorescence intensity (MFI) of ≥ 10,000 cells analyzed from each well was used to calculate the ratio of fluorescence intensity in cells treated with fluorescent P‐gp substrate and experimental drug to that of cells treated with fluorescent P‐gp substrate only (MFI [experimental drug + fluorescent P‐gp substrate]:MFI [fluorescent P‐gp substrate alone]), hereafter referred to as MFI ratio. MFI ratios of ~1 indicate that no competition is occurring between the fluorescent and non‐fluorescent P‐gp substrates while MFI ratios > 1 indicate competition for a particular P‐gp binding site. MFI ratios generated from the experiments using canine versus human P‐gp were not compared statistically, but differences greater than twofold are specifically called out. Comparisons between canine and human P‐gp status were performed with a Bonferroni adjusted t‐test.

Assessment of Drugs as Canine and/or Human P‐gp Substrates

2.4

Cisplatin (negative control), cyclosporine A, digoxin, ivermectin, loperamide, progesterone, PSC‐388 (positive control), and vincristine were assessed as canine and human P‐gp substrates in competitive efflux assays with rhodamine 123 and with calcein AM to assess the R‐ and P‐site, respectively. These drugs are known to be substrates of canine and/or human P‐gp (Table 2) and were selected to identify any qualitative or quantitative species differences. Using the same competitive efflux assay described above, the MFI ratio was determined for each experimental drug after incubation with both hMDR1‐MDCK and MDCK‐PGP cells.

Further Assessment of P‐gp H‐Site

2.5

Because previous studies (Shapiro et al. 1999) identified overlap between the human H‐site and the R‐site, further studies involving the fluorescent (Hoechst 33342) and non‐fluorescent (vinblastine) H‐site probes were conducted. First, vinblastine was tested as a competitive inhibitor of the R‐site probe rhodamine 123 in both human and canine P‐gp expressing cell lines using methods described above. Second, Hoechst 33342 itself was tested as a competitive inhibitor of rhodamine 123 and calcein AM to determine the degree of overlap between the H‐site and R‐site, and the H‐site and P‐site, respectively.

Assessment of Experimental Drugs as MRP1 Substrates

2.6

To determine if experimental drugs were substrates for human and canine MRP1, the fluorescent MRP1 substrate CFDA (Poźniak et al. 2015) was used in competitive efflux assays with and without the MRP1 inhibitor MK 571 (Echevarria‐Lima et al. 2005). Briefly, hMDR1‐MDCK or MDCKPGP cells were incubated overnight in 12‐well plates. Media containing CFDA alone (10 μM), or CFDA plus the MRP1 inhibitor MK571 (100 μM) as a positive control, and CFDA plus each experimental drug (at highest concentration listed in Table 2) were added to wells. Plates were incubated for 1.5 h under standard cell culture conditions (described above). After incubation, the media was aspirated off, cells were trypsinized and washed with ice cold HBSS. Cells were centrifuged as above, the pellet was resuspended in 300 μL ice cold HBSS and subjected to flow cytometric analysis. Cells were analyzed on a Guava easyCyte flow cytometer (Cytek Biosciences, Fremont, CA) with a 50 mW blue laser set at 488 nm. A minimum of 10,000 events were collected for each well. The fluorescence emission of CFDA was collected in the green channel with a 525/30 bandpass filter. Triplicate wells of each were analyzed, with all experiments repeated on at least one additional, separate day to ensure repeatability.

Results

3

Assessment of P‐gp and MRP1 Expression

3.1

Because MRP1 is an efflux transporter that may share some substrates with P‐gp (Laupeze et al. 1999), it was important to determine if hMDR1‐MDCK or MDCK‐PGP cells express MRP1. To determine MRP1's relative expression levels in hMDR1‐MDCK and MDCK‐PGP cells, western blots were performed. The HeLa cell line was used as the positive control for MRP1 expression.3 MRP1, at approximately 190 kDa, appears to be expressed by all cell lines assessed (Figure 1), with higher levels of expression in hMDR1 cells compared to MDCK‐PGP cells. P‐gp, at approximately 170 kDa, is robustly expressed by both hMDR1‐MDCK and MDCK‐PGP cell lines relative to the parental MDCK cells, MDCK knockout cells or HeLa cells There appears to be greater relative expression of P‐gp in hMDR1 cells than in MDCK‐PGP cells (Figure 1).

Western blots assessing MRP1 (~190 kDa; top left panel) and P‐gp (~170 kDa; top right panel) expression in various cell lines. Beta actin (~42 kDa; middle panels) was used as a loading control. Molecular weight markers (MW: 250, 150, 50, 37 kDa) are indicated along the left side of the respective images. Normalized MRP1 and P‐gp expression are indicated in the bottom panels. HeLa cells are the positive control cell line for MRP1 expression (see Note 2); K9KO are MDCK cells with canine P‐gp knocked out (Karlgren et al. 2017), MDCK cells, Madin Darby canine kidney cell line; hMDR1 over‐expresses human P‐gp (Karlgren et al. 2017); K9PGP cells overexpress canine P‐gp (Mealey et al. 2017).

Assessment of Rhodamine 123, Hoechst 33342 and Calcein AM as Fluorescent Substrates for Putative Canine P‐gp R‐, H‐, and P‐Site, Respectively

3.2

Using a competitive efflux assay employing fluorescent and corresponding non‐fluorescent P‐gp substrates for R‐, H‐, and P‐sites (Table 2), incubation with hMDR1‐MDCK cells confirmed that human P‐gp has the three previously reported binding sites (Table 3). Specifically, cyclosporine A competed with rhodamine 123 at the R‐site (Figure 2A), vinblastine competed with Hoechst 33342 at the H‐site (Figure 2B) and progesterone competed with calcein AM at the P‐site (Figure 2C).

Representative histograms representing flow cytometric analysis of a single well of either hMDR1 cells (A–C) or MDCKPGP cells (D–F) treated with the following fluorescent (“probe”) and non‐fluorescent P‐gp substrate combinations to confirm binding to their respective P‐gp binding sites: Rhodamine 123 and cyclosporine A for the R‐site; Hoechst 33342 and vinblastine for the H‐site; calcein AM and progesterone for the P‐site. Additional wells containing the negative control cisplatin and positive control PSC 833 were also evaluated. The x‐axis represents fluorescence intensity, and the y‐axis represents cell number. Experiments were run in triplicate and repeated on a separate day.

To determine if the canine P‐gp binding pocket contains these three binding sites as well, the same fluorescent and non‐fluorescent P‐gp substrate combinations were incubated with MDCK‐PGP cells. As was observed with cells expressing human P‐gp, cyclosporine A competed with rhodamine 123 at the presumptive canine R‐site (Figure 2D), vinblastine competed with Hoechst 33342 at the presumptive H‐site (Figure 2E) and progesterone competed with calcein AM at the presumptive P‐site of canine P‐gp (Figure 2F). However, while the qualitative results between human and canine P‐gp were the same, some ratios differed quantitatively. Cyclosporine A appears to be a stronger substrate for the canine P‐gp R‐site than for the human P‐gp R‐site with ratios of ~40 versus 12, respectively (Table 3). Progesterone and vinblastine exhibited similar substrate strength for human and canine P‐gp P‐ and H‐sites, respectively, as indicated by similar ratios (Table 3).

As can be gleaned from Table 3 there are substantial quantitative differences in the magnitude of MFI ratios achieved depending on the fluorescent P‐gp substrate used. This is best illustrated by comparing the ratios of each fluorescent P‐gp substrate with PSC 833, a drug specifically developed as a P‐gp inhibitor (Advani et al. 1999; Tai 2000). The highest MFI ratios, 44 and 52, are achieved when PSC 833 is co‐incubated with the fluorescent P‐gp substrate rhodamine 123 in hMDR1‐MDCK and MDCK‐PGP cells, respectively. Intermediate ratios, 12 and 8, are achieved when PSC 833 is co‐incubated with the fluorescent P‐gp substrate calcein AM in hMDR1‐MDCK and MDCK‐PGP cells, respectively. The MFI ratios achieved when PSC 833 was co‐incubated with Hoechst 33342 are low, 3 and 3, for hMDR1‐MDCK and MDCK‐PGP cells, respectively.

Assessment of Canine Versus Human P‐gp Substrate Status of Clinically Relevant Drugs via Competitive Efflux Assay

3.3

Figure 3 illustrates important characteristics of canine versus human P‐gp substrate status of experimental drugs. The first observation is that the R‐sites, indicated by competition with rhodamine 123, and P‐sites, indicated by competition with calcein AM, of both human and canine P‐gp bind some drugs in common and some drugs independently. Specifically, progesterone and digoxin have ratios of 1 for both canine and human cell lines when tested against the R‐site fluorescent substrate, rhodamine 123, indicating that neither drug is a substrate at the R‐sites of canine or human P‐gp. Conversely, both progesterone and digoxin have ratios > 1 when tested against the P‐site fluorescent substrate calcein AM (Figure 3), indicating that both drugs are substrates at the P‐site of canine and human P‐gp. Collectively, these results provide evidence of independent, non‐overlapping substrate interactions with the P‐site. Three different drugs, loperamide, ivermectin, and cyclosporine A provide evidence of overlapping substrate interactions involving both the P‐ and R‐site. Specifically, each of these drugs competes with both the R‐site fluorescent substrate rhodamine 123 and the P‐site fluorescent substrate calcein AM for canine P‐gp and human P‐gp mediated efflux as indicated by ratios of > 1 (Figure 3).

*Bar chart of (MFI [experimental drug + fluorescent P‐gp substrate]:MFI [fluorescent P‐gp substrate alone]) ratios for human (x‐axis) and canine (y‐axis) P‐gp at the R‐site (rhodamine 123, left panel) and P‐site (calcein AM, right panel) and standard deviations. Ratios of 1 indicate the compound is not a P‐gp substrate, ratios > 2 indicate the drug is a P‐gp substrate, and ratios > ~5 for calcein AM and > ~10 for rhodamine 123 indicate strong P‐gp substrate status. CIS, cisplatin (10 μM); CYC, cyclosporine A (1 μM); DIG, digoxin (1000 μM); IVM, ivermectin (800 μM); LOP, loperamide (10 μM); PRO, progesterone (10 μM); PSC, PSC 833 (1 μM); VIN, vincristine (50 μM). CIS and PSC are the negative and positive controls, respectively. Experiments were run in triplicate and repeated on a separate day. Indicates statistically significant differences in MFI ratios between canine and human P‐gp.

Another important observation is that quantitative differences in substrate status exist between human and canine P‐gp for several clinically important drugs (Figure 3). Cyclosporine A generated much higher MFI ratios when incubated with MDCK‐PGP cells than when incubated with hMDR1‐MDCK cells indicating they are stronger substrates for canine P‐gp than for human P‐gp at the R‐site, while ivermectin and loperamide are stronger substrates at the human P‐gp R site. Species differences in calcein AM MFI ratios suggest species differences between the human and canine P‐gp P site. Specifically, ivermectin and loperamide appear to be stronger substrates for human P‐gp while vincristine appears to be a stronger substrate for canine P‐gp. Digoxin and progesterone appear to be quite similar with respect to their substrate status for canine and human P‐gp.

Further Assessment of P‐gp H‐Site

3.4

Our data confirmed that vinblastine is a weak H‐site substrate for both human and canine P‐gp and also demonstrated that vinblastine competed with the R‐site probe rhodamine 123 indicating overlap of the H and R binding sites. The MFI ratios generated for cells treated with both vinblastine and rhodamine 123 were 9 (human P‐gp) and 24 (canine P‐gp). Additionally, the fluorescent H‐site probe Hoechst 33342 itself competed with rhodamine 123 and with calcein AM for canine and human P‐gp mediated efflux at the R‐ and P‐site, respectively. The MFI ratios for MDCK‐PGP were 10.8 and 1.4 for the P‐ and R‐site, respectively indicating substantial overlap of the canine H‐ and P‐sites and some overlap of the canine H‐ and R‐sites. MFI ratios for hMDR1 MDCK were 6.6 and 3.1 for the P‐ and R‐site, respectively, indicating moderate overlap of the human H‐site with both the P‐ and R‐site.

Assessment of Experimental Drugs as MRP1 Substrates

3.5

After observing low levels of MRP1 expression by hMDR1‐MDCK and MDCK‐PGP cells, it was important to determine if experimental drugs were also substrates for MRP1 since the P‐site probe calcein AM is an MRP1 substrate. Table 4 shows that when MRP1 inhibitor MK 571 (positive control) is co‐incubated with the MRP1 fluorescent substrate CFDA, a high MFI is achieved for both hMDR1 cells (9.3) and MDCK‐PGP cells (37). However, none of the experimental drugs, when co‐incubated with CFDA, generated an MFI ratio greater than 2, indicating that MRP1‐mediated efflux of these drugs is non‐existent (MFI < 1.5 for cisplatin, cyclosporine, progesterone, ivermectin, PSC‐833) or negligible at most (MFI < 1.6 for vinblastine, digoxin for canine P‐gp, loperamide, or vincristine).

Discussion

4

For dogs with intrinsic, or genetically mediated, P‐gp deficiency, knowing a drug's P‐gp status enables veterinarians to select an alternative drug or adjust the drug dosage and alerts the pet owner to be more diligent in monitoring for adverse events. This knowledge would benefit 75% of Collies, 50% of Australian Shepherds and English Shepherds, and smaller percentages of other breeds. Knowing a drug's status as a P‐gp substrate could prevent adverse drug reactions resulting from acquired P‐gp deficiency in ABCB1 wildtype dogs. Collectively, knowing a drug's status as a canine P‐gp substrate has the potential to benefit 100% of the canine population. Additionally, knowing whether a drug competes for binding to the canine P‐gp R‐site, P‐site or both may provide veterinarians important information for preventing drug–drug interactions. For example, if two P‐gp substrate drugs are co‐administered, and both are R‐site substrates, one would expect competition for P‐gp mediated efflux at the blood–brain barrier and/or biliary canaliculi potentially resulting in an adverse drug–drug interaction. Alternatively, if one of the two concurrently administered drugs binds exclusively at the R‐site and the other binds exclusively at the P‐gp P‐site, they may not compete for P‐gp mediated efflux and would thus not be expected to cause a drug–drug interaction. Further studies would be necessary to confirm if this is indeed the case.

Guidance documents from the U.S. Food and Drug Administration (FDA) and European Medicines Agency (EMA) recommend that test articles intended for human drug development are assessed as P‐gp substrates (Prueksaritanont et al. 2013). It is remarkable that this guidance is in place for humans when only one report exists of an adverse drug reaction resulting from an ABCB1 mutation associated with a P‐gp null phenotype (Baudou et al. 2020), while no recommendations regarding P‐gp substrate status exist for canines, a species with a well‐known genetically mediated P‐gp null phenotype. Granted, the rationale for assessing human drug candidates as P‐gp substrates involves drug–drug interactions and not pharmacogenetics. However, dogs (and cats, for that matter) are also at risk for P‐gp mediated drug–drug interactions. One might argue that human P‐gp substrate data resulting from FDA and EMA guidance could be applied to canine drug candidates. Previous studies and the data reported in the current study indicate otherwise.

In agreement with results reported here, several other groups (Kido et al. 2022; Katoh et al. 2006; Yamazaki et al. 2001; Zolnerciks et al. 2011) have reported species differences in P‐gp substrate efflux capacity. The current study assessed several drugs as canine and human P‐gp substrates at two different P‐gp binding sites (R‐ and P‐site). The H‐site was not explored further because, in the 25 years since the H‐site was discovered, the only two drugs that have been reported to interact with the H‐site are vinblastine and colchicine, and the former has a stronger interaction with the R‐site than the H‐site. While digoxin and progesterone had similar MFI ratios for both canine and human P‐gp, some important species differences were identified. Based on higher MFI ratios cyclosporine (R‐site) and vincristine (P‐site) are stronger canine than human P‐gp substrates. Both drugs are documented to cause serious adverse effects in dogs with P‐gp deficiency (Mealey et al. 2008; Mackin et al. 2020). Relying solely on human P‐gp substrate assessment, one would likely have underestimated both the likelihood and severity of potential adverse reactions associated with vincristine and cyclosporine. Thus, the results of the study reported here add to the growing body of evidence that P‐gp substrate status differs between species and support the recommendation that the target species' P‐gp be used to assess P‐gp substrate status of drug candidates.

This study is the first to examine the R‐, H‐, and P‐site of the canine P‐gp binding pocket. Results demonstrate two apparently independent binding sites for some P‐gp substrates (digoxin, progesterone, vincristine) and overlapping binding sites for others (cyclosporine, ivermectin, loperamide, vinblastine). The clinical implication of this finding is that P‐gp mediated drug–drug interactions are likely to be more severe when both drugs compete for the same binding site. This implies that it is important to determine not only if a drug candidate is a P‐gp substrate, but at which binding site(s) in order to make accurate predictions of potential drug–drug interactions.

A consensus regarding the structure of the human R‐site and P‐site binding sites has not been reached. However, in a modeling of human P‐gp binding to rhodamine 123 and prazosin, a P‐site substrate, 12 and 24 amino acids, respectively, were proposed to be involved in ligand–protein interactions (Pajeva et al. 2013). Interestingly the specific 36 human P‐gp amino acids indicated in that report are identical to those in canine P‐gp, suggesting similarities between the species. This is not surprising given the overlap of canine and human P‐gp substrate status that we identified in our study. However, it is important to point out that while the putative 36 amino acids were identical (12 for rhodamine 123 and 24 for prazosin), differences in neighboring amino acids could affect the tertiary structure of the binding pocket leading to species differences in binding affinity.

The exact conformation and structural relationship between the R‐, H‐, and P‐site of P‐gp and their binding of canine P‐gp substrates is not known. Based on data generated in our experiments and information reported by others, we have proposed a schematic diagram of the functional relationship of P‐gp's R‐, H‐, and P‐sites as a starting point from which to further define canine P‐gp's binding pocket (Figure 4). Notably, the H‐site does not appear to have a great deal of clinical relevance but was included in this study for completeness. Besides Hoechst 33342, the only other drugs that have ever been reported to bind the H‐site are vinblastine and colchicine (Shapiro and Ling 1998; Shapiro et al. 1999). The former is a strong substrate at the R‐site (Table 2) and the latter is rarely, if ever, used clinically. MFI ratios from the current study indicate that vinblastine is actually a stronger competitive P‐gp substrate at the R‐site than the H‐site, raising further questions about the H‐site's clinical relevance. Our study confirms the overlap of the human R‐ and H‐site. Additionally, the MFI ratios achieved when the non‐competitive inhibitor PSC 833 is co‐incubated with Hoechst 33342, the compound the H‐site was named for, are not impressive relative to those achieved when PSC 833 is co‐incubated with either rhodamine 123 or calcein AM. Collectively these data suggest that canine P‐gp substrate assessment should interrogate both the R‐site and P‐site but not the H‐site.

Suggested functional binding relationship of R‐, H‐, and P‐sites within the canine P‐gp binding pocket based on data generated in the competitive P‐gp efflux assay.

It is important to note that the P‐site fluorescent probe calcein AM is a substrate not only for P‐gp but for MRP1 as well. Although both are ABC transporters, MRP1 and P‐gp differ substantially in their biochemical and pharmacological properties as well as their physiological functions. MRP1 and P‐gp share only 23% sequence identity and MRP1 is limited to one ATP hydrolysis site while P‐gp has two (Leslie et al. 2005). The functional role of P‐gp is mostly limited to the extrusion of xenobiotics. MRP1 exports both endobiotics and xenobiotics and thus influences physiological processes other than drug distribution (Cole 2014). The chemical structure of their substrates is also different. P‐gp effluxes hydrophobic substrates, whereas MRP1 primarily effluxes organic anions, usually when conjugated with glucuronic acid, glutathione, or sulfate (Johnson and Chen 2017). Nevertheless, MRP1 efflux of experimental drugs was assessed via competition with the fluorescent MRP1 substrate CFDA to detect any potential contribution of MRP1‐mediated efflux. Based on MFI efflux ratios of < 1.2 cisplatin, cyclosporine A, progesterone, digoxin for human P‐gp, ivermectin, and PSC‐833 are not effluxed by MRP1. Vinblastine and digoxin for canine P‐gp, and loperamide, and vincristine, with MFI efflux ratios between 1.3 and 1.6, undergo negligible efflux by MRP1. Thus, it is reasonable to interpret the MFI efflux ratios for P‐gp's P site that were determined using calcein AM as having been generated by P‐gp‐mediated efflux.

Several potential weaknesses of this study should be mentioned. First, only three concentrations per drug were evaluated. The concentrations used in the current study were based on data from human or rodent cell‐based studies demonstrating P‐gp mediated efflux and incorporated a C max value from pharmacokinetic studies, if available. It is possible that different concentrations would yield different results. Another potential weakness is that the cell lines used are adherent cells, not suspended cell cultures. It is possible that adherent cell lines could have lower basolateral P‐gp expression than cells grown in suspension. This seems unlikely to be a problem for the cell lines included in this study based on robust P‐gp expression observed in western blots and the fact that cells were not allowed to grow to confluency. Lastly, using cell lines in which MRP1 was knocked out would eliminate any potential concerns regarding MRP1‐mediated efflux.

In conclusion, knowing a drug's canine P‐gp substrate status for products intended for canine use provides valuable information for veterinarians and pet owners to improve drug safety. Use of data from other species, that is, human P‐gp substrate data, as a surrogate for canine P‐gp substrate assessment is less than ideal based on data from this study and others. Results from this study suggest that determining whether a drug binds to the R‐ or P‐site might further inform veterinarians about potential P‐gp‐mediated drug interactions. Thus, additional studies to refine the competitive assay approach as a potential screening test for identifying canine P‐gp substrates are warranted.

Author Contributions

Katrina L. Mealey: conceptualization, experimental design, data analysis, statistics, writing and editing manuscript. Neal S. Burke: experimental design, conducting experiments, data analysis, writing and editing manuscript.

Conflicts of Interest

One of the authors (K.L.M.) received royalties from WSU for canine and feline MDR1 genotyping.