Influence of Proteome Profiles and Intracellular Drug Exposure on Differences in CYP Activity in Donor-Matched Human Liver Microsomes and Hepatocytes

Human liver microsomes (HLM) and human hepatocytes (HH) are important in vitro systems for studies of intrinsic drug clearance (CLint) in the liver. However, the CLint values are often in disagreement for these two systems. Here, we investigated these differences in a side-by-side comparison of drug metabolism in HLM and HH prepared from 15 matched donors. Protein expression and intracellular unbound drug concentration (Kpuu) effects on the CLint were investigated for five prototypical probe substrates (bupropion–CYP2B6, diclofenac–CYP2C9, omeprazole–CYP2C19, bufuralol–CYP2D6, and midazolam–CYP3A4). The samples were donor-matched to compensate for inter-individual variability but still showed systematic differences in CLint. Global proteomics analysis outlined differences in HLM from HH and homogenates of human liver (HL), indicating variable enrichment of ER-localized cytochrome P450 (CYP) enzymes in the HLM preparation. This suggests that the HLM may not equally and accurately capture metabolic capacity for all CYPs. Scaling CLint with CYP amounts and Kpuu could only partly explain the discordance in absolute values of CLint for the five substrates. Nevertheless, scaling with CYP amounts improved the agreement in rank order for the majority of the substrates. Other factors, such as contribution of additional enzymes and variability in the proportions of active and inactive CYP enzymes in HLM and HH, may have to be considered to avoid the use of empirical scaling factors for prediction of drug metabolism.

Further, the majority of the proteins quantified with at least three unique peptides across the HL, HH, and HLM (3075 of 3989) were found in all three sample types (Figure 2c).Despite the large overlap in number of quantified proteins, the protein levels in the HLM differed from both the HL and HH with absolute average fold differences (AAFD) of 3.0 and 3.1, respectively (Figure 2d).As indicated from the Proteomaps, HLM proteins with significantly higher concentrations (Figure S1a; Volcano-plot: FDR = 0.01, S0 = 2) were involved in pathways associated with the ER, such as fatty acid and drug metabolism (Data S2).This is in line with that HLM are suggested to be vesicles derived from the ER [2].
An in-depth investigation of the fractional contribution (total protein content %) of proteins from different subcellular locations [3] confirmed that ER-annotated proteins made up a larger proportion of the total protein content in the HLM (19%) than those in the HL (7%) and HH (9%; Figure 2e).However, the HLM also contained large proportions of proteins associated with other subcellular compartments, including mitochondria (14% of the total HLM protein content, compared to 17% and 21% in HL and HH, respectively).Furthermore, the HLM contained a comparable proportion of cytosolic proteins (27%) as to those in the HL (27%) and HH (32%).This was surprising since the cytosolic proteins are assumed to be discarded during the HLM preparation [4].The HLM also contained a large proportion of nuclear proteins (19%) that are expected to be captured in the normally discarded pellet obtained in the first low-speed centrifugation step [4][5][6][7] (Figure 2e).The fraction of nuclear proteins in HLM were comparable to that in the discard pellets (20%), and not much lower than either HL (28%) or HH (24%; Figure S1b-d).This demonstrates that the HLM fractions are "contaminated" with many proteins that are not associated with the ER compartment, an observation that is supported by previous investigations [8,9].
Similar to what was observed for the complete set of quantified proteins, most of the proteins classified as ER-localized were found at substantial concentrations in the three sample types (HL, HH, and HLM) and the discard pellet (Figure S1e-h).The ER-related proteins were in general enriched 2.8-fold (Figure 2f) in the HLM with a large variability in enrichment ranging from 0.03-120-fold across the different proteins.Since ER-associated proteins constituted 7% of the total protein content in our HL, assuming complete isolation and full recovery of the ER-fraction in the HLM, we would expect a 14-fold enrichment of microsomal ER proteins.The lower enrichment of ER-proteins was also reflected in the 2.2 to 4.6-fold enrichment of the ER-membrane markers, CANX and POR, which were similar or slightly lower than previous estimates (Figure S1i) [10].The traditional ER-activity markers glucose-6phosphatase (G6PC) and HMG-CoA reductase (HMGCR) [11], were also enriched to the same extent (3.4 to 7.4-fold) in the HLM (Figure 2g).Furthermore, we found specific membrane markers for other organelles, e.g., lysosome, peroxisome, and plasma membrane, in the HLM, where the enrichment were in agreement with previous findings (Figure S1i) [10].The variability in enrichment of different ERrelated proteins (0.03 to 120-fold) demonstrates the complexity of the enrichment process, which limits the use of specific protein markers as scaling factors for recovery of ER-related proteins.

Enrichment of CYP enzymes and the effect on metabolic clearance
The isolation of ER-localized proteins in liver HLM should lead to enhanced levels of membrane bound drug metabolizing enzymes, such as CYPs [12].The median concentrations of the CYP enzymes in the HLM ranged from 1.3 (CYP2J2) to 76.5 (CYP2C8) fmol/µg protein, and are comparable or higher than previously reported concentrations in HLM (Figure S2a) [9].The CYP enzymes in HLM were enriched on average 3.2-fold (range 1.2 to 56) compared to CYP levels in the HL and HH (HL and HH CYP levels that were in good agreement with that previously reported; Figure S2b-c [13][14][15]).Interestingly, two of the enzymes from the subfamily CYP2C (CYP2C9 and CYP2C19) were very differently enriched in the HLM.CYP2C9, with higher HL concentrations, was poorly enriched (AFD 1.4-fold) in the HLM, while CYP2C19, with lower HL concentrations, had approximately 50-fold higher concentration in the HLM compared to both HL (56-fold) and HH (46-fold) (Figure 2a).In line with this, global analysis showed that proteins with higher initial HL concentrations were less enriched in the HLM compared to those with lower concentrations (r s = -0.33; Figure S3e), suggesting a saturation in the enrichment process.
Since the liver HL, HH, and HLM were obtained from 15 matched donors, we were also able to investigate the preservation of the relative expression of the different enzymes across the donors from the three sample types.In general, the donor rank orders for the different CYPs were highly correlated with median Spearman's correlations (r s ) of 0.87 (Figure 3a; significant Spearman's rank correlations > 0.7, p < 0.006, after Bonferroni correction for multiple comparisons) across the three sample types (range across all CYPs and sample types r s = 0.01 to 0.99).However, some enzymes of great importance for drug metabolism, including CYP3A4 and CYP2C19, displayed weaker rank correlations between HLM and HH (r s = 0.48 and r s = 0.66, respectively), which could give rise to differences in metabolic clearance.

Enrichment of other drug metabolizing enzymes in human liver microsomes
The UGT enzymes were in general not as highly enriched in the HLM as the CYP enzymes, with respective median fold-enrichments of 2.6 (range: 1.3-4.3)and 2.4-fold (range: 0.5-5.8)compared to the levels in the HL and HH.The inter-donor spread of the UGT enzymes were similar in all three sample types with median values of 3 (range: 2-14) in HL, 3 (range: 2-28) in HH, and 3 (range: 2-17) in the HLM.Further, the rank order correlations across the 15 donors were not as well preserved for the UGT enzymes as for the CYP enzymes.The median Spearman's rank correlations were 0.76 (range r s : 0.45-0.81)between HLM and HL, 0.52 (range r s : 0.10-0.83)between HLM and HH, and 0.77 (range r s : 0.17-0.99)between HH and HH (Figure S3a).
The FMO enzymes were enriched to the same extent as the CYP enzymes in the HLM, with median fold-enrichments of 3.4 (range: 2.7-3.5) and 2.8 (range: 1.8-3.1)compared to HL and HH, respectively.High rank order correlations were obtained for the FMO enzymes with median Spearman's correlations of 0.93 (range r s : 0.86-0.94)between HLM and HL, 0.91 (range r s : 0.81-0.91)between HLM and HH, and 0.85 (range r s : 0.81-0.84)between HL and HH.Notably, FMO1 was only found in the HLM at low concentrations ranging from 0.004 to 0.2 fmol/µg protein (quantified with on average four unique peptides; Figure S3b).Additionally, many cytosolic enzymes involved in drug metabolism, such as aldehyde dehydrogenases (ALDHs), sulfotransferases (SULTs), and glutathione S-transferases (GSTs) were also found in the HLMs (Figure S3c, S3d).Although the concentrations were on average lower in the HLM, where the geometric mean was 5.9 fmol/µg protein (range across enzymes 0.01 -257.3) as compared to HL (geometric mean: 11.4; range: 0.02 -312.8) and HH (geometric mean: 10.7; range 0.01 -380.3)fmol/µg protein, several of the proteins were still quantified at high levels in the HLM.For instance, the phase I and II metabolizing enzymes, ALDH1A1 and GSTA2, respectively, had geometrical mean concentrations of 41.4 and 153.1 fmol/µg protein in the HLMs.

Figure S1 .
Figure S1.Global proteomics a) Fold difference of protein expression between different sample types (human liver microsomes, hepatocytes, and homogenates), against P-values, displayed in Volcano plots.Red circles mark proteins with significantly different expression.N denotes number of proteins with significantly different expression.b) Total protein content (%) of proteins from different subcellular compartments in human liver microsomes, hepatocytes, homogenates, and first low-speed centrifugation pellets.c) Protein concentrations of nuclear proteins in human liver microsomes and first low-speed centrifugation pellet, sorted by concentrations in the pellet.d) Number and overlap of nuclear proteins in (c).e) Protein concentrations of ER-related proteins in human liver microsomes, hepatocytes, homogenates, and first low-speed centrifugation pellets, sorted by concentrations in the homogenates.f) Number of ER-related proteins in (e).g) Enrichment of ER-related proteins in human liver microsomes, hepatocytes, and first low-speed centrifugation pellets, compared to in the liver homogenates.h) Comparison of median protein concentrations of hepatocytes, microsomes, and discard pellets against homogenate samples.i) Enrichment of membrane markers in the 15 microsomal samples against homogenate, compared with microsomal enrichment in the literature[10].Gray floating bars denote range, and the line denotes median across the 15 donors.

Figure S2 .
Figure S2.Protein expression of CYPs and UGTs in the a) 15 microsomal samples, compared with literature values [9]; b) 15 hepatocyte samples, compared with literature values [14]; and c) in the 15 homogenate samples, compared with literature values [13].Floating bars denote range, and the line denotes median across the different samples.

Figure S3 .
Figure S3.Protein expression of UGTs, FMOs, and cytosolic enzymes in liver homogenates, isolated hepatocytes, and liver microsomes from the 15 donors.a) UGT b) FMO c) cytosolic enzymes d) glutathione transferase.e) Fold enrichment of proteins in microsomes correlated to initial protein concentration in homogenates.Concentration levels are given in fmol/µg total protein in the respective system.Spearman's rank correlations compare the relative expression of each enzyme across the 15 donors between the respective sample type, where significant correlation coefficients are > 0.7 (P < 0.006, after Bonferroni correction for multiple comparisons).Average enrichment of microsomes compared to homogenates and hepatocytes was calculated based on concentrations from the 15 donors.

Figure S4a .
Figure S4a.Metabolic activity and protein expression.Correlations between CL int (eq.7) in HH of bufuralol, bupropion, diclofenac, midazolam, and omeprazole with CYP concentrations in the HH from 15 donors.

Figure S5 .
Figure S5.Drug accumulation and unbound fractions of drugs.Drug accumulation (Kp), fraction unbound (f u,cell ), and corresponding intracellular unbound drug concentration (Kp uu ) of a) midazolam, b) omeprazole, c) diclofenac, d) bupropion, and e) bufuralol in human hepatocytes from 15 donors.

Figure S6 .
Figure S6.Metabolic clearance of midazolam a) Metabolic clearance of midazolam, measured by depletion of midazolam (fraction remaining) and 1-OH-midazolam formation in HLM and HH from 15 matching donors.b) Amount of CYP3A4 in the incubations of the 15 donors in HLM and HH, respectively.

Figure S7 .
Figure S7.Metabolic clearance of omeprazole.a) Metabolic clearance of omeprazole, measured by depletion of omeprazole (fraction remaining) and 5-OH-omeprazole formation in HLM and HH from 15 matching donors.b) Amount of CYP2C19 in the incubations of the 15 donors in HLM and HH, respectively.

Figure S8 .
Figure S8.Metabolic clearance of bupropion.a) Metabolic clearance of bupropion, measured by depletion of bupropion (fraction remaining) in HLM and HH.b) Amount of CYP2B6 in the incubations of the 15 donors in HLM and HH, respectively.

Figure S9 .
Figure S9.Metabolic clearance of diclofenac.a) Metabolic clearance of diclofenac, measured by depletion of diclofenac (fraction remaining) and 4-OH-diclofenac formation in HLM and HH from 15 matching donors.b) Amount of CYP2C9 in the incubations of the 15 donors in HLM and HH, respectively.

Figure S10 .
Figure S10.Metabolic clearance of bufuralol.a) Metabolic clearance of bufuralol, measured by depletion of bufuralol (fraction remaining) in HLM and HH.b) Amount of CYP2D6 in the incubations of the 15 donors in HLM and HH, respectively.

Figure S11 .
Figure S11.Adjusting CL int,mic and CL int,hep with factors influencing metabolic clearance.a-e) Unbound in vitro clearance was scaled to kg body weight (bw) from HH with hepatocellularity and HLM with literature MPPGL (eq.7), for five probe CYP substrates (midazolam, omeprazole, diclofenac, bupropion, and bufuralol) in 15 donor-matched HLM and HH f-j) Unbound in vitro clearance was scaled to kg bw with specific CYP amount (midazolam -CYP3A4, omeprazole -CYP2C19, diclofenac -CYP2C9, bupropion -CYP2B6, and bufuralol -CYP2D6) in HH and HLM, respectively and CYP concentration in HL (eq.8).k-o) HLM clearance scaled with CYP amount was adjusted with Kp uu determined in HH (eq.9).r s = Spearman's rank correlation coefficient; AFD = Average fold difference.

Table S1 .
Characteristics of the 15 liver donors

Table S2 .
LC-MS/MS conditions in the compound quantification.

Table S3 .
Kinetics and proteomics data for the 15 donors

Table S3 .
Kinetics and proteomics data for the 15 donors cont.

Table S3 .
Kinetics and proteomics data for the 15 donors cont.