Fig 1.
Cooperation and collaboration between wet-lab and virtual experiments in improving mechanistic explanations of phenomena.
The workflow in each cycle has the same objective: challenge an explanatory hypothesis. Knowledge generated from right-side cycles is dependent on the combined strength of the four characteristics in Fig 3A. Knowledge gained from multiple right-side cycles can guide design of more efficient wet-lab experiments.
Fig 2.
Mouse Analog components and their organization (A) Mouse Analog comprises a Liver, Mouse Body, as well as a space to contain dose; the space enables simulating intravenous, intraperitoneal, and intragastric dosing. During execution, each discrete time step maps to 1 second. (B) A Liver comprises Monte Carlo-determined Lobule variants. (C) A Lobule comprises a directed graph with a concrete Sinusoid Segment (SS) object (a software agent) at each graph node. The Lobular configurations used herein were validated earlier; they are the result of cycling many times through the Iterative Refinement Protocol (described later) and successfully achieving several quantitative Target Attributes having stringent Similarity Criteria [9–12]. All flow paths follow the directed graph. Bile (dotted green) flows separately from blood (solid red) but is not a factor for the hypotheses tested herein. Periportal (PP) to CV gradients provided intra-Lobular location information to each Hepatocyte. (D) Each Sinusoid Segment configures a parsimony-guided multilevel variety of components so that during execution it functions as an analog of sinusoid components and features averaged across many lobules; Sinusoid Segment dimensions are Monte Carlo determined to mimic hepatic variability. Cell objects occupy most of Endothelial Cell (99%) and Hepatocyte (90%) spaces. APAP objects enter and exit a Sinusoid Segment via Core and Interface, percolate stochastically through accessible spaces influenced by configuration-controlled local flow, and, if not metabolized, exit to CV and return to Mouse Body. (E) Cells in Endothelial space control APAP entry and exit and contain a probability-specified number of Binders; for this work, we only required that they bind and release APAP. Hepatocytes use three previously validated event management modules [13], which control 1) material entry and removal, 2) binding and object transformations, and 3) up- and down-regulation of events such as Metabolism (not used for this work).
Fig 3.
Intra-Hepatocyte events and Analog–mouse relationships
(A) Experiments capable of challenging the NAPQI zonation hypothesis must demonstrate four characteristics. 1) Components are concrete and biomimetic. 2) Mechanism events during execution are observable and independent of phenomena being generated. 3) Qualitative and quantitative similarity (or lack thereof) can be established between target and Mouse Analog phenomena. 4) Means exist to incrementally strengthen claims that details of causal cascades in mice are strongly analogous [14]—quantitatively similar—to details of Mouse Analog’s causal cascade within and across multiple levels. (B) Virtual experiments designed to challenge the NAPQI zonation hypothesis focus on Metabolism Phase events and key early events within the Toxicity Phase of injury [1]. Although illustrated as a sequential cascade, each event executes independently in pseudo-random order each time step. All events are stochastic. Some event probabilities are Lobule location-dependent. An APAP object maps to a small fraction of an actual APAP dose, which for this work maps to 300 mg/kg. G&S objects represent APAP-glucuronide and APAP-sulfate plus all other inactive metabolites. A glutathione (GSH) Depletion event maps to depletion of a portion of a hepatocyte’s basal GSH. Mitochondrial Damage objects (mitoD) map to conflation of all influential damage products occurring within mitochondria [15]. Each mitoD may undergo one amplification event resulting in ≤ 6 additional mitoD; doing so enables downstream events to be finer grain than Metabolism Phase events. A mitoD Mitigation event maps to processes that advance recovery; it maps to an incremental reduction in mitochondrial disruption and damage.
Fig 4.
Location-dependent features (A) Only APAP metabolism configuration and the type of Metabolite formed are location-dependent in NZ-Mechanism. Target Phenomenon: pericentral necrosis begins first adjacent to CV and then moves (radially) outward in the PP direction. (B) GSH Depletion Threshold and probability of mitoD Mitigation event configurations are location-independent for NZ-Mechanism. However, one or both exhibit zonation (Methods) for GNZ-, MNZ-, and MGNZ-Mechanisms. (C) The relative locations of Hepatocytes are plotted in two different ways.
Fig 5.
Three of four plausible Mechanisms falsified
Shown are measurements from identical Mouse Analog experiments for which one of the four different Mechanism configurations was implemented. Average distances from CV (in Sinusoid Segment grid spaces) of Death trigger events each time step was recorded. Hepatocyte Death (plotted in Fig 6) becomes detectable Death Delay hours following its Death trigger event. Values shown are 100-second moving averages to reduce the considerable variability within and between simulation steps. Only MGNZ-Mechanism achieved the Target Phenomenon. Circled trends 1 and 2 help falsify NZ- and MNZ-Mechanisms.
Fig 6.
Cascading events within Lobules (A) During a Mouse Analog experiment that used MGNZ-Mechanism and started with a single toxic APAP dose (maps to approximately 300 mg/kg), measurements were made within the three illustrated 5-grid-space-wide regions: PP region adjacent to lobule entrance; CV region adjacent to CV; and Midzonal region in between. The experiment used 332 Monte Carlo variants of the same Mouse Analog. (B) Amounts in Mouse Body. The function of Extracellular Marker is analogous to that of an internal standard. It behaves the same as APAP, except that it is excluded from Cells and is not eliminated. The APAP profile during the experiment maps quantitatively to blood level profile in mice. G&S are transported out of Hepatocytes and accumulate in Mouse Body. Data in C and G-J are 100-second moving averages from the experiment described in A. (C) NAPQI profiles in each region reflect APAP profiles. APAP Blood levels adjacent to PP spaces are dramatically reduced as it distributes into the large number of accessible Hepatocytes. APAP in Blood in CV region partitions into far fewer Cells, so that per Cell concentration adjacent to CV at early times is actually greater than that in Cells adjacent to lobule entrance. (D-F) Histograms for number of Hepatocyte (Hep) Death events per second. Earliest Hepatocyte Deaths are seen at 1.2 hours after APAP dosing; Death trigger events occur earlier. (G) Amounts of G&S in each region. (H) Cumulative mean GSH Depletion events. (I) Mean amounts of mitoD: more than 5 mitoD per Hepatocyte triggers Death. Significant mitoD accumulation begins only after GSH Depletion. (J) Cumulative mitoD Mitigation events. Total Hepatocytes for 332 Monte Carlo variants: 4,648,000.
Fig 7.
Mechanism diversity mimics diversity of toxicity among genetically different mouse strains (A) Total Dead Hepatocytes (out of a total of 168,000) are plotted 24 hours after simulated oral APAP dosing using 64 plausibly biomimetic variants of MGNZ-Mechanism. Examples of the configuration used are shown in Supporting S3 Fig. Each variant represents a virtual mouse strain. Based on reported variability between mice in the same experiment, we specified that mean toxicity measurements for any variant that was within 20% (a judgment call) of mean values from MGNZ experiments (shaded green) could be determined experimentally indistinguishable. Inserts: Mechanistic consequences of changes to configurations are brought into focus by comparing Deaths per Zone rather than totals. (B) Shown are mean necrosis scores (left axis; n = 3–4/strain) data from Harrill et al. [7] along with one of the 64 Mouse Analog variants from A selected as described in the text. Their identifying Mechanism variant numbers are listed at the top.