Blood Cells, Hematocrit and Clotting

An endurance athlete can have a lower hematocrit than an untrained friend and more red cells. This lesson follows red cells from kidney signal to iron recycling, then how platelets and plasma proteins stop a leak without clotting the whole circulation.

Hematocrit is a ratio, not a cell count

Blood’s components are in blood vessels, blood and lymph. The hematocrit is the percentage of a volume of whole blood occupied by packed red cells after centrifugation.

hematocrit = red cell volume / (red cell volume + plasma volume)

Made-up round numbers for an untrained adult: 2.2 L of red cells and 2.8 L of plasma give 2.2 / 5.0 = 44 percent. Let training raise red cell volume 10 percent, to 2.42 L, and plasma volume 20 percent, to 3.36 L. Hematocrit falls to 2.42 / 5.78 = 42 percent, with more red cells. Reviews describe this pattern in trained athletes, so “sports anemia” is not anemia. Pregnancy also expands plasma more than red cells (pregnancy, birth and lactation). In the first days at altitude the change runs the other way: diuresis shrinks plasma, and hematocrit rises before red cell mass has changed much (control of breathing).

A concentration cannot give a mass; that needs plasma volume by indicator dilution (body fluid compartments) or hemoglobin mass by carbon monoxide rebreathing (exercise physiology).

Hemoglobin carries about 1.34 mL of oxygen per g (oxygen transport), but red cells also thicken blood: in vitro, viscosity rises steeply and nonlinearly with hematocrit, most of all at low shear, where red cells aggregate (doubling hematocrit raises it about 3- to 4-fold at high shear and almost 10-fold at low shear). Red cell deformability offsets part of that cost in microvessels. Oxygen delivery is flow times content, and a high hematocrit raises content while tending to lower flow.

The kidney senses oxygen, not red cells

The signal for red cell production is erythropoietin (EPO), a glycoprotein hormone made mostly by the adult mammalian kidney, though not by its tubules. In situ hybridization on anemic mouse kidneys (Lacombe and colleagues, 1988) found Epo mRNA in peritubular cells of the cortex and outer medulla, not in glomerular or tubular cells. They judged the cells most likely endothelial; later work identified them as fibroblast-like peritubular interstitial cells.

These cells read local tissue PO2, not the red cell count. They sit in a relatively hypoxic zone, so a small fall in blood oxygen content, from anemia or low arterial PO2, lowers their oxygen enough to matter. The switch is HIF-2 alpha, a subunit of a hypoxia-inducible transcription factor. In well-oxygenated cells, prolyl hydroxylases that require oxygen, iron and 2-oxoglutarate mark it for destruction; when oxygen falls it accumulates and turns on the EPO gene. The machinery is in ubiquitin and proteolysis. In rodent kidneys, hypoxia raises EPO mainly by recruiting more producing cells; human data are limited.

In lowlanders exposed to hypoxia, plasma EPO activity peaks at about 48 hours, then falls while staying somewhat above baseline. EPO keeps late erythroid progenitors in the marrow alive and dividing, and release of reticulocytes (young red cells that still hold RNA) rises over the following days, so the hormone peaks long before red cell mass does. Platelets answer to thrombopoietin, the main regulator of megakaryocyte development. HIF-2 alpha also directly raises duodenal iron absorption. Hypoxia lowers hepcidin too, but mostly indirectly: the erythropoiesis that EPO drives sends signals, such as erythroferrone, that suppress it.

Red cells wear out and their iron comes back

Human red cells circulate about 120 days on average, but survival varies: in a biotin-labelling study of six people without diabetes, the mean age of circulating red cells ranged from 38 to 60 days, which for cells that all live the same time would mean lifespans of roughly 76 to 120 days. Aged cells are removed by macrophages, mainly in the red pulp of the spleen and in the liver; how red pulp macrophages choose which cells to take is not settled. Heme yields iron and bilirubin (absorption and the liver).

The iron leaves through ferroportin, the only known cellular iron exporter, and returns to the marrow on plasma transferrin; duodenal enterocytes export absorbed iron the same way. Hepcidin, a 25-amino-acid liver peptide secreted when iron stores rise and in inflammation, binds ferroportin. In cultured cells this drives ferroportin’s internalization and degradation, and structural work shows hepcidin also occludes the exporter’s central cavity. Too much hepcidin traps iron in macrophages (low plasma iron, the anemia of inflammation); too little lets iron flood in (overload).

Worked problem. Made-up round numbers for an adult human at steady state: blood volume 5.0 L; 5.0 x 10^12 red cells per L; hemoglobin 150 g per L; mean lifespan 120 days; 3.46 mg iron per g hemoglobin (4 x 55.85 = 223.4 g of iron per mol of a tetramer of about 64,500 g per mol). Assume production equals removal, every cell lives the same time, no blood loss, and all hemoglobin iron is recycled.

1. Total red cells = 5.0 x 10^12 per L x 5.0 L = 2.5 x 10^13 cells.
2. Production = 2.5 x 10^13 / 120 days = 2.08 x 10^11 cells per day; divided by 86,400 s per day, 2.4 x 10^6 cells per second.
3. Total hemoglobin = 150 g per L x 5.0 L = 750 g; turnover = 750 g / 120 days = 6.25 g per day.
4. Iron moved = 6.25 g per day x 3.46 mg per g = 21.6 mg per day. Reviews give 20 to 25 mg a day recycled by macrophages against 1 to 2 mg absorbed: over 90 percent of the marrow’s iron is recycled.
5. Iron in circulating hemoglobin = 750 g x 3.46 mg per g = 2,600 mg, about 2.6 g.
6. Perturbation: if hemolysis shortened lifespan to 30 days with hemoglobin held constant, production must rise fourfold, to 8.3 x 10^11 cells per day, and hemoglobin iron flux to 750 / 30 x 3.46 = 86.5 mg per day. If reticulocytes stay recognizable in blood for about one day, their share rises from about 1/120 = 0.8 percent to 1/30 = 3.3 percent.

Two highland peoples, two answers to thin air

In a lowlander at altitude, hemoglobin mass rises only gradually, about 1 percent per 100 hours of exposure in pooled data from athletes, so early hematocrit changes mostly reflect plasma volume. Does every highland population end up with high hemoglobin?

Beall and colleagues (1998) measured Tibetans and Bolivian Aymara at 3,800 to 4,065 m by the same technique, with samples screened to remove iron deficiency, abnormal hemoglobins and thalassemia. If high hemoglobin were the universal human answer to chronic hypoxia, both groups should be about equally elevated. Tibetan men averaged 15.6 g per dL and women 14.2, against 19.2 and 17.8 in Aymara men and women. The Tibetan distribution resembled a sea-level US sample. The authors attributed most of the variance within each population to genetic factors (0.86 in Tibetans, 0.87 in Aymara).

In 2010 Beall’s group and collaborators reported that in Tibetans at 4,200 and 4,300 m, variants at EPAS1, the gene for HIF-2 alpha, were associated with lower hemoglobin, about 0.8 g per dL lower in major-allele homozygotes than in heterozygotes. That region was later traced to interbreeding with Denisovans (human evolution).

Ruled in: two highland populations differ in hemoglobin, and in Tibetans lower values are associated with variants of the gene for EPO’s transcription factor. Not shown: that low hemoglobin is itself the trait selection acted on (the 2010 authors allowed another effect of the variants), how the variants change HIF-2 alpha signalling, or that either population is better adapted overall. Tibetan hemoglobin does rise above about 4,000 m. In the excessive erythrocytosis of some Andean residents, predict hematocrit, viscosity and oxygen content per liter of blood at a given saturation to rise, and resistance to flow to rise too.

A plug, a thrombin burst, then brakes

A cut small vessel first constricts, through reflex neurogenic mechanisms and local endothelin and platelet thromboxane A2. Under arterial flow, von Willebrand factor (VWF) then binds exposed collagen and the platelet receptor GpIb, tethering platelets. Activated platelets release ADP, acting on P2Y12, a Gi-coupled receptor (G protein-coupled receptors), and make thromboxane A2 from arachidonic acid through COX-1 and thromboxane synthase. Both recruit platelets and switch the integrin GpIIb/IIIa to bind fibrinogen, which bridges neighbouring platelets. Intact endothelium nearby releases prostacyclin, which inhibits aggregation.

The cell-based model (Hoffman and Monroe, 2001) places coagulation on cell surfaces in three overlapping stages. In initiation, tissue factor on cells beneath the endothelium meets blood when the wall breaks, binds factor VIIa and calcium, and activates factors X and IX; Xa with Va makes a little thrombin, and tissue factor pathway inhibitor (TFPI) soon shuts this phase off. In amplification, that thrombin activates platelets and factors V, VIII and XI. In propagation, on the activated platelet surface, IXa with VIIIa (tenase) makes Xa and Xa with Va (prothrombinase) makes a burst of thrombin, which cleaves fibrinogen to fibrin and activates factor XIII to cross-link it.

The old “intrinsic” and “extrinsic” pathways survive as two in vitro tests, not parallel routes in the body: the prothrombin time starts plasma clotting with tissue factor, the aPTT with an artificial contact activator, through factor XII.

Brakes keep the burst local. Antithrombin inactivates thrombin and factors IXa, Xa, XIa and XIIa. On intact endothelium thrombomodulin binds thrombin and turns it into an activator of protein C, which then inactivates Va and VIIIa. Later plasmin dissolves fibrin. Endothelial tPA makes plasmin from plasminogen docked on fibrin lysines, and fibrin potentiates tPA about 500-fold, so lysis concentrates on the clot. Plasminogen activator inhibitor restrains tPA; alpha-2 antiplasmin restrains free plasmin.

Vane’s 1971 finding that aspirin blocks prostaglandin synthesis is in fatty acid synthesis. But aspirin can acetylate many proteins, so did it block cyclo-oxygenase by covalently modifying that enzyme? Roth, Stanford and Majerus (1975) incubated microsomes from sheep and bovine seminal vesicles, a rich source of the enzyme, with aspirin whose acetyl group carried tritium, and separated the proteins by SDS gel electrophoresis. Nonspecific acetylation would spread label over many bands. Instead one protein of about 85,000 molecular weight took it up, with the same time course and aspirin concentration dependence as the loss of cyclo-oxygenase activity, and a corresponding protein was labelled in aspirin-treated human platelets. That ruled in aspirin as an active-site acetylating agent, a covalent and so irreversible block. It did not show the antiplatelet effect itself, which the authors said the acetylation may account for, nor the residue (later work places it at serine 529 or 530, depending on species and numbering), nor anything about bleeding.

A platelet has no nucleus and makes little protein, so the block lasts the platelet’s life, and function returns only as new platelets arrive.

Platelet lifespan L 8 days (made up for this sketch; the lesson gives none) Fraction blocked at time zero, b 100 %

Kind of block

irreversible, covalent (aspirin on COX-1; clopidogrel’s active metabolite on P2Y12) reversible, fades as the inhibitor clears (half-life 12 h, made up for this sketch)

Swap platelets for red cells (120 days, the lesson’s average)

Working at time zero

0 %

Half of the lost function back

day 4

All of it back

day 8

In this sketch: steady state, constant production, cells removed at a fixed lifespan so their ages are spread evenly, and no drug left to block new cells. Made up for the sketch: the platelet lifespans, the blocked fractions and the 12-hour half-life. “Target” means COX-1 for aspirin and the P2Y12 receptor for clopidogrel. The 120-day red cell figure is the lesson’s average, and the lesson notes that red cell survival varies between people. The straight-line and 12-hour half-life curves are this sketch’s own rules, not data.

Rule to keep: a platelet has no nucleus and makes little protein, so an irreversible block lasts its life and function returns only as new platelets arrive; a reversible block fades as the inhibitor clears. A model of recovery of a blocked target (an enzyme or a receptor), not of bleeding; how many working platelets are enough is not shown.

Each drug class has one main molecular target, though some targets touch several clotting proteins. Aspirin acetylates COX-1. Clopidogrel’s active metabolite binds P2Y12 covalently and irreversibly. Heparin, through a specific pentasaccharide, accelerates antithrombin about 1000-fold. Warfarin inhibits vitamin K epoxide reductase, so new factors II, VII, IX and X lack the gamma-carboxylation that lets them bind calcium. The anticoagulant proteins C and S are vitamin K-dependent too, and protein C falls early; the full effect waits days while existing factors are cleared. Direct factor Xa and thrombin inhibitors occupy their enzyme’s active site.

Hemophilia tests the model

Hemophilia A (factor VIII) and B (factor IX) are X-linked recessive, so they mainly affect males (sex linkage and beyond).

Platelets are normal, so a plug forms. Tissue factor with VIIa still makes a little Xa and thrombin, so the prothrombin time is normal. But TFPI soon ends initiation, and Xa leaving the tissue factor-bearing cell is quickly inhibited, so the platelet surface depends on IXa-VIIIa for its Xa. Without either factor the burst falls short, fibrin forms late as a loose network of thick, sparse fibers, and the plug does not hold, fitting the joint and muscle bleeding of severe disease. The aPTT, which runs through IX and VIII, is prolonged. A two-pathway scheme cannot explain this: with parallel routes, the intact extrinsic one should suffice.

Factor XII deficiency prolongs the aPTT but causes no abnormal bleeding: factor XII drives the test’s contact activation but is not required for hemostasis. Severe factor XI deficiency usually causes milder bleeding, mostly after surgery or injury, though it varies and can be serious. That fits XI as an amplifier that thrombin can activate on platelets, though how XI is activated in the body is not settled. Von Willebrand factor deficiency impairs adhesion and also lowers factor VIII, which VWF carries. Emicizumab, a bispecific antibody that bridges IXa and X in place of VIIIa, was first approved in the US in 2017: the cell-based model put to work.

Design problem

In a study modelled on Renne and colleagues (2005), factor XII-null mice are compared with littermates. Illustrative results: the null mice have a prolonged aPTT, the same tail-bleeding time, and fewer occluded arteries after a chemical injury to the artery wall, and they are protected from thromboembolism induced by injected collagen plus epinephrine. What do these results rule in and out, and what would you do next?

One way to do it

Tail bleeding measures hemostasis, stopping blood loss from a cut; occlusion and thromboembolism measure thrombosis, an occlusive clot inside a vessel. Normal on the first and protected on the second, in this model factor XII contributes to thrombosis but not hemostasis.

The long aPTT does not contradict this: the assay is started by contact activation on an artificial surface, which requires factor XII, while a cut exposes tissue factor. As predicted, humans lacking factor XII are not reported to bleed abnormally.

Next, a rescue: restore factor XII to null mice, for example by infusing the purified protein. If occlusion returns, the missing protein caused the defect, not a linked gene or a developmental effect. Use littermates, blind scoring and two injury models.

The same reasoning, with the usually milder bleeding of human factor XI deficiency, made factor XI a target for new anticoagulants.

What it cannot show: mouse injury models are not human thrombosis, and the result says nothing about safety in people.