Protein folding, targeting and modification

A polypeptide leaving the ribosome is not yet a working protein. It must fold into the right three-dimensional shape, reach the right place in the cell, and often be cut, decorated with chemical groups or joined to partners. Each of these steps can be regulated, and each can go wrong. This lesson follows a protein from the ribosome exit tunnel to its destination: how folding happens and how chaperones assist it, how sorting signals direct proteins to membranes, organelles and the outside of the cell, how post-translational modifications change protein function, and how unwanted or damaged proteins are destroyed. It is deliberately brief; BIOL-214 covers organelles and protein sorting in depth, and BIOC-310 covers protein structure in detail.

Sequence determines structure

In the late 1950s Christian Anfinsen, working with the enzyme ribonuclease A, showed that a protein unfolded by urea and a reducing agent, which broke its four disulfide bonds, could refold spontaneously into its active form when the chemicals were slowly removed. He concluded that the amino acid sequence contains all the information needed to specify the folded structure, which is the most stable arrangement under physiological conditions. This principle earned him a share of the 1972 Nobel Prize in Chemistry and underlies the modern prediction of protein structures from sequence. In 2024 the Nobel Prize in Chemistry recognized David Baker for protein design and Demis Hassabis and John Jumper for AlphaFold, a machine-learning system that predicts structures from sequences with high accuracy for most proteins.

Folding is driven largely by the hydrophobic effect: nonpolar side chains cluster in the interior, away from water, while polar and charged side chains face outward. Hydrogen bonds in the backbone create the regular secondary structures, α helices and β sheets, and these pack together into the tertiary structure. Many proteins also assemble from several polypeptide chains into a quaternary structure, as hemoglobin does from four.

Chaperones

Inside a cell, a growing polypeptide faces a crowded environment, with protein concentrations of 200 to 300 mg per mL. Partly folded chains expose hydrophobic surfaces that tend to stick to each other and aggregate. Molecular chaperones prevent this. Hsp70 proteins bind exposed hydrophobic stretches as chains emerge from the ribosome and release them in cycles driven by ATP, giving them chances to fold correctly. Chaperonins, such as the bacterial GroEL–GroES complex and the eukaryotic TRiC, form barrel-shaped chambers in which a single protein can fold in isolation. Many chaperones were first identified as heat-shock proteins, made in large amounts when cells are heated, because heat unfolds proteins and increases the need for help.

When folding fails

Misfolded proteins cause disease in two broad ways. Some mutations prevent a protein from folding properly so that it is destroyed before it can work. The most common cystic fibrosis mutation, ΔF508, deletes one phenylalanine from the CFTR chloride channel (BIOL-201); the protein misfolds, is retained in the endoplasmic reticulum and degraded. Drugs called correctors, combined with potentiators that improve channel opening, now help the mutant protein fold and reach the cell surface; the triple combination elexacaftor, tezacaftor and ivacaftor has greatly improved lung function for many people with cystic fibrosis. In the second category, misfolded proteins form toxic aggregates, as in Alzheimer, Parkinson and Huntington diseases, or are infectious, as with prions.

Sorting signals

Every protein begins synthesis on a ribosome in the cytosol, yet proteins end up in the nucleus, mitochondria, chloroplasts, peroxisomes, membranes or outside the cell. They are directed by sorting signals, usually short stretches of amino acids in the protein itself, a concept for which Günter Blobel received the 1999 Nobel Prize. Proteins without any signal stay in the cytosol.

The secretory pathway illustrates sorting at work. As a protein destined for secretion or for a membrane begins to emerge from the ribosome, its N-terminal signal peptide is recognized by the signal recognition particle (SRP), a ribonucleoprotein that pauses translation and brings the ribosome to the SRP receptor on the endoplasmic reticulum (ER) membrane. Translation resumes and the growing chain threads through a channel, the Sec61 translocon, into the ER lumen. A signal peptidase then usually removes the signal peptide. Membrane proteins contain additional hydrophobic segments that stop transfer and become membrane-spanning helices. From the ER, proteins travel in vesicles to the Golgi apparatus, where they are further modified and sorted to the plasma membrane, lysosomes or secretory vesicles.

Signal peptides share no exact sequence, but they share a pattern: a few positively charged residues near the start, a core of hydrophobic residues, and a polar region before the cleavage site. Hydropathy plots, which average the hydrophobicity of residues in a sliding window along a sequence, reveal such segments. The scale of Kyte and Doolittle (1982) assigns each amino acid a hydropathy value, from +4.5 for isoleucine to −4.5 for arginine.

# Kyte-Doolittle hydropathy in a sliding window of 9 residues
KD = {"A": 1.8, "R": -4.5, "N": -3.5, "D": -3.5, "C": 2.5, "Q": -3.5, "E": -3.5, "G": -0.4,
      "H": -3.2, "I": 4.5, "L": 3.8, "K": -3.9, "M": 1.9, "F": 2.8, "P": -1.6, "S": -0.8,
      "T": -0.7, "W": -0.9, "Y": -1.3, "V": 4.2}

# Human preproinsulin (110 residues); the first 24 form the signal peptide
preproinsulin = ("MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN")

w = 9
scores = [sum(KD[a] for a in preproinsulin[i:i+w]) / w for i in range(len(preproinsulin) - w + 1)]
peak = max(range(len(scores)), key=scores.__getitem__)
print("length:", len(preproinsulin), "residues")
print(f"most hydrophobic window: residues {peak+1}-{peak+w} ({preproinsulin[peak:peak+w]}), mean {scores[peak]:.2f}")
print("windows with mean > 1.6 (candidate hydrophobic segments):",
      [i + 1 for i, s in enumerate(scores) if s > 1.6][:12], "...")
print("signal peptide (residues 1-24):", preproinsulin[:24])

length: 110 residues
most hydrophobic window: residues 7-15 (LLPLLALLA), mean 2.76
windows with mean > 1.6 (candidate hydrophobic segments): [5, 6, 7, 8, 9, 10, 11, 35, 36] ...
signal peptide (residues 1-24): MALWMRLLPLLALLALWGPDPAAA

The most hydrophobic windows fall within the first 24 residues of preproinsulin, the signal peptide, confirming what the hydropathy pattern predicts. Software used in research combines such scores with machine learning to predict signal peptides and membrane-spanning segments across entire genomes.

Proteolytic processing: insulin

Insulin, the hormone whose discovery in Toronto by Frederick Banting, Charles Best, James Collip and John Macleod in 1921 transformed the treatment of diabetes, shows how cutting a precursor creates an active protein. The gene encodes preproinsulin, 110 amino acids. Removal of the 24-residue signal peptide in the ER produces proinsulin, 86 residues, which folds and forms three disulfide bonds. In secretory vesicles, proteases called prohormone convertases cut out the middle segment, the C-peptide, leaving two chains, A (21 residues) and B (30 residues), held together by two disulfide bonds. Mature insulin and C-peptide are secreted together, so measuring C-peptide in blood indicates how much insulin a person’s own pancreas is making, which helps doctors distinguish type 1 from type 2 diabetes.

In 1978, insulin became one of the first human proteins made by recombinant DNA technology, with the A and B chains produced separately in E. coli. Today recombinant insulins, including engineered analogues with altered absorption, have replaced insulin from pig and cow pancreas. Many other proteins are made as inactive precursors and activated by cutting, including digestive enzymes such as trypsinogen, blood clotting factors and the caspases that drive programmed cell death.

Post-translational modifications

More than 400 kinds of chemical modification of proteins are known. A few are especially widespread.

Phosphorylation deserves special mention. The human genome encodes about 520 protein kinases, and roughly a third of all human proteins are phosphorylated at some point. Adding a phosphate, with its two negative charges, can change a protein’s shape, create a docking site for other proteins or alter its location. Because kinases and phosphatases act quickly and reversibly, phosphorylation is the main currency of intracellular signaling. Many cancers are driven by kinases that are stuck in the on state, and kinase inhibitors such as imatinib, which targets the BCR–ABL kinase produced by the Philadelphia chromosome (BIOL-201 lesson 9), have become important cancer drugs.

Protein degradation

Cells constantly destroy proteins: damaged ones, misfolded ones and regulatory ones whose levels must change quickly. In the ubiquitin–proteasome system, a cascade of three enzymes (E1, E2 and E3) attaches chains of ubiquitin to target proteins, and the 26S proteasome, a barrel-shaped protease complex, recognizes the chains, unfolds the target and cuts it into short peptides. Humans have more than 600 E3 ubiquitin ligases, which provide specificity by recognizing particular targets. The discovery of ubiquitin-mediated degradation earned Aaron Ciechanover, Avram Hershko and Irwin Rose the 2004 Nobel Prize in Chemistry. Proteasome inhibitors such as bortezomib are used to treat multiple myeloma, a cancer of antibody-producing cells that rely heavily on the proteasome to cope with the huge amount of protein they make.

Protein half-lives range from minutes for regulatory proteins such as the cyclins that drive the cell cycle to months or years for structural proteins such as the crystallins of the eye lens and collagen. Protein levels depend on the balance between synthesis and degradation, just as mRNA levels do (lesson 5), and the same mathematics applies.

import math

def steady_level(synth_per_hour, half_life_hours):
    return synth_per_hour * half_life_hours / math.log(2)

print("Proteins made at 1,000 molecules per hour:")
for name, h in [("cyclin-like regulator", 0.5), ("typical enzyme", 20), ("long-lived structural protein", 500)]:
    print(f"  {name:30s} half-life {h:6.1f} h -> steady state {steady_level(1000, h):10,.0f} molecules")

# How fast does a protein disappear after synthesis stops?
h = 0.5
for t in (0, 0.5, 1, 2):
    print(f"t = {t:3.1f} h: {100 * 0.5 ** (t / h):5.1f}% of the regulator left")

Proteins made at 1,000 molecules per hour:
  cyclin-like regulator          half-life    0.5 h -> steady state        721 molecules
  typical enzyme                 half-life   20.0 h -> steady state     28,854 molecules
  long-lived structural protein  half-life  500.0 h -> steady state    721,348 molecules
t = 0.0 h: 100.0% of the regulator left
t = 0.5 h:  50.0% of the regulator left
t = 1.0 h:  25.0% of the regulator left
t = 2.0 h:   6.2% of the regulator left

Bringing it together

Lessons 5 to 8 have followed the flow of information from gene to working protein. In a eukaryotic cell, that path includes transcription with general factors and the CTD, capping, splicing and polyadenylation, export, scanning and initiation, elongation and termination, folding with chaperones, sorting, modification and eventually degradation. Every step offers the cell a point of control and every step can fail in disease. Lessons 9 and 10 turn to the first and most important control point: deciding which genes are transcribed.