One - DARING TO THINK DIFFERENTLY
Bullmore opens with the central problem: depression is common and disabling, yet psychiatry still cannot usually tell an individual patient why they are depressed or predict which treatment will work. He contrasts the familiar “serotonin imbalance” story with the absence of a clinical serotonin measurement. His recurring example is “Mrs P,” a woman with severe rheumatoid arthritis who also had low energy, anhedonia, disturbed sleep, guilt and pessimism. In 1989 the standard interpretation was psychological: of course she was depressed because she knew she had a painful chronic disease. Bullmore’s alternative is that the inflammatory disease itself could directly generate depressive symptoms.
The chapter introduces the causal claim the rest of the book tries to build: peripheral inflammation can signal to the brain and alter mood, cognition and behavior. Cytokines are the key messengers. Bullmore rejects the old picture of the blood-brain barrier (BBB) as an impermeable wall separating immune system and brain; cytokine signals can reach or communicate with the brain. He distinguishes association from causation and lays out the required evidence: inflammation should precede depression, there should be plausible biological steps linking immune activation to altered brain function, and manipulating inflammation should alter depression-like outcomes.
Several results preview that case. A longitudinal study of about 15,000 children found that children with higher inflammation at age nine were more likely to be depressed at 18. In animals, injecting bacteria - or cytokines without the germ itself - causes “sickness behavior”: social withdrawal, reduced activity, disturbed sleep/eating and other depression-like changes. In humans, healthy volunteers given typhoid vaccine showed a cytokine increase, mild depressive symptoms, and altered activity in emotion-related brain regions. Bullmore therefore proposes a continuous physical chain from immune challenge to cytokines to brain changes to depressive experience, while explicitly acknowledging that gaps remain. He also introduces an evolutionary question: if inflammation can cause depressive behavior, why would natural selection preserve that coupling?
Two - THE WORKINGS OF THE IMMUNE SYSTEM
This chapter supplies the immunology needed for the later argument. Inflammation is presented as an evolved defense of “self” against “non-self.” Macrophages are widely distributed front-line immune cells that engulf pathogens and release cytokines. Lymphocytes provide more specific adaptive responses, including antibodies and immune memory. Cytokines act as broadcast signals: a macrophage encountering infection can release them into circulation, recruiting and activating other immune cells. Lymph nodes provide hubs where macrophages carrying antigen fragments interact with lymphocytes, allowing the immune system to recognize particular threats and build a targeted response.
Bullmore emphasizes that inflammation is useful but potentially destructive. The same machinery that kills pathogens can damage host tissue - “collateral damage” - and immune regulation therefore matters as much as immune activation. Vaccination exploits immune learning: exposure to an antigen generates a response that makes later encounters faster and more effective. Autoimmunity represents a failure of self/non-self discrimination, in which immune mechanisms attack the body’s own tissues.
The neuroscience relevance is structural: the immune system is not a single organ isolated from the brain. It is distributed throughout the body, communicates chemically through cytokines, and includes macrophage-like cells in the brain (microglia). The chapter establishes the cellular vocabulary - macrophages, lymphocytes, cytokines, antibodies, endothelial cells and microglia - that Bullmore later uses to explain how an inflammatory event outside the brain can influence neural function.
Three - HIDING IN PLAIN SIGHT
Bullmore returns to patients such as Mrs P to argue that the co-occurrence of inflammatory physical illness and depression has been clinically obvious but conceptually discounted. In rheumatoid arthritis, he notes that fatigue is extremely common and depression is also common; conventional medicine has tended to treat the joint disease as biological while interpreting fatigue, low mood and “brain fog” as psychological reactions to being ill. His claim is not that psychological reactions are unreal, but that this explanation may miss a shared inflammatory mechanism.
Bullmore discusses the development of anti-TNF therapy for rheumatoid arthritis. Once TNF was identified as an important inflammatory cytokine, antibodies were engineered to neutralize it. In a randomized clinical trial, four weeks after assignment to high-dose anti-TNF, low-dose anti-TNF or placebo, 79% of the high-dose group responded, versus 44% of the low-dose group and 8% of placebo. This established that selectively interrupting one cytokine signal could dramatically alter an inflammatory disease.
Bullmore then asks why mental outcomes were not studied with the same intensity. Clinicians reported a rapid improvement in mood after anti-TNF infusion - the “Remicade high” - sometimes apparently occurring before major physical recovery. That observation is intriguing but not, by itself, proof of a direct antidepressant mechanism: expectancy/placebo effects and relief about treatment could explain it. Bullmore highlights this as an example of the book’s broader methodological point: the critical experiment must separate improvement in mood caused directly by reduced inflammation from improvement in mood secondary to feeling physically better.
Four - MELANCHOLIA AFTER DESCARTES
This chapter traces how depression became divided conceptually between mind and brain. Bullmore moves from ancient melancholia and humoral medicine through Cartesian dualism, Freud and psychotherapy, then into biological psychiatry and antidepressant drugs. His target is not psychotherapy or neuroscience themselves, but the persistent tendency to treat mental and physical explanations as competing domains.
The antidepressant history is especially important. Iproniazid was developed as an anti-tuberculosis drug after screening hydrazine-derived compounds in infected mice. In a 1952 TB trial at Sea View Hospital, patients became strikingly more energetic, sociable and hungry while their TB improved. The mood effect could have been interpreted as a consequence of recovering from a lethal disease, but it prompted direct psychiatric testing. In 1957, Nathan Kline’s group reported improvement in roughly 18 of 24 psychiatric patients treated for five weeks; Bullmore stresses how weak this evidence looks by modern standards - there was no placebo control and many participants had schizophrenia rather than depression.
Mechanistic work showed that early antidepressants increased monoamine signaling, leading to the catecholamine hypothesis and later a focus on serotonin. Lilly deliberately developed a selective serotonin reuptake inhibitor; its first small depression trial failed to beat placebo, but subsequent trials did, and fluoxetine/Prozac was licensed in 1987. Bullmore’s central criticism is an inference problem: a drug increasing a transmitter and sometimes improving depression does not establish that depression was caused by a deficiency of that transmitter. He notes that direct evidence for low brain serotonin as a general cause of depression never materialized.
That connects to his major criticism of psychiatry: lack of biomarkers. There is no routine serotonin test that diagnoses depression, identifies who has a serotonin abnormality, or predicts SSRI response. Blood serotonin has not reliably served that purpose; CSF sampling is invasive; specialized neuroimaging of serotonin-related targets is impractical and has not produced a clinical biomarker. Treatment therefore remains largely trial-and-error, and biologically heterogeneous patients are grouped under the same diagnostic label.
Five - HOW?
Bullmore first reviews case-control studies measuring cytokines or C-reactive protein (CRP). Across thousands of participants studied from the early 1990s through 2014, depressed groups had moderately but significantly higher concentrations of CRP and several cytokines than controls. In a Copenhagen population study of 73,131 people, higher CRP tracked with more depressive symptoms in a dose-response-like relationship; Bullmore reports the probability of that association occurring by chance as less than one in a trillion. He is careful that this demonstrates association, not that every depressed person is inflamed or that inflammation necessarily caused the depression.
Longitudinal data strengthen temporal inference. In the approximately 15,000-child cohort, those in the top third of cytokine levels at age nine were about 1.5 times as likely to develop depression by 18, despite not being more depressed at baseline. In a study of about 2,000 older UK civil servants measured in 2004, 2008 and 2012, persistent high CRP preceded first-onset depression; women with high CRP in both earlier assessments were about three times as likely to become depressed in 2012 as women without prior CRP elevation. Bullmore still treats these as supportive rather than conclusive because unmeasured processes could connect the distant measurements.
Interferon treatment provides a more direct human perturbation. Interferon intentionally drives inflammation in hepatitis treatment. Patients acutely develop sickness behavior, and after several weeks about one-third develop clinical depression despite not being depressed immediately before treatment. This is presented as unusually clear evidence that experimentally/therapeutically induced immune activation can precede and cause a depressive syndrome in humans.
The chapter then asks how peripheral inflammation reaches the brain. Bullmore describes several routes rather than literal free passage of immune cells through an absolutely open BBB: cytokine signaling at endothelial interfaces, communication at less restrictive regions, and neural signaling. The vagus nerve is central to an “inflammatory reflex”: it detects inflammatory signals and can feed back onto immune organs such as the spleen. In a rheumatoid arthritis study he describes, electrical vagus stimulation for 20 minutes per day rapidly reduced cytokines and disease symptoms; stopping stimulation for 10 days increased cytokines and symptoms, and restarting it reduced them again - an especially striking manipulation of brain-body immune communication.
At the systems-neuroscience level, Bullmore discusses emotional-face fMRI. Depression is associated with exaggerated activation of regions including amygdala and cingulate cortex to sad emotional stimuli, and successful SSRI treatment has been associated with reduced amygdala response. More directly, 20 healthy young volunteers were scanned after typhoid vaccination and after placebo injection. Vaccination increased circulating cytokines and mild depressive symptoms, increased cingulate activity, and altered emotional-network connectivity; stronger inflammatory responses and greater depressive symptoms were associated with larger network changes.
Animal experiments provide cellular resolution unavailable with fMRI. LPS injection produces acute sickness behavior; BCG vaccination can produce longer-lasting social withdrawal and reduced preference for sweetened water, used as a behavioral index related to anhedonia. Peripheral immune activation activates microglia in the brain. Activated microglia release cytokines and can damage neurons, reduce neurogenesis and synaptic plasticity, and alter neurotransmitter metabolism. One specific pathway highlighted is diversion of tryptophan away from serotonin synthesis toward kynurenine and related metabolites; this simultaneously reduces substrate available for serotonin and can generate neuroactive/toxic products. Bullmore uses this multilevel evidence - epidemiology, longitudinal cohorts, immune challenge, fMRI, animal behavior, cells and molecular pathways - to argue that a plausible causal chain now exists from peripheral inflammation to depressive changes.
Six - WHY?
Bullmore next asks why depression and inflammation would be coupled at all. He first considers sources of low-grade inflammation that overlap with depression risk. Adipose tissue contains many macrophages, so obesity increases inflammatory signaling and is also associated with depression; age increases both inflammation and depression risk; inflammatory activity also varies seasonally. He is cautious about causal direction for many of these associations.
Social stress is the major focus. Major life events involving loss, rejection or humiliation strongly increase depression risk; Bullmore cites estimates that roughly 80% of depressive episodes are preceded by a stressful life event, with especially large risks after interpersonal loss/rejection. Bereavement is associated not only with depression but physical disease and mortality; he cites evidence that recent bereavement approximately doubles risk of acute cardiovascular events. Studies reviewed in the chapter show that social stress can increase macrophage activation, cytokines and CRP; elevated inflammatory markers are reported in contexts including caregiving, poverty, isolation and childhood adversity. Animal social-defeat work supplies a mechanistic bridge, linking repeated social stress to peripheral immune changes, trafficking of monocytes and microglial activation.
Bullmore proposes a vicious cycle: stress can increase inflammation; inflammation can alter brain and behavior toward depression; depression can then increase social stress and behaviors that further sustain inflammation. The “why” answer is evolutionary and explicitly more speculative than the mechanistic evidence in Chapter Five. In ancestral environments where infection was a dominant mortality threat, genes producing stronger innate immune responses would improve survival. The same cytokine programs also produce sickness behaviors - withdrawal, reduced activity, appetite loss, altered sleep and reduced pleasure. Temporarily conserving energy and reducing social contact during infection could have protected the sick individual and perhaps reduced pathogen transmission. Traits that were adaptive under high infectious pressure may therefore persist even when modern triggers are psychosocial or chronic rather than acute infection.
The chapter also connects immune genetics to depression. Bullmore’s broader proposal is not a simple “gene for depression,” but pleiotropy: variants selected because they improved host defense could also increase inflammatory and depressive responses. Modern depression can therefore be understood partly as a mismatch or cost of ancient immune-defense programs, not as depression itself having been directly selected because severe depression improves reproductive fitness.
Seven - SO WHAT?
The final chapter asks what immuno-psychiatry changes clinically. Bullmore is notably cautious about what could be done in 2018. A CRP test can identify low-grade inflammation - he uses >3 mg/L as a practical threshold in an example - but there was no licensed anti-inflammatory treatment for depression. He explicitly warns against simply prescribing aspirin: evidence for antidepressant efficacy was insufficient and adverse effects such as gastrointestinal bleeding are real. He instead discusses identifying and treating sources of inflammation, while emphasizing that lifestyle recommendations are not yet a precision treatment for “inflamed depression.”
Existing trials provide suggestive repurposing evidence. Reanalyses/meta-analyses of placebo-controlled anti-inflammatory/anti-cytokine trials in tens of thousands of patients with inflammatory diseases found an average antidepressant effect size of about 0.4; Bullmore contrasts this with about 0.2 for SSRIs. But these studies were designed primarily to improve physical disease, and mood was often a secondary endpoint measured months later. Thus improved mood could be secondary to improved joints, skin or other physical symptoms rather than a direct anti-inflammatory brain effect.
The most important depression-specific test randomized 60 patients with treatment-resistant depression to an anti-TNF antibody or placebo. After eight weeks, both groups improved substantially and there was no significant average treatment-placebo difference. However, baseline inflammation mattered. Patients with higher pretreatment CRP showed a stronger antidepressant response to anti-TNF than those with lower CRP. Bullmore treats this subgroup result as the model for a new strategy: not an anti-inflammatory “panacea,” but biomarker-defined treatment for the subset of depressed patients whose illness is inflammatory.
He extends the same logic to vagus-nerve stimulation, which was already licensed for depression but whose mechanism remained uncertain; an anti-inflammatory mechanism via the vagal inflammatory reflex is presented as a testable alternative to the traditional monoamine explanation. The chapter also applies neuroimmune reasoning to Alzheimer’s disease and schizophrenia, emphasizing heterogeneity and the possibility that immune mechanisms matter only in subsets.
The book ends with a precision-medicine agenda: stop treating depression as one biologically uniform disease; develop better inflammatory and genetic biomarkers; test treatments in biologically selected patients; integrate mental and physical healthcare; and design trials that can distinguish direct immune effects on mood from indirect benefits of improved physical health. Bullmore presents this as a research program and prediction, not an established replacement for standard depression treatment.
That’s a similar message to the NIMH’s RDoC (Research Domain Criteria) initiative, which Insel helped launch while directing NIMH. The idea is don't necessarily start with “depression” as the unit of analysis at all. Start with underlying functional systems - reward processing, threat/fear, cognition, arousal, etc. - and study their genes, molecules, circuits, physiology, behavior, and symptoms across traditional diagnoses. Symptom categories can lump together people whose illnesses have different mechanisms.
Discussion Questions
1. Which evidence in Chapter Five most strongly supports causality rather than correlation: longitudinal CRP/cytokine studies, interferon-induced depression, the typhoid-vaccine fMRI experiment, vagus-nerve manipulation, or animal immune-challenge studies? What does each design still fail to establish?
2. In the 60-patient anti-TNF trial, the overall drug-placebo comparison was negative, but higher baseline CRP predicted a stronger drug response. How convincing is that as evidence for an “inflamed depression” subtype, and what would you require in a replication?
3. Bullmore repeatedly uses sickness behavior in rodents as a model for aspects of depression. Which features - social withdrawal, reduced activity, disrupted sleep/appetite, reduced sucrose preference - translate well to human depression, and where does the analogy break down?
4. Does the absence of a clinically useful serotonin biomarker undermine the serotonin hypothesis itself, the way SSRIs have been marketed/explained, or neither? What evidence would actually demonstrate a serotonin-defined subtype of depression?
5. The typhoid-vaccine study links an experimentally induced cytokine response to mild depressive symptoms and altered cingulate/network activity. How much can a transient inflammatory challenge in 20 healthy volunteers tell us about chronic major depressive disorder?
6. Bullmore’s evolutionary account argues that inflammatory sickness behavior may have been adaptive during infection and is maladaptive in some modern contexts. Which parts of this account are experimentally testable, and which remain evolutionary storytelling?
7. If you were designing the decisive clinical trial today, how would you define “inflamed depression,” which biomarker(s) would you use for enrollment, what treatment would you test, and what timing/outcomes would distinguish a direct antidepressant immune effect from simply feeling better because physical illness improved?
The SSRI story:
In the 1950s, the anti-TB drug iproniazid unexpectedly improved mood. Researchers discovered that it increased monoamine neurotransmission, and other early antidepressants did too. This led to the hypothesis that depression might result from too little monoamine signaling.
Eli Lilly then focused specifically on serotonin. The target seemed biologically plausible because serotonin-producing neurons in the brainstem project to areas disrupted in depression and regulate functions including sleep, appetite, and mood. So there was a plausible neural system connecting serotonin to depressive symptoms.
Lilly deliberately designed a drug to selectively block serotonin reuptake. This was appealing because it followed a rational drug-development sequence: proposed mechanism → specific molecular target → drug that selectively hits the target → clinical trial. Their first small trial failed, but later trials beat placebo, and fluoxetine/Prozac was licensed in 1987.
Prozac's success helped cement the simple story: low serotonin → depression; SSRI → restores serotonin → improvement.
But that last inference was the problem. Researchers never established that depressed people generally start with a serotonin deficiency. Knowing that serotonin regulates depression-relevant brain functions - and that manipulating it can help - doesn't prove that deficient serotonin caused the disorder.
SSRIs are modestly effective on average but can be much more effective for some individuals. Bullmore uses this to argue that depression is likely biologically heterogeneous rather than a single serotonin-deficiency disease.